Plurk paste

Part Three ENDGAME
16 Mors ex Machina Digging Up Death – From Human Exploitation to Human Extinction – Goliath Becomes Death – When Our Luck Runs Out – Reviving (and Upgrading) our Greatest Killers – Automated Cognition – The AI Accelerant DIGGING UP DEATH Your plastic flip-flops slip in the dirt as you swing the pickaxe. The earth is red, but down here deep in the bowels of Congo, it looks a shadowed, bloody brown. It’s dark and the walls seem to press around you. Those walls have always seemed closer, more ominous, ever since your oldest brother, Akile, died when a tunnel collapsed on him. Your other brother, Kijani, was pulled out but hasn’t walked since. You remember that day, the way the tears felt on your dusty face, that moment you realized that Akile’s smile wasn’t coming up from the ground ever again. The crying didn’t last long. You have to work, especially now that Kijani can’t.
Your stomach rumbles as you swing again. You skipped lunch and now your head aches with exhaustion as you pass your tenth hour of work. It seemed like a good idea at the time: you’ll make an extra dollar without the lunch break and by working overtime. It wasn’t a hard choice since lunch is just two tasteless bread rolls and a carton of sugary water that barely passes as juice. The air is heavy with black dust which sticks to your clothes and coats your throat. If you’re lucky, you might make a bit over $4 today.
1 The cobalt – that bright, blue-green-grey rubble that you smash out from the rocks a hundred metres below the surface – will be worth far more than your tears. After being carried by wheelbarrow to the truck it’ll be shipped to China for transformation into those tiny, patterned silver and gold squares that we call microchips.
Some of the cobalt you have mined eventually ends up in a chip that is soldered to a computer now in San Francisco, powering a new Large Language Model for an AI company. The owner of that company has over 1,000,000 times more wealth than you do. That won’t change, no matter how long you spend toiling underground.fn1 Other grains of cobalt line the chips inside a drone that cuts through the dry desert air of Libya at 72 kph. Packed snugly next to those microchips are explosives.
When it spots its target, the multicopter drone will dive down at 120 kph until it smashes into them. It won’t even need a human to approve of that kamikaze murder. The microchips are all the authorization it needs. It’s killing autonomously. Blood and bone will blend with the cobalt you once dug out of the ground.
This is just one story of the unseen faces that lurk behind the production of global catastrophic risk today. The specific names are fictional, but the story is not. And that one tale doesn’t do justice to the scale of the exploitation. Around three-quarters of the global cobalt supply – used in the manufacture of batteries for phones, computers, and electric vehicles – today comes from the Democratic Republic of Congo. Some 15 to 30 per cent of that is mined by what are referred to as ‘artisanal’ miners. It’s a euphemism for people working in unregulated and dangerous conditions.
Up to 40,000 of them are children who spend their days in squalid tunnels that are prone to collapse. Around the world there could be up to 100 million informal miners of all ages digging up not just cobalt, but many other rare earth metals such as the lithium used in your phone, or the tantalum used in semiconductors. The costs of quarrying cobalt are carried across bloodlines. Workers have such high rates of toxic metals in their blood that they are five times more likely than average to have children with birth defects.2 Over eight decades ago, in what was then the Belgian Congo, a different mine was birthing a different global threat.
The Shinkolobwe mine was worked by Congolese miners day and night in appalling conditions under the draconian supervision of its colonizing authorities. Medical and birth records were seldom kept, but there are reports of many children born with congenital defects (as with cobalt).
The mine is named after a kind of apple that, if squeezed, leaves a skin burn. It’s a telling name, since many of the people working in the mine suffered from the hallmarks of radiation sickness. That’s because Shinkolobwe is the world’s purest uranium mine.
Shinkolobwe was the mine that provided around two-thirds of the uranium for the Manhattan Project: the 1940s US research and development programme to build the first atomic weapons. Uranium from Shinkolobwe was present in the first atomic test bomb – the Trinity bomb – that was detonated in the deserts of New Mexico in 1945.
Shinkolobwe uranium was also in the two bombs (named ‘Little Boy’ and ‘Fat Man’) dropped on Hiroshima and Nagasaki later that year. Some of the 1,000 tons of silverywhite rock that was carved out of the Congo would eventually be vaporized and its particles burned into the bodies and buildings of Japan.3 There are many links in this chain of production from exploited miners to global risk. Another is the coal miners in the Shanxi province of western China during the rise of that country’s economy two decades ago. There, workers lived in shared dormitories with unclean sheets and beds as hard as stale bread. They were paid $7.50 a day to risk their lives. One man whose injuries when a mine caved in resulted in him becoming a paraplegic received no compensation from his employers, nor even a car ride to the hospital.
Another, already without front teeth owing to mining accidents, continued to work despite having a broken shoulder.
4 The coal the Shanxi miners excavate is shipped both across China and around the world to be burned at around 400–500°C. Coal is around 45–86 per cent carbon, and that carbon reacts with oxygen during the burning process to create carbon dioxide (CO2 ), a greenhouse gas that floats into the atmosphere and acts like a blanket, trapping solar radiation and heating the earth. The earth’s atmosphere now includes 427 parts per million of CO2 , and that figure is growing every year. The coal dug up in Shanxi will eventually become the gases that supercharge hurricanes and melt the polar ice caps. FROM HUMAN EXPLOITATION TO HUMAN EXTINCTION It is deceptively easy to think about killer drones, AI, nuclear weapons, and climate change as being separate, isolatable threats. Scholars and politicians alike have tended to regard these as a list of distinct global problems – unfortunate side-effects spewed out by innovative technology, whether that be the machinery of the Industrial Revolution or the algorithms of today.
5 They are not. Each is a link in an economic chain from toiling miners to the cleaners sweeping the floors of a military base or a fossil-fuel-company headquarters. These threats are not unavoidable; they are consciously created by powerful groups who profit handsomely from the endeavour. Global catastrophic risk is the product of a worldwide system of extraction: the Global Goliath.
Global catastrophic risk means the likelihood of a severe, unprecedented, and irreversible catastrophe that could happen in the space of decades. Past generations have experienced the global decimation of the Black Death and the invasion of the Americas. A worldwide disaster that kills more than 10 per cent of the global population while disrupting the critical systems the world depends on, such as food and telecommunications, would be an unprecedented global catastrophe. If that were to happen, we would enter into dark, uncertain territory and a potential spiral into global societal collapse.6 Scholars today aren’t worried just about global catastrophes or even a global societal collapse; some also fear even worse fates, such as human extinction. Think of this as a spectrum of catastrophe, from precedented global decimation risk through to unprecedented global catastrophic risk, including global societal collapse and human extinction. The upper end of this spectrum, from a deep collapse to human extinction, is often called ‘existential risk’. This would be a catastrophe that threatens the existence of either global society or our species, Homo sapiens (Figure 36).7 This is not about individual threats like AI or nuclear weapons. This is about the overall level of risk produced by the Global Goliath. The world is not just facing one individual challenge like climate change, thermonuclear war, autonomous armed drones, enormous volcanic eruptions, or dangerous new technologies. It is facing all of them at the same time. And alongside these threats are new vulnerabilities, such as interconnected global financial markets where a failure in one country can quickly spread across the world.
Then there are responses that could make a collapse worse, such as our tendency to turn to autocrats during times of crisis. We call issues that can significantly contribute to catastrophic risk across these risk areas (threats, vulnerabilities, exposure, and responses) ‘drivers’ (for instance, climate change creates certain hazards such as lethal heatwaves, spawns new vulnerabilities like droughts, deepening debt, and may worsen other threats like conflict). All these drivers are interrelated.
Figure 36: The spectrum of global catastrophe Think about climate change. We should not ask whether climate change is an existential risk but rather try to discover how much climate change contributes to the level of existential risk. Whether the world can handle 3°C of warming above pre-industrial temperatures depends on how the future looks. A world characterized by strong international cooperation, high trust in public institutions, and few other problems may be able to handle that level of global heating. A world that is riven with conflict, distrust, autocracy, rampant inequality, and out-of-control technology might well collapse.8 These drivers share another feature: many are driven by modern technology. The Latin phrase deus ex machina, a god from the machine, was used in the theatres of antiquity to describe a situation in which a character required divine intervention to escape from an impossible situation. A god figure was then lowered onto the stage from above. Gods don’t just bring salvation, of course; they also mete out punishment and death. The world today faces a set of problems that could be termed mors ex machina: death from the machine. Advanced technology can help improve our standard of living and manage global problems, but it has also introduced unprecedentedly destructive threats.
Collapse in the future is likely to be global and far worse than in the past.
That is in large part because of the technological threats being spawned by the Global Goliath.
Threats such as nuclear weapons, engineered pandemics, AI, and killer robots.
GOLIATH BECOMES DEATH On 16 July 1945 Robert J. Oppenheimer uttered the words ‘now I am become death, the destroyer of worlds’, quoting the Bhagavad Gita. He was watching a fireball erupt in the deserts of New Mexico, the first explosion of a nuclear weapon. This was the Trinity Test. Within a month two atomic bombs had been dropped on the cities of Hiroshima and Nagasaki, killing between 110,000 and 210,000 people.9 The blast in Hiroshima was so powerful that it etched the shadows of some of the victims into stone.
Less than a decade later, in 1954, the US conducted the Castle Bravo test, which ignited a thermonuclear bomb around 1,000 times more powerful than the one dropped on Hiroshima. Since then, over 2,000 nuclear weapons have exploded across a dozen or more sites around the world. Tests have been conducted by at least eight countries, although two have dominated: the US and Russia. There was a distinct colonial flavour to the test sites. France picked Algeria (the Algerians won their fight for independence within two years of the first test); the UK used its former colony Australia, and the US used the Marshall Islands in the central Pacific as a test ground for Castle Bravo.
The constant pursuit of bigger arsenals that could be more easily concealed from the enemy, and deliver missiles more quickly, continued throughout the Cold War. By 1986 – just over four decades since the Manhattan Project – over 70,000 nuclear weapons had been built.10 Imagine just a single 800-kiloton weapon dropped in the middle of New York City. Everything within a five-kilometre radius would be vaporized.
Everyone within twelve kilometres would be killed almost instantly. Even concrete buildings would be demolished. All those within eighteen kilometres who survived would die within a month or two of radiation poisoning. Within the first twenty-four hours a minimum of around 1.5 million people would be dead. Another 1.5 million would be injured and the Manhattan skyline would be essentially empty: a flat line of smouldering rubble.11 Nuclear war is unlikely to end after the first explosion. Imagine in this case that the US is struck by Russia. Russia is unlikely just to bomb New York City. They are more likely to send a swarm of missiles with the aim of crippling the US economy and military, ideally minimizing the ability of the US to retaliate. However, the US has nuclear submarines scattered around the world, hidden silos, and more than sixty nuclear bombers (the forces of land, sea, and air are referred to as the ‘nuclear triad’), all of which could be quickly mobilized. Nuclear war will not be easily contained: even a single nuclear missile is more likely to spiral into thousands of explosions than remain unanswered. A plausible simulation of a nuclear war between the US and Russia predicted that within the first few hours 34.1 million people would die and another 57.4 million would be left injured.12 Covering the earth with mushroom clouds brings a new, even darker threat into play: nuclear winter. Nuclear blasts turn cities into infernos and skyscrapers into billowing chimneys of burned materials and soot. These particles float up high into the atmosphere and act as heat sponges, absorbing incoming solar radiation and preventing it from reaching the surface below. This cools the planet. A war between India and Pakistan (both countries with nuclear weapons) resulting in around 47 million tons of soot in the skies would cause global temperatures to drop by around 4°C. A war between the US and Russia would throw up about 150 million tons of soot, dropping global temperatures by 9–10°C. The latest modelling suggests that the ensuing famines would kill up to 5 billion people, well over half the world’s current population.13 It could be worse, as there are many impacts not included in such models.14 When a nuclear bomb explodes it produces a pulse of electromagnetic radiation (commonly called an EMP). This EMP explodes outwards, a bubble that expands at the speed of light. It acts like a powerful lightning strike, overcharging and frying any electrical infrastructure it touches. Depending on the power of the bomb and the height of the detonation, one EMP could affect the entirety of the contiguous US. Think of hundreds of these across the world. What regions would be engulfed by these EMPs depends on who is involved in the war. One involving the widespread participation of nuclear powers would see most of the northern hemisphere fried. Much of the electrical infrastructure in these areas would be rendered useless. Electricity and the internet would be lost. For how long, we don’t know.
It wouldn’t be easy to replace all the now impotent electrical infrastructure. Large power transformers are highly customized and take up to two years to manufacture. It would take far longer to produce one in a world without the internet and with most of its factories and supply chains destroyed. It’s also difficult to tell how long it would take to recover the internet. The internet relies on Border Gateway Protocol (a standardized set of rules that determines how connections between computers are routed), which assumes that any servers that go offline can reconnect to a functioning network. If most servers have been destroyed, the protocol may fail.15 In the aftermath, people would, of course, move, seeking shelter, food and loved ones, relying on our old power of mobility, as always.
Unfortunately, unlike in the Palaeolithic, this could have a severe downside.
A large movement of people with compromised immune systems (due to both malnutrition and radiation sickness), alongside a rapid change in the climate and animal movement, would turn the world into a breeding ground for new zoonotic infections. Medicines and vaccines would dry up and healthcare would be limited in a world where the internet, electricity grids, and supply chains are severed.
It’s hard to imagine what the knock-on effects would be on war and politics. Fewer than 3,000 people died in the 9/11 terrorist attacks and that led to $4–6 trillion dollars’ worth of wars in Iraq and Afghanistan, as well as the approximately 777,000 killed and the 13.4 million displaced.
Imagine what millions dead and a levelling of the richest city in the world might do.
Or what would happen if thousands of cities were annihilated, with the sky becoming a web of criss-crossing missiles and then a veil of ash.
The existential risk and collapse expert Florian Jehn has explored how a nuclear winter could impact the natural world. The release of nitrogen oxide and smoke from the nuclear firestorms would degrade the ozone layer that shields the earth from ultraviolet radiation. We could lose up to threequarters of ozone, resulting in UV levels of 35–45. (The current advice is to stay indoors if UV levels are above 11.) The oceans would turn more acidic (by around 0.05 on the fourteen-point pH scale). The acidified oceans, winter conditions, and radiation from both sun and bombs would strangle an already struggling biosphere. How much extinction rates would further increase is unknown (and perhaps unknowable), but it would be devastating.16 WHEN OUR LUCK RUNS OUT A nuclear war would be the worst event in human history. It is difficult to see how such a conflict between the US and Russia would not result in a global collapse. That day could still occur at any time.
Currently nine countries have stockpiled around 10,000 nuclear warheads. That’s a welcome relief and a sevenfold reduction from the high of over 70,000 weapons some four decades ago. Mass protest movements and successful arms-reduction agreements (such as the 1979 Strategic Arms Limitation Treaty and the 1988 Intermediate-Range Nuclear Forces Treaty) drove these reductions. The spread of nuclear weapons has also been restricted. One country (South Africa) has willingly dismantled its stockpile, Ukraine gave up 1,900 warheads it had inherited from the Soviet Union, and several countries with the capability to produce those weapons chose not to.
Despite these achievements, the threat remains. The nuclear war scenarios outlined above are all possible with existing nuclear arsenals. Unfortunately, all signs now point towards these arsenals growing in both size and strength. China has expanded its nuclear arsenal from around 200 in 2020 to 500 in 2023. The US and Russia are both developing new technologies that could make the use of nuclear weapons more likely and the global situation more precarious. Russia’s hypersonic missiles are harder to detect and intercept than older, slower-moving weapons, and they reduce the time available for any target to react to an incoming barrage. The US is planning to spend $1.7 trillion over three decades to upgrade and modernize its own nuclear arsenal.
All this creates unnecessary risk. There is in any case no strong evidence that nuclear weapons are effective in preventing wars or ensuring peace.
Some have argued that the deterrent effect has ushered in a ‘long peace’, and at first glance the second half of the twentieth century does seem relatively peaceful. But this is an illusion that has been debunked on numerous occasions. The number of wars increased for three decades after nuclear weapons were built, and didn’t decline even at the point when nuclear arsenals were largest. Yes, international conflict has decreased slightly since the fall of the Berlin Wall, but that might be due to random chance. (If it was attributable to nuclear weapons, we should have expected that fall to coincide with their invention, not the end of the Soviet Union.
Even if the world has seen a decrease in conflict between countries since the world wars, that would need to last another 100–140 years before it became statistically significant.)17 What is true is that we haven’t seen any great-power wars – such as a direct conflict between the US and Russia – since the invention of nuclear weapons. Yet even here, we’d still need a longer time frame to know that this was due to nuclear deterrence and not just luck.18 We do have evidence that the absence of nuclear war is largely down to sheer dumb luck. On 26 September 1983, the Soviet duty officer Stanislav Petrov saw on his screen a wave of nuclear missiles coming towards Russia. He faced a choice: escalate the decision to his superior officer (who might have launched a massive retaliation, given tensions at the time), or not.
Luckily for all of us, Petrov decided not to. He reasoned that the US would have sent a far larger fleet of warheads if this was a first strike intended to wipe out the USSR. He was right: it was a false alarm caused by a malfunction in the satellite warning system. Later Petrov reflected that if someone else had been on shift, it could easily have resulted in a global nuclear holocaust. Even at the time, he said it was a 50/50 decision for himself.19 Such nuclear ‘near misses’ have occurred at least a dozen times throughout history. As my colleague Professor Benoît Pelopidas (the leading expert on luck and nuclear weapons) suggests, this may be an underestimate since nuclear-armed states have been notoriously secretive and many details are still classified.
States also show few signs of learning from these near misses.20 And each time one of these near misses occurs we should consider the existing peace to have been shattered.21 Today, our luck looks increasingly fragile. Russian President Vladimir Putin has already warned that a deployment of NATO troops into Ukraine would result in ‘tragic consequences’ – one of several not-so-subtle allusions to a nuclear strike. Meanwhile China has more than doubled its nuclear arsenal. A senior US official announced that the US may expand its own stockpile of nuclear weapons. That expansion would be in addition to the $1.5-trillion overhaul and upgrade of the US nuclear arsenal. The fact that we have yet to face a thermonuclear war is not because of strategic brilliance or a sign of a functioning balance of power. It’s mainly a matter of chance. If tensions continue to escalate between nuclear-armed states, then it becomes ever more likely that our luck will finally run out.22 REVIVING (AND UPGRADING) OUR GREATEST KILLERS Manipulating subatomic particles for weapons is a dangerous affair that threatens the entire world, and so is manipulating viruses. Avian influenza, also known as the H5N1 virus, has a mortality rate of over 50 per cent. It is over fifty times deadlier than Covid-19. Luckily, we know of only a handful of cases making a leap from birds to humans, in none of which it has been contagious. Had it been, then potentially the worst pandemic in history would have begun.
It is the sheer lethality of H5N1 that made a series of experiments in 2011 so concerning. A group of researchers across the US, Japan, and the Netherlands found a way to mutate the virus and make it transmissible between ferrets. In one study they spliced together genes from the H5N1 virus with swine flu (H1N1); in another they modified H5N1 and then ran it through numerous generations of ferrets until it had mutated. The scientists made a virus that could now spread through the air between mammals.
While they didn’t create a virus that could be passed between humans, and the lethality of what they had made is debated, for many it was one step too close. This is called ‘gain-of-function’ research: experiments that add to or enhance the functions of an organism, such as the transmissibility, lethality, or types of hosts a virus can infect compared with the original versions found in the wild. Advocates claim that such research can help us identify and prepare for the next pandemic; critics say it isn’t worth the risk of starting the next pandemic.
Gain-of-function research is just one example of ‘dual-use research of concern’: research in the life sciences that can be misused to do great harm.
There are others. In 2017 a small group of Canadian scientists spent less than $100,000 to recreate the then-extinct horsepox virus in six months using mail-order DNA. Synthesizing a virus from scratch (‘de-novo synthesis’) has now been done with poliovirus and influenza virus. There are fears that this could be done with variola virus, the virus that causes smallpox. That is a disease that kills roughly a third of its victims and leaves most of the survivors with deep pitted scars. It has through decades of global effort now been eradicated, but its genome is publicly available and there are no substantial technical barriers to reviving it like the small group did with horsepox.23 It’s plausible that such a high-mortality pandemic could trigger a global collapse. If we were hit by a plague such as the Black Death, we would be likely to fall apart in a way that medieval Europe did not. The people of medieval Europe were not dependent on collective infrastructure such as electricity grids and mass transit systems, and their governments didn’t base their legitimacy on providing public goods like healthcare. The modern world is different. Even a virus with a lower fatality rate, like smallpox, would almost certainly break the world’s healthcare system.
Existing systems are already under strain and increasingly overburdened by climatechange-related chronic illnesses and infectious diseases. They are also just not equipped to handle a huge influx of patients. Bear in mind that Covid19 had a fatality rate of less than 1 per cent. Even then, during the height of the pandemic Italian hospitals could not always provide ventilators for those who were struggling to breathe. A high-mortality pandemic would also hit front-line health-service workers incredibly hard. If a third of them were to die, while others were left incapacitated for weeks, how would an already overwhelmed healthcare system cope?
Without the guarantee of a functioning system to take care of the ill, the legitimacy of states would begin to dissolve, and those employed in many other essential sectors would probably stop working: why expose yourself to a one-in-three chance of dying, especially if you won’t even be able to get professional treatment? Suddenly, rubbish piles up in the streets. Power lines are no longer repaired. Deliveries fail to arrive.
Water and sewage systems begin to falter. Undoubtedly, our old altruistic ways would kick in: many would still go outside and brave the risk. Yet even courage and compassion have limits. Our entire economic system would begin to fray.
The peasants of medieval Europe didn’t expect hospital care from their states as part of a ‘social contract’, nor did they rely on large, complicated, collective infrastructure like we do. Global society appears to have become better-equipped to produce vaccines, yet we are also more vulnerable to a large enough disease shock.24 That situation is likely to get worse. In 2020 I led a WHO process to identify the most significant issues in dual-use research of concern. The process didn’t just highlight the reconstruction of the variola virus as one frightening possibility; the convened group of experts also foresaw several others, including the use of machine-learning algorithms that trawl through vast amounts of data to identify new dangerous pathogens that don’t yet exist, and platforms that allow for the quick reconstruction and engineering of different diseases and bacteria.25 Our ability to create new, dangerous pathogens is increasing rapidly alongside our technology. Sequencing the human genome, a project that began in 1990, took thirteen years and cost $2.7 billion. By 2020 sequencing a human genome cost less than $300 and could be performed in days. We can now also modify and edit genomes in real time thanks to CRISPR-Cas9 (a technology that essentially allows us to cut, copy, and paste genes). Both the new dual-use problems on the horizon and the rapid development of biotechnology have led many to fear that a small, untrained group may soon be able to create something far worse than smallpox. Evolution tends to constrain naturally occurring viruses such as smallpox: those that are too lethal kill off their hosts, while those that kill too quickly don’t spread far enough or infect enough people. An engineered virus would be designed rather than evolved and, at least in theory, would not face such limits. Rather than reviving a natural virus, a group could create an unnaturally lethal one that kills more of its victims than the plague but spreads quicker than Covid-19. In other words, an engineered doomsday disease.
Luckily, carrying out a large-scale bioterror attack is exceedingly difficult.
It requires researchers with extensive training, tacit knowledge of how to work in a lab with pathogens, access to advanced labs and equipment, substantial experience with clandestine operations to avoid being caught (probably someone who is a former military or intelligence operative), and the fortunately rare desire to murder all of humanity.
Finding a highly trained virologist who is capable of recreating a virus, let alone designing and creating a new one, is already exceptionally difficult.
Finding all these skills and experience in one individual is nigh impossible.
And more people means more coordination and an even greater chance that someone will defect and tell the authorities.26 Even with current advanced technology, bioterror is surprisingly rare and ineffective.
Currently we have a biocrime or bioterrorism incident around every 2.9 to 3.5 years. Most are hoaxes, and even those that do kill usually only claim a few casualties.27 The largest one to date was carried out by followers of the cult leader Rajneesh (also known as Osho) who poisoned ten salad bars in Oregon with salmonella, resulting in 751 cases of food poisoning and luckily no deaths. This had led some researchers to conclude that even if you have the right amount of money and resources, killing large numbers of people with germs is just incredibly hard.28 We should still guard against the possibility of catastrophic bioterrorism, but be aware that this is a highly unlikely possibility. For now, the risk comes from large, powerful organizations such as militaries and giant bioengineering firms. These are the institutions that will have the resources to create and release an engineered virus. That’s true historically as well.
While the greatest bioterror incident injured fewer than a thousand people, Japan killed around 300,000 Chinese during the Second World War by experimenting with typhus, cholera, the plague, and a host of other bioweapons.29 Even if there was a collapse-inducing bioterror attack because of cheap, dangerous technology, the bigger cause would be the governments who fail to regulate these technologies appropriately, and their failure to invest in public healthcare and pandemic preparedness. Our most robust line of defence against pandemics – whether they be natural or lab-made – is still the simple, often overlooked and underfunded work of having wellresourced functioning healthcare systems, a healthy populace with strong immune systems, strong biosecurity practices in labs, ample vaccine capacity, effective contact-tracing processes, and other pandemic preparedness measures. These remain largely absent, despite the fresh scars of Covid-19.
The risk of collapse is likely to rise over the coming decades, driven in part by advances in another potentially dangerous technology: AI.
AUTOMATED COGNITION In 2023 a long list of AI luminaries signed a letter on the risks of AI.
These included Geoffrey Hinton (known as the ‘Godfather of AI’), Stuart Russell (who co-wrote the textbook on AI: Artificial Intelligence: A Modern Approach), Demis Hassabis (head of Google DeepMind), Sam Altman (head of OpenAI), and Dario Amodei (leader of Anthropic). The letter stated that ‘Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.’30 That concern has been expressed in more modern language by the rapper Snoop Dogg: ‘And I heard the dude, the old dude that created AI [referring to Hinton] saying, “This is not safe, ’cause the AIs got their own minds, and these mother----ers gonna start doing their own s---.” I’m like, are we in a f- --ing movie right now, or what? The f--- man?’ One problem is that it’s unclear what Hinton or Snoop are talking about.
In general, AI tends to be used as a vague term to cover the technologies that make machines do things that, if a human did them, we would consider intelligent. That’s a capacious and ambiguous definition.31 Intelligence alone has dozens of potential definitions, from achieving goals in a complex environment through to acting appropriately with foresight.32 AI is, of course, also a marketing term that can overstate the capabilities of many algorithms and make them seem almost ethereal.33 Let’s not get distracted by abstract, marketing terms and stick to what these machines actually do. They are material systems that vastly speed and scale up specific cognitive tasks. That could be moving a robotic arm (spatial navigation), recognizing faces (image recognition), or stringing together words to maximize the probability of providing the correct answer to a question (text processing). While intelligence is a broad, fuzzy term, cognition refers to specific, concrete abilities that different algorithms can excel at. As the Oxford cognitive scientist Carla Zoe Cremer observes, we are better off thinking of this as ‘automated cognition’.
What Hinton and others truly fear is that one algorithm will be able to automate and combine a large number of different cognitive abilities. It will start to possess the kind of broad, flexible abilities that a human has.
It may even be able to learn and develop new cognitive abilities. Think of this as ‘automated general cognition’. Then there is the much more speculative chance that an algorithm might not just become general, but also orders of magnitude quicker and better performing than a human.
From now on, when I use ‘AI’ or ‘AGI’, I’ll be referring to automated cognition or automated general cognition.34 The idea of machines annihilating or enslaving humans is by no means new. In 1872 the novel Erewhon by Samuel Butler depicted a society that had destroyed most of its mechanical inventions after the machines began to gain consciousness, self-replicate, and rise up owing to natural selection among machines.35 In 1965 the Oxford-based AI researcher I. J. Good published an essay on ‘Speculations Concerning the First Ultraintelligent Machine’. He hypothesized that if a machine were to become more intelligent than its human programmers then it could design even smarter algorithms, triggering an explosion in AI. Humans would be quickly left behind. Good warned, ‘Thus the first ultraintelligent machine is the last invention that man need ever make provided that the machine is docile enough to tell us how to keep it under control.’36 In the past few decades, experts, including leading computer scientists, have joined Good in taking this risk seriously. The perceived threat is the creation of an ‘unaligned’AGI. That is, an AGI with a mission that is either not under human control or does not reflect human values (as difficult as they may be to define across 8 billion people). The concern is not that the machine will be malevolent or murderous; rather, we’ll just be collateral damage for them to achieve their programmed goals.
This wouldn’t be a mere collapse, but potentially complete human extinction, or perhaps an even worse fate, like perpetual enslavement.37 Unlike nuclear weapons, this is a hypothetical concern. We don’t know exactly what such a machine would look like. We don’t know how it would control or destroy all humanity. The ways that scholars have suggested all tend to be sci-fi-like, such as the algorithm manufacturing and releasing billions of microscopic drones to simultaneously assassinate all 8 billion of us.38 The ultraintelligent machine quickly begins to sound like a malevolent technological god. And we should always be sceptical of appeals to gods, whether they reside in real clouds or the digital one.
We don’t even know if such a machine can be built, or how long it would take. Numerous attempts have been made to find out. A 2022 survey asked 352 AI experts when unaided machines would be able to complete every task better and cheaper than a human: 90 per cent said it would occur within a century and 1.1 per cent said it would never be built. The most recent survey of 2,778 researchers found a 5 per cent chance that superhuman AI would result in extremely bad outcomes by the end of the century, such as human extinction. Superforecasters put the risk of AI causing human extinction lower, at around 0.38 per cent.
Few of us would hop on a plane with a 5 per cent (or even 0.38 per cent) chance of crashing. In this case, a crash would potentially kill not just a few hundred, but billions.39 There are reasons to be sceptical about these results: surveys are one of the least accurate forms of forecasting.
We have no evidence that they work on long time frames for such a complex question as the creation of an unprecedented technology. They don’t allow experts to deliberate, and individual experts don’t have a great track record of prediction, even in their own fields.40 Importantly, these surveys show an enormous range of opinions, all the way from ‘it will be built within years’ to ‘it will never be built’.41 AI could also slow or stagnate for a range of reasons, including increasing scarcity of the materials needed for labs and microchips, hard physical limits (such as how small one can make a microchip), or innovations simply becoming too difficult and costly. The polymath Vaclav Smil, in his magisterial book Growth, puts forward the case that the growth of all systems eventually stops. Diminishing returns doesn’t just apply to extraction, but also to innovation: every dollar of investment results in fewer and fewer benefits.
That halting point for new technologies, whether it be in bioengineering or AI, may be far sooner than anticipated.42 AI may already be facing limits. Projections suggest that large text models like ChatGPT will be using datasets equivalent to all the available online, public human-generated text by 2026 to 2032. Within a few years we could run out of data to develop bigger, better algorithms. This limit to data could be overcome by relying more on ‘synthetic data’ created by AI, but whether it’s possible and what problems this would create is uncertain.43 There could also be limits to energy use. One report estimated that energy usage by data centres could more than double between 2023 and 2027, causing 40 per cent of data centres to face energy constraints.44 The approach of building bigger algorithms to get better results could also plateau. Some researchers in the field of machine learning think that the approach of simply scaling up existing algorithms (with more data and more parameters) may already be hitting a wall.45 AI itself is an industry product that is already prone to hyperbole and exaggeration. The heads of AI companies, such as OpenAI’s Sam Altman, are happy to feed the concerns about AI safety because it makes their products seem far more powerful than they are.
They are artificial hypemen.46 Today we have forecasts of AI significantly boosting the economy and speculative visions of a world run by virtual AI workers. Yet we haven’t seen AI have much impact on growth or productivity. Even going back to 1987 the Nobel Prize-winning economist Robert Solow quipped, ‘You can see the computer age everywhere but in the productivity statistics.’ That fundamentally hasn’t changed: we still haven’t seen any significant economic boost from AI.
We also don’t know what is necessary to build intelligence. One hypothesis is that intelligence only evolves in a body. It is an outcome of an embodied interaction with an environment using different senses.
This would make the proposition of building machine intelligence far more difficult. Robotics lag far behind the development of large text models, while computer simulations would need to be run at an incredibly high resolution (requiring an enormous amount of data and raw materials like semiconductors) to produce an embodied intelligence.47 Despite all these reasons for caution, we still can’t rely on the assumption that an AGI is impossible to build or will face limits to growth. History shows how quickly previously unthinkable technology can be built. On 12 September 1933 the physicist Leo Szilard read a newspaper that would shape history. An article in it quoted Lord Rutherford, the man known as the father of nuclear physics, who dismissed the use of nuclear fission as ‘moonshine’. Szilard took a walk afterwards. During that stroll he came up with the idea of a nuclear chain reaction. That became central to the creation of the atomic bomb, which was built just a dozen years later. AGI may seem unlikely now, but in an uncertain world we can’t blithely dismiss fears of powerful new machines. We need to be cautious. The stakes are so grave that even an incredibly low likelihood of building a catastrophic AGI should give us concern. More importantly it should give us a reason to pause efforts to construct such a machine. We need to stop efforts to build something that could potentially kill everyone until we can make an informed, genuinely democratic decision about whether it is worth the risk.48 Stopping AI is possible. All it may require is simply addressing the economic exploitation that underpins increasingly powerful machines.
Even the most impressive algorithms are still built on cobalt miners in the DRC, unfathomable amounts of data provided by artists, writers, and ordinary people, and hundreds of hours of training and reviewing by flesh-and-blood humans. Some of that work is done by humans known as ‘content moderators’. These poor souls need to trawl through thousands of sentences and images, telling the algorithm whether or not material is appropriate.
Every day they are exposed to the worst of humanity, ranging from torture to sexual abuse. OpenAI hired many of its content moderators from Kenya, paying them between $1.46 and $3.74 per hour and offering them no support. Many of the workers have complained of psychological trauma.49 Data annotators in Uganda are paid similarly low wages, despite the fact that building an algorithm requires far more hours of work in annotating, organizing and checking data than coding an algorithm.
The data that is annotated in Uganda is then exported along a fibreoptic cable that runs close to the old East African railway. That was a railway financed by the British to take resources from Uganda’s interior to the port of Mombasa, from where it was shipped back to the imperial core. Lurking even under AI there is the global, rimless wheel of empire. Now, data, finance, cobalt, and silicon are moving along these spokes, not just colonial commodities.50 All this data is, of course, stolen.51 Data is fast becoming the latest lootable resource that can be easily seen, captured, transported, and stored.
While we celebrate the CEOs of leading AI companies, the algorithms themselves are models of collective intelligence founded on our collective data. All this data is extracted from the web without our consent, while none of us receives a single dollar in compensation for our valuable data. AI is more extractive than artificial: it’s dependent on cheap workers, cheap materials, and stolen information. It wouldn’t be possible (or at least inexpensive and highly profitable) without the Global Goliath. If we started paying workers fairly and requiring consent and compensation for the use of public data, the development of AI would both slow and become far more controllable.52 THE AI ACCELERANT AI doesn’t even need to reach such heights to be dangerous. It already poses enough problems. That is because AI accelerates many existing concerns, such as allowing for sophisticated automation of hacking, amplifying the threat of cyberwarfare between countries, and allowing for the construction of deepfake videos, causing problems for both our political and personal lives.53 That is just the beginning of a far longer list of potential woes. One particularly worrisome possibility is that the responsibility of whether to fire nuclear missiles is placed in the digital hands of an algorithm. There is already a precedent for this. Russia has since 1985 possessed the Perimetr system, more charmingly known as ‘the Dead Hand’. When activated it has contact with Russian command.
If that connection were to be lost, the computer system would assume that they had been destroyed and automatically launch a radio-tipped rocket that would shower down radio signals instructing all nuclear silos below to launch. The intention is to provide foolproof deterrence: even if the Soviet commanders are killed in a first strike, the rest of their arsenal will still launch on the US. It’s a pre-programmed doomsday device.
Fortunately, it’s active only during crises, though we don’t know whether it is active at the moment.
In 2024 Adam Lowther, former professor of the US Army’s School of Advanced Military Studies, and former US Air Force Colonel Curtis McGiffin advocated for the US to adopt its own dead-hand system through the integration of AI. The US military already has a worrying fondness for automation and integrating AI into its forces. This has led to a growing chorus of scholars warning the government not to hand over the nuclear launch codes to an algorithm. It would be the automation of nuclear holocaust.54 The military has already shown a willingness to use AI to scale and speed up murder. In 2016 the National Security Agency (NSA) of the US hooked up a new machine-learning algorithm to its drones hunting across the skies of Pakistan. For over a decade the US had been using drones to launch missile strikes against people and compounds across the country. This new programme was different. It used the metadata (information about data, for instance where and when a call took place) of 55 million Pakistanis to determine how likely they were to be a messenger for terrorist cells. It analysed how they moved, who they called, and who they met, to decide whether they should die.
The programme would then create a ‘final kill list’ of the most suspicious suspects for the drones to target. The program’s name was SKYNET.
We don’t know how many were killed because of SKYNET, or even whether they were guilty of aiding terrorist organizations.55 Using AI to accelerate bombing campaigns isn’t just a US pastime. In 2023, in the opening weeks of Israel’s bombing campaign of Gaza, a new set of AI systems was deployed. It had been developed by Unit 8200, Israel’s version of the NSA. The ‘Lavender’ system would use swathes of data to predict who was an appropriate target connected to Hamas or the Palestinian Islamic Jihad. At its height Lavender identified as many as 37,000 Palestinian men. These had to be vetted by a human operator, but the procedure was inappropriately brisk. One Israeli using the Lavender system admitted they would spend around twenty seconds checking a target before giving approval for them to be bombed. Another AI system called ‘The Gospel’ would identify buildings and facilities that were most likely used by targets. The targets were then tracked and a third algorithm called ‘Where’s Daddy?’ would alert the Israeli military when the targets had arrived back at their family homes and were ready to be assassinated.
Dropping a bomb on an unguarded house was seen to be more reliable than attacking military facilities. According to internal sources, the Israeli military used cheap ‘dumb bombs’ that wiped out entire houses and were permitted to kill up to twenty civilians for every target. This was largely done as a cost-cutting measure. We don’t know how many innocent lives were lost due to these AI systems, but it could easily be in the thousands.56 Lavender at least required some oversight. There are some AI military systems that require no human approval at all to decide who lives and who dies. These are usually called ‘lethal autonomous weapons’, but for clarity let’s just call them ‘killer robots’. The first confirmed attack from a killer robot came in 2020, when a Turkish-made Kargu-2 drone (the same kamikaze drone mentioned earlier) hunted down and assaulted people in Libya. A year later in 2021, Israel launched a swarm of AI-coordinated autonomous drones during an attack on Gaza. AI-enhanced drones packed with explosives are, of course, already being used in the war between Russia and Ukraine.
This is just a small taste of things to come. The US is developing fleets of unmanned ships and swarms of up to 250 AI-managed drones. This is likely to be boosted by the $1 billion Replicator Initiative that plans to build even larger AI-managed drone swarms that will hunt down targets.
Not to be outdone, the US Air Force is also looking to construct a fleet of 1,000 unmanned fighter jets to be flown by AI systems. Russia is also working on AI-run drones, vehicles, and even missiles. Perhaps the most concerning is an autonomous, nuclear-armed underwater vehicle called Poseidon.57 Proponents of AI warfare suggest that the use of killer robots and integration of AI into military operations could make war more restricted and humane. Algorithms are not expected to be clouded by emotion or bloodlust and could be hypothetically programmed to abide by strict rules.
After looking at operations like Lavender and SKYNET, it is hard to take such a position seriously. The goal of using these systems is speed, scale, and cost, not compassion, restraint, or a desire to abide by the rules of warfare.
The debate over whether killer robots will be humane also overlooks bigger problems. Killer robots won’t stay on the battlefield. It is only a matter of time before one is used to assassinate a head of state, or a swarm is outfitted with facial recognition to exterminate a rival political group (a scenario chillingly depicted in a short video titled ‘slaughterbots’). Even that may not be as bad as how a government could use killer robots to intimidate and even kill its own citizens. Historically, larger, non-violent movements are more likely to succeed.58 Part of that is because violent action is less likely to be carried out by police and the military when a gathering has such a broad cross-section of society. Will you still fire into a crowd when you might shoot a cousin or a friend? Killer robots, however, would have no such qualms. They would, at least in theory, be obedient to whoever controls the code.59 The scenario of killer robots becoming widespread across a population is terrifying. The scenario in which they become the ultimate monopolizable weapons – a new, modern Goliath fuel – is perhaps worse.60 Machine learning could also supercharge our global ecological crisis.
One search on Google uses as much electricity as is required to power a lightbulb for two minutes. One query to ChatGPT is twenty-five times more costly: you could light up a bulb for fifty minutes. That is an enormous increase, and one that can accelerate the already substantial emissions of the digital world. If all our digital technology was a single country, it would be the third highest consumer of electricity, behind the US and China.61 Major fossil-fuel companies such as Shell intend to use machine learning to improve their ability to find and tap new sources of oil and gas.62 Any new fossil-fuel developments are completely incompatible with limiting global heating to 1–2°C. To have just a one in two chance of keeping the planet below a safer 1.5°C level of warming, around 60 per cent of oil and gas, and 90 per cent of known coal reserves, need to remain in the ground.63 Using AI to search for more at this stage is a public commitment to heating the planet.
An automated nuclear response, killer robots, mass surveillance, digging up and burning more coal: these are all outcomes of AI. Then there are the myriad other problems that just the first wave of chatbots and generative algorithms have created: supercharging plagiarism in school, recommendation algorithms that deepen political divisions, bots that speed up cyberwar and cybercrime, the theft of intellectual property from artists and writers, and the spectre of mass unemployment due to automation.
These systems offer numerous benefits as well. AI-enhanced detection of cancers and identifying new drugs is important, AI-generated art is cool, chatbots can be nifty, and AI could help progress areas of science, but it’s hard to see these outbalancing the shadow of a crowd of protesters being cut down by a buzzing cloud of drones. This is the biggest threat of AI: it accelerates the creation of global catastrophic risk. Arms races and military operations – whether it be the race to a bigger swarm of autonomous armed drones or deciding which families to drop bombs on – get speeded up. The hunger of the global economy for data, energy, and resources will deepen. The need for more and more data for better algorithms will drive mass surveillance. While thermonuclear war or engineered pandemics (or, more speculatively, AGI) could directly cause collapse, this AI accelerant is closer to a collapse by a thousand cuts. It is the equivalent of pouring a tank of petrol onto an inferno.
AI will also slowly and irrevocably take our fate out of our own hands.
As the world speeds up, humans can no longer keep pace. That slow theft of decision-making power is already happening on battlefields and in courtrooms across the world. Speed comes at a price. Algorithms may be quicker and more efficient than humans, but we just need to think of Stanislav Petrov to remember the costs. If an algorithm had automatically responded to that nuclear alert forty years ago, it would have made a different decision from Petrov’s. You wouldn’t be reading this book.
17 Gaia vs Goliath Forever Corrupted – Playing With the Planetary Life-Support System – How Bad Could Climate Change Get? – Tipping into Catastrophe – Breaking Boundaries – The Roots of the Anthropocene FOREVER CORRUPTED If you’ve been out in the rain recently, or drunk a glass of water, then you have probably been exposed to poly- and perfluoroalkyl substances, also known as PFAS. They have been called ‘forever chemicals’ owing to their persistence. They can last for years, as the environment has no clear way to break them down. There are around 4,500 of these fluorine-based chemicals that were originally intended to be used in heat-, water-, and stain-resistant products. Since then, they have spread from these materials into the earth and our bodies. PFAS are now found in the bodies of 97 per cent of Americans. They have also been found in soils and rainwater around the world. Often in levels that exceed European and US safety guidelines. Even if you are in remote regions of Antarctica, you will be exposed to PFAS.
There is simply no way to avoid them.1 Substances such as PFAS have been linked to a range of health outcomes globally, including higher cancer rates as well as declining fertility and sperm count levels. They have also been associated with an increase in birth defects, less-effective vaccination responses, raised levels of liver enzymes and cholesterol, and kidney and testicular cancer. What is more worrying is what we don’t know. We lack long-term safety assessments for most of these long-lasting pollutants. We know even less about how they interact with the environment and other compounds. Such assessments may be impossible, given the quantities in which these artificial compounds are produced: over 100,000 types and 2 billion tonnes every year.
2 Also, they may not be reversible, since PFAS cycle continuously through rivers, rain, and other water sources. Concerningly, the costs of trying to clean them up may simply be prohibitive. It is currently estimated that trying to remove just one type of PFAS, perfluoroalkyl acids (PFAAs), at the same rate they are produced every year would cost between $20 trillion and $7,000 trillion.
In other words, the cost of just cleaning up one type of PFAS probably exceeds the global GDP (which is around $106 trillion). And that wouldn’t even get rid of what has already been released. These truly are forever chemicals.3 PFAS are part of a wider basket of synthetic compounds and chemicals that are called ‘novel entities’. They are geologically novel, as they are synthetic and have never previously existed on earth. Novel entities include PFAS alongside different types of plastics (including microplastics), bisphenol A (BPA), and persistent organic pollutants such as aldrin (a pesticide), highly carcinogenic polychlorinated biphenyls (PCBs), and DDT. DDT was banned after decades of use since it was found to be highly persistent and toxic to many forms of life, as well as carcinogenic to humans.4 The amount of novel entities like PFAS that the earth system can handle is what researchers now consider to be a ‘planetary boundary’. Planetary boundaries are zones of human impact on the earth system. Once you pass a boundary, the risk of abrupt and irreversible change increases dramatically. Change that could have disastrous consequences for humanity and the planet. We are now thought to be beyond the boundary for PFAS: we have already entered the uncertain danger zone for novel entities. This is not the only planetary boundary we have crossed. We are already feeling the effects of a more visible trespassed boundary: climate change.5 PLAYING WITH THE PLANETARY LIFESUPPORT SYSTEM The history of life on earth is a story about the climate. We’ve had four mass extinction events in the last 541 million years (the Phanerozoic period of earth with larger-scale multicellular lifeforms).6 Each of these is associated with a loss of around 75 per cent or more of species.
Climate change, namely global warming, played a key role in all of them.7 The human story is also entwined with the climate. For the first 95–99 per cent of our species’ history, we lived in climatic conditions in which sustained, intensive agriculture was almost impossible. The dry, cold conditions with abrupt, volatile climatic shifts weighed against intensive food production, or even just settling down in one place.8 Then, 12,000 years ago, the entry into the warmer, more stable Holocene allowed for Goliath fuel, especially lootable resources, leading to the incredibly slow but fundamental shift from hunting and gathering to farming, from civilization to Goliath.
Ever since then we have been fundamentally constrained by the climate.
For the past 6,000 years the vast majority of humans have lived in a narrow climatic envelope of ~11–15°C (along with a second smaller envelope of ~20–25°C, for India’s monsoon region).
This is the ‘climate niche’ of Goliath. The first five original political states of the world all arose in this peculiar climatic bandwidth. Since then, most cities, crops, people, and livestock have stayed clustered in these areas. Even today, this climatic zone contains the greatest amount of economic activity. While we can live in numerous other environments, such as the Arctic and the Sahara, those areas don’t tend to support cities or industrial infrastructure.9 Shifts in temperature have played a major part in shaping human history.
We’ve repeatedly seen the fingerprints of climate change, from the fall of the Late Bronze Age during a megadrought to the synchronized collapse of Goliath systems across the north of America during the Little Ice Age.
This was never a straightforward relationship of climate causing collapse, but we shouldn’t ignore the role that climate plays. Even the rise of great empires often coincided with a change in the climate.
Rome’s expansion coincides with the ‘Roman climatic optimum’: conditions that were favourable for growing crops and supporting a larger army.
10 The formation of the Mongol Empire was also given a helping push by the climate. The rise of Chinggis Khan’s empire, the largest contiguous empire in history, occurred during a period of persistent wet and warm conditions.11 These conditions meant more productive grasslands that Mongol horses grazed. More and healthier horses meant a stronger army. The contraction of the Mongol Empire and its sudden retreat from Hungary in 1242 occurred during a period of colder conditions that undermined the pastureland and mobile cavalry.
12 Climate is one important factor that moved the needle of history.
It would be a mistake to feel as though our industrial Global Goliath is now outside the reach of the effect of climatic shifts. That overlooks how dramatic the amount of warming we are facing is. The changes that rocked and directed the world in the past were regional, short term and usually less than 1°C. We are now sitting at the precipice of global heating of perhaps 2–6°C above pre-industrial levels within the next hundred years. By the end of this century we could experience temperatures that have not been felt since the Early Eocene, around 55–50 million years ago.13 Even a rise of just 2°C would lead to temperatures that have not been experienced for over 2.6 million years, since the Pleistocene epoch.14 We are facing an atmosphere and a climate that will be radically different from the one in which our species and Goliath evolved (Figure 37).15 Figure 37: Temperature change from 1000 to 2100 CE The greatest mass extinction event in the earth’s history, the Great Permian Dying (around 252 million years ago), was sparked by an initial temperature rise of 6–8°C over tens of thousands of years; 80–90 per cent of all life on earth was eventually lost. We are releasing carbon today ten times faster than the carbon pulse that caused the Great Dying. What we are doing to the climate is geologically unprecedented.
We are playing with the atmospheric life-support system of the planet.
HOW BAD COULD CLIMATE CHANGE GET?
The declaration that followed the 1988 Toronto Conference, one of the first climate summits, described the consequences of climate change as potentially ‘second only to a global nuclear war’. We’ve already seen how cataclysmic nuclear war could be. Could climate change approach that level of global destruction?
In truth, we don’t know. Both higher-temperature scenarios and climate risk are dangerously understudied. Analysis of reports by the Intergovernmental Panel on Climate Change (IPCC) and the wider scientific literature shows that they disproportionately focus on warming of 2°C or less. We are betting on the best case.16 What we do know is that the warming we are facing is severe and unprecedented. Our current emissions trajectory puts us on track for a temperature rise of around 2.1–3.9°C above pre-industrial levels by 2100.17 These numbers are highly uncertain. It all depends on the tremendously difficult task of predicting economic growth, energy technologies, and geopolitics decades ahead. It also depends on how the earth reacts. The earth system could be more sensitive to our emissions than we are expecting, meaning we get more warming than anticipated for every additional tonne of carbon put into the atmosphere. Our best estimates forecast that a doubling of CO2 will see temperatures rise by around 2.5– 4°C.
Yet that leaves an 18 per cent chance of warming of more than 4.5°C.18 This is not exactly what anyone would call ‘low probability’. One recent (and controversial) paper led by NASA scientist James Hansen suggests that a doubling is highly likely to cause warming of 4.5°C. In the long run, over more than a century, this could result in heating of 10°C.19 It could be even worse than that if the earth system springs any hidden surprises. One poorly understood factor is how clouds respond. One recent simulation suggests that at very high levels of greenhouse gas concentrations in the atmosphere (roughly a tripling of current levels), the stratocumulus cloud decks (fluffy, low-lying clouds) around the tropics will begin to break up. Clouds usually reflect sunlight, effectively cooling the earth. Their loss will result in even more warming, perhaps up to 8°C of it.20 This would be in addition to the warming of around 6°C already caused by high concentrations, resulting in a spike of heating of about 14°C. This is, for now, a speculative simulation. Let’s hope we don’t get the chance to test its accuracy. So far, it appears that we are on track for higher levels of warming and may have been underestimating how sensitive the climate system is. As of 2025 we have already temporarily exceeded 1.5°C. Over the period June 2023 to May 2024 global temperatures were 1.63°C above pre-industrial levels. This is a significant acceleration of previous warming trends, but whether it is temporary is unclear.
Even a rise of 1.5°C would have dramatic consequences. It would significantly increase the likelihood of ‘multiple breadbasket failures’: simultaneous failed harvests in the major crop-production centres of the world, such as the US Midwest.21 This would shock the price of staple crops like maize and wheat, with repercussions for the world. One small example of such a ripple effect comes from over a decade ago. In 2010 a climate-change-driven heatwave scorched Russia. Wheat yields dropped, causing Russia to institute a cereal export ban. This caused a spike in global cereal prices, especially wheat. In the UK, food-bank usage surged as people struggled to pay for increasingly expensive staples. In Egypt, where 40 per cent of people lived below the poverty line, the effect was even more noticeable. Bread is crucial in Egypt, where it is referred to as aish, meaning ‘life’. The increasing price of bread contributed significantly to the 2011 Arab Spring revolts there, which were initially dismissed as yet another bread riot. Now imagine that rather than a single heatwave in Russia, a far worse surge in temperatures occurred simultaneously in Russia, Brazil, and the US Midwest. We don’t know what the knock-on effects would be, and our models don’t even try to capture such possibilities, but it’s safe to assume they wouldn’t be good.22 That is just for 1.5°C. The impacts of climate change do not scale linearly with temperature. That means a temperature rise of 3°C would be more than twice as damaging as 1.5°C. Such levels of warming would violently shift the climatic niche that our societies are adapted to, with over a third of people pushed beyond the niche by the end of the century.
23 Some areas would be especially hard hit. Currently only about 30 million people live in areas that have average yearly temperatures above 29°C. These are largely non-urban peoples living in the Sahara Desert or on the Gulf Coast. Under a medium-high emission scenario, by 2070 around 2 billion people are expected to live in regions where the average yearly temperature exceeds 29°C.24 Much will depend on whether they can successfully adapt. This extreme heat will make for a far more volatile world and could even trigger other catastrophic threats. These domes of extreme heat will also encompass two nuclear states (Pakistan and India) and seven Biosafety Level 4 labs (labs that handle the most dangerous pathogens in the world).
They would also disproportionately fry the countries that are already politically fragile and likely to fall into state failure (see Figure 38).25 What would happen next is necessarily speculative. Pakistan and India have always had a tense relationship, with several border skirmishes over the past decades. They also share a water resource in the Indus River. Around 70 per cent of Pakistan’s economy relies on the Indus, which is set to dry up under future climate change.26 The chances of a nuclear conflict, accidental or intentional, increase alongside geopolitical tensions. Without remedying action, climate change in the Indus Basin is likely to inflame these tensions.
(This doesn’t mean that climate change will directly cause a nuclear war between the two, simply that it heightens the conditions for conflict that could escalate into one.)27 Then there is, of course, the heat. Humans have a physiological limit to how much heat they can withstand.
Prolonged exposure to heat can be lethal, even for young healthy adults.
Such lethal effects are usually reached once the wet-bulb temperature – the temperature at 100 per cent humidity – of 31°C is reached.28 Under global warming of 4°C vast swathes of the Middle East, South and East Asia, and North and South America would experience extreme heat conditions that reach lethal levels. Every year 2.7 billion people would have to endure weeks of such heat during the daytime; 1.5 billion would experience a month of it and 363 million would face an entire season of such intolerable outdoor temperatures.29 Global supply chains would be strained as people are limited to working outside for a few hours during darkness for weeks or months on end. Air conditioning isn’t a sustainable solution as it would simply increase energy demand, emissions, and blackouts.30 (It is also an option that the poor of the world rarely have access to.) The ripple effects of such a rise would make the 2010 Russian heatwave look mild.
A rise of 6°C or more would lead to mass extinction. Lethal heat would engulf even more than 3 billion every year. The Arctic would be ice-free year-round while Antarctica would lose around 70 per cent of its ice, leading to sea levels twelve metres higher than today. 31 Billions would be displaced from the climate niche. As the world warms, animals and infectious diseases will also move with the changing climate niche. This will all heighten the risk of new zoonotic infections (like Covid-19) as we get novel mixes of people and animals in hot conditions.
This picture of the future is bleak enough, but it gets worse once we consider tipping points in the climate system.
TIPPING INTO CATASTROPHE Tipping points are thresholds where just a small push can propel a system into a totally different state, sometimes with no possibility of return, just as when you rock too far back on a chair you commit yourself to toppling over. Tipping points are one of the key reasons that 1.5°C is considered the planetary threshold for dangerous climate change. Even at this temperature, six global tipping elements are at a high risk of being passed. These include the collapse of the Greenland and West Antarctic ice sheets as well as the loss of low-latitude coral reefs.32 In other words, we are already careening into a world that will lose its coral reefs and two of its largest ice sheets.
These tipping points don’t just threaten to create significant disruption.
They could also propel further warming. They are what we call ‘positive feedbacks’: self-amplifying changes. Think of the loss of an ice sheet.
The bright white of snow reflects sunlight back out into space. Dark seawater absorbs sunlight, heating the earth. As ice sheets shrink and more seawater is exposed, the earth warms up even more, melting even more sea ice. It becomes a self-propelling, sometimes runaway, change.
We already have numerous reasons to worry about such self-perpetuating cycles. Currently, the Arctic is heating nearly four times faster than the global average rate and twice as fast as previously expected. The permafrost across the Arctic hides a ticking time bomb: it holds nearly twice as much carbon as the atmosphere. As it thaws, increasing amounts of methane are released. It’s a process that is irreversible over the space of centuries (it would take far longer to reverse). According to one model, we may have already reached the point of no return: methane will continue to leak from the Arctic for hundreds of years even if we don’t release one more tonne of carbon.
These are just the tipping points we know of. A world that is 3–4°C warmer than pre-industrial temperatures would be radically unfamiliar. If it happens, it will be so different that it is immensely difficult to know when the warming will stop.33 Even tipping points that could occur this century with lower levels of warming are alarming. One of the most concerning is the shutdown of the Atlantic meridional overturning circulation, known as the AMOC. The AMOC acts as a huge, liquid conveyor belt, taking warm salty water from the tropics and transporting it to the North Atlantic. This warmer water releases heat before cooling and becoming denser, sinking down before moving south back towards the tropics. Melting ice from the polar regions and higher temperatures in the ocean are combining to slow the AMOC.
A complete shutdown of the AMOC was once considered a speculative possibility. That is no longer the case. The latest evidence using sea surface temperatures over the past two decades suggests that the AMOC could pass the threshold towards collapse during 2025–95.34 A shutdown of the AMOC would reshape the world. It would plunge Europe into an ice age while warming the tropics. London could be cooled by around 10°C, while areas further north could be worse. Bergen in Norway could see temperatures drop by 15°C.35 Much of the arable farming land in the UK would become unsuitable for growing crops.
Globally, land that can be used for growing corn and wheat would be halved compared to a world without climate change.36 This is not a far-fetched or distant scenario. Once the AMOC passes a tipping point the collapse can unfold within decades. While the numbers are highly uncertain, Stefan Rahmstorf, one expert studying the AMOC, put the odds of crossing the tipping point by the end of this century at 50/50, the flip of a coin. As with AI, we shouldn’t rely on the estimates of a single individual, but all the evidence points towards an AMOC collapse within the next century as being plausible and not even low probability. While the AMOC would lead to a regional ice age in western Europe, unfortunately the rest of the globe would continue to heat.37 The danger of the AMOC is that, once we cross a certain threshold, we are committed to its collapse, even if we stop global carbon emissions entirely. This is the unsettling dynamic of how most tipping points work.
The AMOC is only one among several interlinked tipping points that are now coming to light, including the Antarctic Circumpolar Current (a current surrounding Antarctica that could be 20 per cent slower by 2050 owing to climate change). Tipping points don’t just characterize the climate system, but also much of the earth system that we are degrading. We aren’t just risking pushing the climate system over the edge, or pushing into hazardous unknown territory with novel entities; we are wading into the danger zone for multiple planetary boundaries.
BREAKING BOUNDARIES There are nine planetary boundaries. These are estimated thresholds past which we will leave the safety of Holocene-like conditions for the planet and enter a state in which the earth can no longer effectively self-regulate.
We’ve transgressed six of these, four of which are in the high-risk zone (Figure 39).38 None of these show any signs of abating. The high-risk zone for these boundaries means two things. First, we’ve moved beyond the conditions in which our societies developed (as with our shift of the human climate niche). Second, the potential for uncertain, abrupt, and irreversible change and damage increases significantly. We’ve already seen this for novel entities and climate change: tipping points and uncertainties are in many ways the scariest part of what we are doing to the earth system.39 Regardless of tipping points and uncertainties, we are already heading towards the next mass extinction event. Current rates of species extinction are around 10,000 times what they would be without human intervention.40 In other words, species are going extinct 10,000 times quicker than they otherwise would because of human activity. Thirty per cent of all mammals are now at risk of extinction.
Some species are suffering far more than others. Insects are dying out at rates eight times those of birds, reptiles, and mammals. Total insect populations are decreasing by around 2.5 per cent per year. Based on current trends, there is the real possibility of most insect species disappearing from the earth by the end of the century.
41 Again, there are potential tipping points. The fear here is that the loss of enough critical species will start a cascade across the webs of life. As one prey species dies, its predator goes extinct, and the losses of both ripple across ecosystems.
Figure 39: Transgressing the planetary boundaries All planetary boundaries are interconnected; transgressing any one of them could plausibly be a key contributor to collapse. We are transgressing six so far. One study of extinction rates among marine fossils found that a temperature rise of around >5.2°C tends to be a trigger for mass extinction.42 Similarly, other studies have found that large changes in the earth’s carbon cycle are associated with mass extinction events. The current burst of carbon we are releasing may have already passed the threshold for previous mass extinction events.43 That is just the interaction between biodiversity loss and climate change; there are many others.
These are just the boundaries we have studied and know about. There may be others that we are already beginning to transgress. These include deoxygenation of the oceans and the cycling of different elements in the earth such as boron, arsenic, and mercury. One could even be in space.
Competing companies are racing to fill the atmosphere with fleets of satellites. Projects such as SpaceX’s Starlink will lead to up to a million new spacecraft being launched in the coming decades. This wave of floating metal has foreseeable dangers. As the defunct satellites drop from orbit, they burn up in the lower atmosphere, leaving conductive, electrically charged particles. The earth is surrounded by a magnetic sphere that holds in place two belts of charged particles (known as the Van Allen Radiation Belts). The debris from burned satellites will leave a trail of conductive particles that is billions of times greater than the current Van Allen Belts.
Adding billions of tonnes of unknown chemicals to the atmosphere hasn’t worked out well so far. Doing the same to the earth’s magnetic field is also unlikely to be a wise move. There is already evidence that this could weaken the earth’s magnetic field, and the re-entry of spacecraft could deplete the earth’s repairing ozone layer. The weakening of both the magnetic field and the ozone layer will leave humanity even more vulnerable to solar storms (which could destroy electrical infrastructure) and cancer-causing solar radiation. It could eventually push us into the high-risk zone for a fifth planetary boundary: stratospheric ozone depletion.44 Then there are the unknown effects, for instance whether billions of floating metal particles littering the atmosphere could warm or cool the planet.45 From the oceans and soils to the skies and beyond into space, the earth system is being pushed into a perilous state.
Infringing on some of these planetary boundaries, such as climate change, could directly lead to collapse. Others act as indirect drivers of collapse by slowly weakening societies. For instance, novel entities could lift the rate of chronic illnesses and disease, overburdening already struggling healthcare systems and creating further distrust in public institutions. This doesn’t make such drivers less relevant to collapse.
Remember that the fall of Rome and Chinese dynasties were occasioned by both numerous vulnerabilities (such as corruption and elite competition) and bigger shocks (like disease and invaders). The important lesson is that these interacting contributors to collapse all emerge from an extractive Goliath system.
THE ROOTS OF THE ANTHROPOCENE We no longer live in the Holocene but in the Anthropocene. Ours is a geological epoch in which human activities have become a key shaper of the earth system. Today you can find not only PFAS in the Arctic but also plastic 10,975 metres below sea level in the deep-ocean Mariana trench.
Our gouging environmental fingerprints are now inescapable.46 While our actions now will determine both the climate and the fate of most species (including our own), the Anthropocene is not a recent phenomenon. It is a process with deep historical roots. Early hominids were habitually using fire at least 400,000 years ago, and perhaps as early as 800,000 years ago. Fire remade the landscape. It not only favoured species that could withstand burning (pyrophytes), or benefited from it (pyrophiles), but also selected for certain types of animals. As we spread from Africa, our footsteps on the land became heavier. Hunting animals, burning forests, and harvesting wild crops started to transform the earth. Before the first plough had been picked up, foragers had already modified over three-quarters of the world’s regions.47 As we raised sickles and scythes our prints on the land deepened. The turn towards domesticated crops and animals meant a new kind of landscaping. The wetlands of Mesopotamia were slowly replaced with manmade irrigation canals and pastures for goats, sheep, and cattle.
These new lands looked wild and diverse compared with the cash-crop plantations of spices, sugar, tea, and cotton that would follow 4,500 years later under colonization. The biggest acceleration of our environmental impact came with the end of the Second World War. If a future archaeologist were to examine this point in the fossil record, they would notice some peculiar trends. From 1952 to 1962 there would be a thin layer of radioactive materials such as plutonium spread across the world due to nuclear tests. Plastics and previously unseen novel entities would begin to appear. The fossils of animals become increasingly scarce. As the Global Goliath grew the entire world was re-engineered.48 One leading candidate for the date that should mark the beginning of the Anthropocene is 1945. Others think it should be the invasion of the Americas, when the sheer scale of death cooled the climate. Some believe it began when our ancestors begin to pick up ploughs to farm the land. Yet, any single date would be misleading. The Anthropocene isn’t an event; it’s a long, accelerating process. We’ve now entered a thick-and-fast Anthropocene that is rapidly pushing beyond our planetary boundaries.49 This quickening, deepening stamp on the land is not mainly due to human numbers or the general greed of humans. When we look at human activity across the tropics, whether it be the Yucatán, or Africa, there are clear thresholds past which our environmental impact intensifies. The introduction of domesticated crops such as rice, colonization, and the post1945 acceleration are the three clearest turning points. Each of these is marked by the creation and evolution of Goliath: the first states, colonial empires, and the forging of the Global Goliath. At each point there was a clear acceleration in the acres of forests cleared, the number of species wiped out, and the amount of pollution released into the soil, rivers, and oceans. The Anthropocene is the environmental bootprint of Goliath.50 Figure 40: Land use over the past 12,000 years Land use is a measurement of how humans have modified the global environment. It is one key indicator of the Anthropocene. The story of land use is simple. The conversion of land into fields for crops and pasture for grazing animals began to pick up around 8,000 years ago as lootable resources were adopted and Goliaths began to emerge. It has grown and accelerated over time with two big upticks in the eighteenth century (the Industrial Revolution and spread of Goliath) and the 1950s (the entrenchment of the Global Goliath) (see Figure 40).
Not everyone bears equal culpability for this planetary conundrum. The rich are disproportionately responsible for our current ecological predicament. The richest 1 per cent worldwide account for more emissions than the poorest 66 per cent.51 Those living in high-income countries use six times as many resources and produce ten times more emissions than those living in low-income countries.52 One study across five planetary boundaries found that the top 10 per cent of income earners globally were responsible for between 31 and 67 per cent of planetary degradation.53 The richest 1–10 per cent not only exert the biggest environmental impact but also tend to set status and consumption norms for everyone else to strive for. There is a reason why the most expensive perfumes and designer bags are marketed by the richest and most glamorous celebrities. It is also the wealthiest who hire lobbyists, make hefty political donations, and even run for office (Trump being one recent example). Lurking beneath the crisis of planetary boundaries is inequality and status competition.
The Anthropocene is largely driven by the long historical process of turning the principle of least effort of our ancestors on its head and instead using the environment and the labour of others for conspicuous consumption. The principle of least effort was all about expending as little energy as possible to meet your needs so you can maximize leisure.
Conspicuous consumption is about wastefully using as much energy as possible to signal power and status to others. Chiefs and kings of hierarchical hunter-gatherers regularly cajoled, coerced, and encouraged their wider communities to intensify the production of food for feasts.
This may have even been a historical contributor to the uptake of domesticated crops and animals. It continued from there with growing a greater surplus (usually through conquest or intensifying agriculture) to fuel more material displays of status. Whether it be throwing a feast in Mycenaean Greece (or essentially anywhere else for that matter), a maize-beer party in Monte Albán, human sacrifices by the Shang Dynasty, a parade of war captives through Rome, the colonial gold and silver that lines the churches and palaces of Spain, mountains of plantation-grown sugar and tea in Britain, or showing off a lavish set of private jets and yachts in New York, material growth is needed when the underlying goal is status. From feasts and pyramids to rockets and skyscrapers (especially those emblazoned with one man’s name), such conspicuous consumption is largely status competition that has become manifest in grain, stone, metal, and concrete.
Today, this growth fetish underpins our environmental problems.
Economic growth is one of the few factors that pushes against each of the planetary boundaries. To date, economic growth and the material footprint of societies are closely correlated. Growth across countries is set to increase material extraction and is expected to rise by 60 per cent by 2060, endangering our ability to meet climate, biodiversity, and pollution targets.54 Economic growth de facto makes us less likely to keep the climate sufficiently cool. Higher growth means we need to cut emissions more rapidly, a task we are already failing at abysmally.
Plausibly high economic growth rates could make the worst-case emissions trajectory (which puts us on track for warming of around 4–6°C) 35 per cent likely.
55 Although it is possible to grow the economy while reducing emissions (usually called ‘decoupling’, a phenomenon we’ve seen in countries such as the UK) the rates required go far beyond anything we’ve seen.
Equitably limiting warming to 1.5°C means that high-income countries would need to decouple emissions from growth an average of ten times faster than current rates.56 There is just the raw, inescapable reality that more growth means sharper, more difficult emissions cuts. Staying within 1.5°C of temperature rise means annual reductions of global emissions of 7.6 per cent, a rate faster than that achieved by Covid-19 (6.3 per cent), for decades.57 Doing that while rapidly expanding the economy – which would increase the amount of reductions required – seems at best unwise and at worst impossible. Moreover, even if we did miraculously decouple emissions sufficiently, we’d still have the problem of such material consumption pushing against the rest of the (too frequently ignored) planetary boundaries. The more we remain enslaved to Goliath’s impulse to grow, the less likely we are to keep the planet stable.
This is not to say that we should stop all economic growth: the poorest countries – from Afghanistan to Zambia – still desperately need to grow economically to eliminate poverty and increase prosperity. Instead, to stay within planetary boundaries we’ll need to eventually overcome the growth fetish and rein in conspicuous consumption. We’ll also need to realize that, past a certain point (which rich countries have already passed), growth doesn’t mean progress. Simon Kuznets, the inventor of Gross Domestic Product (GDP), our modern measure of economic growth, remarked in 1934 that ‘the welfare of a nation can scarcely be inferred from a measure of national income’. GDP was never intended to be a yardstick of progress, and it is a mistake to consider it as such: Portugal, for example, has 65 per cent less income per capita than the US but outranks it on almost every social indicator.
58 Rather than focusing on raw numbers of GDP, we should focus on what we are growing. Renewable energy, vaccine capacity, affordable housing, and leisure time are all things we should all be striving to increase and innovate in. Not bullets, bombs, cigarettes, SUVs, private aeroplanes, fossil-fuel infrastructure, or algorithms to keep teenagers glued to their screens. We need to start focusing on progress rather than growth. That means moving beyond growth and putting our production back under democratic control.
Growth simply measures economic activity, which includes the production of risk. The construction of every nuclear warhead, the manufacturing of every killer robot, and the chemical synthesis of every tonne of novel entities contributes to GDP. The biggest contributors to global catastrophic risk, whether it be in emissions, dangerous new technologies, powerful new AI models, nuclear arsenals, carbon emissions, or novel entities, are also the biggest economies: the US and China.
This isn’t a problem of humans in general eating the earth. It’s Gaia vs Goliath.59
18 The Death-Star Syndrome When the World Stood Still – The Hyperconnected World – The Accelerating World – The Concentrated World – Falling from the Zenith WHEN THE WORLD STOOD STILL As a species, we are more powerful than ever. In the last 10,000 years the yields of our food supply have grown by an order of magnitude. We can move 50 to 150 times faster than walking. Our life expectancy has tripled, thanks largely to reducing infant mortality and combating bacteria through better sanitation and antibiotics. Economic activity per person in wealthy nations is around a hundred times higher than at the peak of the Roman Empire. Availability of knowledge for the average individual has increased by an unfathomable amount. Today, a simple internet connection gives you access to 64 trillion gigabytes of data, a complete compendium from astronomy to zoology.
1 We are more powerful than ever, yet in some ways we are also more vulnerable than ever, as we discovered in 2020. New Year’s Day that year was much like any other, but within four months over 3 million people had died, entire countries had been turned into fortresses and hundreds of millions had been locked in their homes.
This wasn’t just a moment where we realized our own collective biological vulnerability. It was also a triumph of sorts. Vaccines were developed within weeks and it ended up being the fastest vaccine development and rollout in history. It could have been far quicker if countries had invested more in vaccine production and companies had not hoarded the intellectual property of life-saving drugs.2 In March 2021, while the pandemic was still unfolding, another shock hit the globe when the Ever Given, a 200,000-tonne, 400-metre-long container ship, became stuck in the Suez Canal. Around 369 ships, representing approximately 30 per cent of global container traffic, backed up at either end of the waterway, costing the world $6–10 billion in lost trade.3 Both these events could have been much worse: a virus that had killed 10 per cent of its victims rather than <1 per cent would have had a very different outcome, and the canal could have taken far longer to clear.
The world appears to be growing both increasingly robust and more fragile. I call this the ‘Death-Star Syndrome’, after the space-weapon in the film Star Wars: A New Hope that can annihilate entire planets but can itself be destroyed by a single well-placed blow. Our world is incredibly powerful and robust, yet surprisingly fragile if hit hard enough in the right place.
THE HYPERCONNECTED WORLD Both the spread of Covid-19 and the economic fallout from the blockage of the Suez Canal were worsened by the interconnectedness of the world.
During the twenty years from 1950, commercial air traffic grew in volume by over 10 per cent per year. By 2014 around 3.3 billion passengers were being carried every year. In 2024 that number was around 5 billion, and it could double in the 2030s. We can see a similar trend in shipping.
Over the past 150 years, global shipping capacity has increased around sixty-fold.4 Today, exported goods account for around 25 per cent of global economic activity by value compared with 8 per cent or less before the Industrial Revolution.5 The world is economically more connected than ever. This can mean more connections between people, cheaper goods, and more countries dependent on one another and so less likely to go to war. Hyperconnectivity is also a key reason why collapse is suppressed in the modern world. When a state begins to fall, the international community can and usually does mobilize resources to prop it back up (although, as we can see in Palestine, this depends on geopolitics). Yet this is not a complete good-news story.
The same links that can aid in recovery can also transmit a crisis around the entire system. One of the most striking examples was the WannaCry ransomware attack. WannaCry was a computer virus that could replicate itself through a network. The initial attack occurred at 07:44 UTC on 12 May 2017. In less than eight hours it had corrupted 300,000 computers across 150 countries, including networks in FedEx and the UK’s National Health Service. Again, we got lucky: the virus just demanded a ransom payment from victims rather than destroying data, and it was able to be quickly defused. At 15:03 UTC the hacker turned cybersecurity expert Marcus Hutchins found a way to prevent the virus from spreading further.
The problem with connectivity can be seen not just in computer viruses, but also in biological ones: the hyperconnected world enabled Covid-19 to move across entire continents and leapfrog oceans in just a few days.
Connectivity can often make a system resilient to small shocks but vulnerable to large ones. It is a pattern we see across systems, be they financial, digital, or ecological.6 Think of how quickly the global financial crisis of 2008 spread from a single housing bubble in the US across most of the world in the space of months.7 In the Late Bronze Age, interdependence was a blessing for elites in the best of times but became an Achilles heel in the worst. Their close trade in tin and copper – the prerequisites for making bronze – and their close diplomatic links, led to a state of independence. This meant that just the loss of a few key hubs, such as Ugarit and the Hittite Empire, was potentially enough to destabilize the system. The countries of today are, of course, far more technologically advanced and resilient than the Hittite and Assyrian empires of the Late Bronze Age, but the world is also far more interconnected and fast-moving. The Late Bronze Age also reveals that the problem is not connectivity per se but rather deep interdependence: when one part falls, so too do all those that are overly reliant on it. THE ACCELERATING WORLD The Black Death of the fourteenth century spread at the speed of medieval sailing ships; the Covid-19 virus, carried on a commercial airliner, travelled one hundred times faster; our online financial systems communicate around a hundred times faster than the speed of sound; most stock-market trading is outsourced to algorithms that can react far faster to minuscule microsecondto-microsecond changes in the market and maximize profit.
A market ruled by lightning-quick machines can lead to lightning-quick disaster. In 2010 the US stock market lost over $1 trillion on stock indices within thirty-six minutes. That flash-crash was caused by algorithmic trading, which accounts for about three-quarters of stock-market trades. The system operates far too quickly for any human to intervene and prevent a crash.8 Algorithmic trading is one example of ‘tight-coupling’: when two components in a system are closely interdependent and respond quickly to each other’s moves. Another common example is ‘just-in-time’ supply chains, whereby manufacturers avoid keeping expensive inventories by having parts arrive when needed. As with algorithmic trading, it makes for an efficient but vulnerable system. In 2011, floods in Thailand engulfed a central hub for producing car parts and hard-disk semiconductors that most of their customers relied on for just-in-time delivery. There were factory slowdowns and closures across Japan, Malaysia, and Indonesia. The price of hard drives rose by up to 190 per cent, exports from Thailand to the European Union decreased by 35 per cent (and over 20 per cent for the US), and car manufacturers in Japan saw a decrease of up to 50 per cent in profits as well as 174 days of disrupted production.9 Having more back-up stockpiles and slower trading with more human involvement would lessen the potential for factory shutdowns and flashcrashes, yet countries and companies have little incentive to do so.
Efficiency, low costs, and speed are essential in a deregulated, highcompetition, profit-driven economy, and building resilience costs money.
Under the tyranny of the quarterly earnings report, long-term planning takes a back seat.
That quickening pace is also evident in the environmental footprint of the Global Goliath. After 1945 most socioeconomic trends speeded up rapidly, whether it be population growth, the production of paper, the consumption of fertilizers, or the constructions of dams. Between 1950 and 2022 the global economy increased fourteen times over, going from around $10 trillion to approximately $140 trillion.10 All of those rapid increases required huge material extraction from the environment.
Whether it be the amount of water dammed, greenhouse gases released into the atmosphere, or the number of shrimps farmed, human impacts on the environment all soared dramatically. This is sometimes called ‘the Great Acceleration’ (Figure 41).11
But it wasn’t just a great acceleration; it was a great coupling. The fate of the earth system – whether it be through the proliferation of novel entities into water streams, the number of species going extinct, or changes in the temperature of the planet – is now tightly intertwined with the Global Goliath. And whether we like it or not, the Global Goliath is still dependent on the earth system (remember where all that cobalt, silicon, uranium, steel, and coal comes from).
THE CONCENTRATED WORLD The world is becoming increasingly monopolized. Economic power, the goods and services we use, and even our infrastructure are all becoming more monopolized and concentrated. You can taste it in every meal you eat.
Around 60 per cent of calories consumed globally come from just four crops: soy, corn, rice, and wheat (the last three were the original lootable resources and tax crops), and around three-quarters of global production of these four staples comes from just five countries.12 In the private sector, two companies own about 40 per cent of the global seed market, four control 44 per cent of the global agricultural machinery markets, and four dominate 62 per cent of the market for agrochemicals.13 That concentration of market power and control shows no signs of reversing: mergers and acquisitions between companies are making the global food system more and more concentrated. It’s a process driven by dynamics we’ve already seen throughout history: those who have power (or in this case, market share) find it easy to acquire more power.
14 The concentration and monopolization of market shares can be seen everywhere one looks. In the US in 1993 there were 193 defence contractors. Now it is essentially just five colossal firms. In the US, over half of printed book sales and 80 per cent of eBook sales happen through Amazon (the numbers globally are about the same). In 2015, 64 per cent of internet advertisement revenue went to either Google or Facebook (although Google’s share triples that of Facebook). One city (Hsinchu) and two companies (UMC and TSMC) in Taiwan satisfy 60 per cent of the world’s appetite for semiconductors. And finance is so concentrated that some institutions are now deemed ‘too big to fail’. This is a problem in itself: banks and airlines can often indulge in risky and irresponsible behaviours because they know they can rely on government bail-outs (a phenomenon known as ‘moral hazard’).15 These concentrated, homogeneous systems tend to be the most vulnerable to shocks.16 If diets are highly dependent on one crop, then one disease or shock can create a famine, as in 1845–52, when Ireland lost about a quarter of its population due to potato blight in combination with the policies of the British government.
There is also some evidence to support the old truism ‘the bigger they are, the harder they fall’. A 2020 study found that larger ecosystems tend to shift more sharply than smaller ones. Think of the change from a coral reef to an algae-dominated ecosystem after bleaching. After analysing twentyfive marine ecosystems, fifteen freshwater ones, and five terrestrial environments, one study found that that any increase in ecosystem size meant quicker change. The team concluded: ‘The collapse of large vulnerable ecosystems, such as the Amazon rainforest and Caribbean coral reefs, may take only a few decades once triggered.’ Whether this applies to societies is unknown.17 A more concentrated world poses vulnerabilities that could be catastrophic. Some scholars think that a volcano would need to rank as an 8 on the volcanic explosivity index (VEI) (a phenomenon colloquially called a ‘supervolcano’) to trigger a global catastrophe in the modern world.18 Yet a smaller volcano could also have an equally high impact owing to the structure of our world.
There are spots critical to the global economy where shipping lines, underwater internet cables, telecommunications cables, and other international infrastructure are clustered. These are what the University of Cambridge volcanologist Lara Mani has called ‘pinch points’.19 A category 7 volcano hitting one of these vulnerable pinch points could be calamitous.
The chance of one of these erupting in the next century is around one in five.20 Vulnerable points, interdependence, and rapidly reacting systems are all key ingredients that allow for a big shock or crisis to spiral into collapse.
FALLING FROM THE ZENITH ‘From morn to noon he fell, from noon to dewy eve/ a summer’s day; and with the setting sun/ dropped from the zenith like a falling star.’John Milton penned those lines in 1667 for the epic poem Paradise Lost, yet they are surprisingly relevant to our modern world. While the current Global Goliath may seem unassailable, it contains several chinks in its silicon armour.
More hyperconnected, hyperconcentrated, hyperhomogeneous, and hyperaccelerated systems make sufficiently large shocks more likely to occur and spread. This is systemic risk: the potential for individual disruptions or failures to cascade into a system-wide failure.21 It’s something that scholars have been warning about for over a decade.
Such systems create ‘synchronous failures’ in which crises spill across systems.22 We see that in the global, industrial world in the close association of oil crises with financial crises: financial crises tend to impact fossil fuels heavily, and often vice versa (such as when the 1973 oil crisis helped trigger the 1973–5 financial crisis).23 Today, crises across global systems – like food, energy, and finance – are entangled in causes and effects, creating what some call a ‘global polycrisis’.24 Underneath all this are just some simple structural features that are fundamental to the Global Goliath: speed, connectivity, and concentration.25 None of this is about bad luck; it’s about the structure of the system. The Global Goliath is like a stack of ladders. Imagine a hundred ladders lined up against a wall. Individually, it’s easy to make any of them topple, but you must hit them one by one. Now, imagine you tie them together into a single giant ladder. It takes an act of great strength to push this one over, but everything falls together in a single gargantuan crash.26 That is the situation we find ourselves in today.
19 The Rungless Ladder Swords of Damocles – Fattening Tails – Termination Shock – The Problem With Techno-Fixes – The Somalia-Denmark Rule – Not as Improbable as You Might Think – Falling Fossil Fuels – The Rungless Ladder SWORDS OF DAMOCLES In Greek mythology, Damocles, a courtier of the ancient King Dionysius I of Syracuse, Sicily, is fawning over the monarch and telling him how lucky he is to be king. Dionysius offers to let him take his place on a throne at the head of a table overflowing with the finest food and surrounded by flatterers. The catch is that he hoists a sword directly above Damocles’ head. Suspended by a single horse hair, it could fall at any time. It is the king’s way of showing that, while he may have wealth and power, he lives every moment under constant threat of assassination. Power sits in the dark shadow of precariousness.
This is a fable made for our modern world. Dan Zimmer, a catastrophicrisk researcher at Stanford University, has compared the creation of nuclear weapons and surging economic growth in the post-war period to a contemporary Sword of Damocles. Distracted by growing material comfort, we ignore the atomic sword hanging overhead.
We live under not one sword, but many. As riches multiply, so too do the existential threats delicately suspended above us.
Violence and deaths from infectious disease are at historical lows and life expectancy is at a historical high. Yet we also live in the most dangerous time in human history. This is because of catastrophic technological threats, planetary boundaries, systemic risk, and our responses to the prospect of catastrophe. Together these are all building up ‘latent risk’: risk that does not cause harm today but would do so in a post-collapse world. Our Global Goliath is continuously inflating the costs of collapse, meaning that a future societal breakdown will be far worse than any we have seen in the past.
FATTENING TAILS You are about twice as likely to take your own life than to be murdered.
You are more likely to die from cancer than from any infectious disease.
Unfortunately, this does not mean that we live in the safest of times or are on the cusp of defeating disease. Far from it.
Deaths from disease and violence follow a ‘fat-tailed distribution’.1 They are marked by a few outsized events, such as the Spanish flu, the Black Death, and the Columbian Exchange. This makes observing an average from one year, or even ten, essentially meaningless. The average will be brutally skewed by these ‘tail events’. The extremes, not the average, are the main source of risk. Importantly, these tails are becoming fatter and longer over time. Even if the world on average seems safer and healthier, the extremes are growing in likelihood and severity.
When it comes to violence, you may have a lower probability of being bludgeoned to death than a citizen of Uruk in 4000 BCE. But in 4000 BCE there was no weapon that could kill half of the people on the planet.
The risk of violence hasn’t decreased per se; we’ve simply lowered the number of observed killings per year while enormously inflating the risk of global mass murder. All it takes is one bad day, one mistake, and suddenly the violence statistics begin to look very different.
Every action or scenario has risks that vary according to their likelihood and severity. The combination of those will form a ‘risk distribution’.
Imagine you are forced to play one of two games of chance. In one, you are flipping a coin for a prize and penalty of $10,000. There’s a 50 per cent chance you lose $10,000 and a 50 per cent chance you win $10,000.
Imagine a second game: Russian roulette. You sit around a table with five others, a revolver loaded with a single bullet in front of you. If you survive, you win $1 million. There is a five in six chance you walk out a millionaire, and a one in six chance that you die. Technically, your odds of victory are far higher, and the monetary payoff is far larger, yet most of us would opt to flip the coin instead. That is because the risk distribution of the two games is overwhelmingly different, especially the ‘tail risk’ of a low-probability, high-impact bad outcome. The two games have different risk distributions: Russian roulette has a better average outcome, but a huge tail risk.2 What we have done with violence is to shift the risk distribution much more towards a game of Russian roulette. We have inflated the tail risk, while making the average outcome better.
It is illogical to say that the world is becoming safer while tail risk is growing. It’s like saying that your neighbour is a pacifist, except for that one stabbing spree when they killed fifty-five people.3 It is a fallacy to believe that we live in a peaceful age with the nuclear Sword of Damocles hanging overhead. This is true for disease as well. Infectious disease has followed a fat-tail distribution throughout history.
4 A mass-fatality engineered virus would, like nuclear weapons, be growing the tail risk (Figure 42).
The greatest threats we face today, whether it be from AI, climate change, or nuclear weapons, all began after the Industrial Revolution. They all require combustion engines, electricity, and international supply chains fuelled by oil, coal, and gas. Coal, cobalt, and uranium are all built on cheap, exploited labour arrangements that began with colonization. These threats are the tail risk of the Global Goliath. Figure 42: Fat Tails The taller curve is roughly a ‘normal’ distribution, while the dashed line is a ‘fat tail’ distribution.
The key difference between the two is that the fat tail distribution has higher probabilities of big losses or big gains. This can be seen in the gap between the tails of the two distributions.
This is pivotal for collapse. These tail risks are either key triggers for a collapse or could be set off domino-like during a societal breakdown. A nuclear war could trigger a global breakdown, or, conversely, a period of international instability and conflict due to climate change and runaway technology could trigger a nuclear war. These planetary Swords of Damocles make collapse a dark prospect. Latent risk is growing not only due to fatter tail risk, but also because of our responses to managing catastrophe.
TERMINATION SHOCK In the future, we may live in a world where a fleet of planes patrol the skies 24/7, pumping out chemicals that reflect sunlight, such as sulphur dioxide, to cool the earth – a process known as stratospheric aerosol injection. It is the cheapest and most likely form of solar geoengineering: large-scale efforts to change the reflectiveness of the earth and so lower global temperatures and combat climate change. (For simplicity I’ll be using ‘solar geoengineering’ in place of ‘stratospheric aerosol injection’).
It involves spraying particles that reflect sunlight, such as sulphur dioxide, to reduce temperatures. These aerosols act as minuscule space mirrors, reflecting incoming solar radiation back out to space and cooling the planet. Solar geoengineering would reduce global temperatures.
Solar geoengineering will be experimented with, and perhaps even globally deployed, in the coming decades. It’s relatively cheap: solar geoengineering would cost only around $2–10 billion per year.
5 To put that into perspective, the US Department of Defense requested a budget of $849.8 billion for 2025. It’s also likely to work. Solar geoengineering replicates volcanic eruptions, which also spew reflective particles into the sky with similar effect, as in the 1991 Mount Pinatubo eruption that cooled the earth by around 0.5°C during the following year.
Oil-producing countries (and their rich customer states) would support such a move, of course, since it would allow more breathing space to use fossil fuels. The most vulnerable might eventually see it as the lesser of two evils.
Ironically, human-emitted pollutants are already cooling the planet by about 0.5°C. If we were to ban them, we would experience a surge in warming that would push up temperature increases close to 2°C. In 2020 the International Maritime Organization (IMO) introduced new regulations that lowered the sulphur content of shipping fuels from a maximum of 3.5 per cent to 0.5 per cent. It effectively cut sulphur emissions from shipping by 70 per cent.6 Since then, the world has experienced an acceleration of global warming, with temperature increases now exceeding 1.5°C. This speeding-up of warming has probably been due to the reduction of shipping pollution that has lowered the amount of coolants in the atmosphere.
Whether we like it or not, we are already solar geoengineering the climate by emitting both aerosols and greenhouse gases.
Is solar geoengineering safe? Or could the cure be worse than the disease? In 2021 Dr Aaron Tang and I examined what its catastrophic downsides could be. We discovered that, unlike the potential of global warming, surprisingly little is known about the impacts of cooling by solar geoengineering. It could slow the monsoon and change rainfall patterns across the world, but much remains uncertain. The risk will also depend on how it is deployed, the amount of cooling, and the kind of world it is used in. Most models bet on the best-case scenario by assuming solar geoengineering will be deployed with perfect international cooperation for decades on end. This isn’t a safe assumption. It is more likely that a small cadre of countries will resort to a rushed deployment because the world is warming even more rapidly than expected.7 Solar geoengineering will also pose other unique problems such as political instability. After a solar geoengineering project is initiated, every extreme weather event may be attributed to it. Conspiracy theories and accusations of wrongdoing will abound, especially if there are no reliable scientific methods to determine just how much a solar geoengineering experiment contributed to a drought, flood, or hurricane. If the project were to be undertaken unilaterally, the country responsible would be blamed by the rest of the world.
Finally, solar geoengineering creates the problem of termination shock.8 Imagine that the world is facing a temperature rise of 4°C by 2100. In one future, we didn’t use solar geoengineering. It took 250 years to go from preindustrial temperatures to this hothouse earth state. That is because temperature rises alongside increasing greenhouse-gas emissions. Now think of a second future where the world implemented a solar geoengineering system in 2040. We’ve had six decades of cooler, more stable temperatures. Then, in 2100, the system is knocked out by a major event such as a geomagnetic storm, a nuclear war, or a severe pandemic.
The solar-reflective particles unfortunately wash out of the atmosphere with a half-life of around eight months. Within a year and a half, over 75 per cent of the chemical mirrors that have been cooling the earth are gone, yet we still have high levels of greenhouse gases blanketing the atmosphere.9 The world begins to heat up fast. Rather than taking 250 years to reach 4°C of heating, it only takes about two decades. This failure of solar geoengineering would result in a warming rate a hundred times faster than the greatest mass extinction event in history. This is termination shock: it is the latent risk of solar geoengineering.
Termination shock is not inevitable. Perhaps we won’t be hit by a debilitating catastrophe; perhaps we’ll wisely use the time that solar geoengineering provides to reduce emissions until we can safely phase out the system. Yet it is nonetheless a gamble. Solar geoengineering does not solve climate change per se: it reduces the average risk but fattens the tail risk. The average outcome of a solar geoengineering effort might be far less damaging than the climate change it is preventing, but in the worst-case scenario termination shock could make matters far worse.
Solar geoengineering would also only be a partial fix. It wouldn’t stop the oceans from turning more acidic because ocean acidification is primarily caused by a chemical reaction between ocean water and carbon in the atmosphere. Solar geoengineering would also be helpless to reverse the effects of passing certain tipping points. Lowering global temperatures can’t stop the AMOC from shutting down once a threshold has been breached or reverse it once it has shut down because the tipping-point-induced shutdown is driven by dynamics in the ocean, not by global temperature.
Similarly, if we lose ice sheets in the future, we won’t regain them by lowering the temperature back to present levels.10 Even so, solar geoengineering could be one way to significantly reduce climate risk. We could also try to design a system to safeguard against different potential catastrophes, such as strong insulation against electromagnetic pulses and multiple back-ups. A well-constructed solar geoengineering system that is implemented with global cooperation, that doesn’t act as an incentive to keep burning fossil fuels with impunity, that is robust to most global catastrophes, and that is used temporarily to shield the world while we rapidly decrease emissions, would be better than facing the direct impacts of climate change. We need to question how likely such a scenario is and how much of an increased tail risk we are willing to tolerate.
There is no right or wrong answer as to whether that trade-off is worth it.
We simply need to carefully and democratically pick our poison.
THE PROBLEM WITH TECHNO-FIXES Solar geoengineering is representative of a wider set of techno-fixes to environmental problems. Almost every environmental threat has a matching engineering solution that mitigates the hazard without addressing the causes of the risk.
This holds true for even regional-scale disasters such as cyclones.11 There are several engineering interventions to prevent the formation or decrease the intensity and duration of cyclones. These include cloud seeding (injecting silver iodide or other chemicals that help form clouds and precipitation), and aerosol injection (firing up sunlight-reflecting or other particles to de-energize the cyclone or influence clouds). This has yet to be done at scale, but it would probably work, and a group of Australian scientists have already floated the idea.12 Similarly, there are many ways to prevent the threat of volcanoes, whether at a regional or global level. These include drilling, fracturing volcanic rocks, and cooling off the magma that surrounds the central reservoir. These volcanic interventions could help release gas that is pent up, lower the intensity of an eruption, or even completely prevent it.13 For now, all these remain understudied and none has been implemented at scale. There are even ways to use engineering to combat the biodiversity crisis.
Imagine a world in which you go to smell a flower and a robot bee flies off the stamens. This is not sci-fi. The world is losing pollinators such as honey bees at an alarming rate and potential replacements include mechanical ones.14 It has also been suggested that we could artificially resurrect or ‘deextinct’ lost species. There are multiple ways of doing so, but the one that has received the most attention involves using preserved genetic material to create a clone embryo that is then carried to birth and raised by a near relative (an elephant raising a woolly mammoth, for example).
These are all part of a wholesale drift towards a more artificial world.15 In this case, deeper institutional, or behavioural policies are overlooked in favour of a techno-fix. This doesn’t mean that all these interventions are undesirable. Volcanic engineering is something we’ll have to turn to eventually if we want to avoid the devastation that would accompany a large-scale eruption. It is something that we should be putting far more money into researching and exploring. Similarly, given the rate of warming and consequences of climate change, we should be exploring and ramping up technical efforts to either delay or reverse warming. This could include attempts to restabilize melting ice sheets or to draw down large amounts of carbon by growing trees and using mechanical filters.
But we also need to be aware that an increasingly artificial world comes with a cost. Technological solutions to environmental problems are more likely to fall apart during a collapse. If we rely on using aerosols to prevent cyclones, then imagine what the loss of such methods during collapse would mean: an unprepared and unprotected world would be facing cyclones intensified by global heating. Becoming dependent on mechanical pollinators means exactly that: relying on a pollinator that can be lost owing to fracturing of the energy supply or a breakdown of industrial infrastructure. A nuclear war, or even an explosion of solar energy from the sun, could knock out the global electricity supply and all our techno-fixes with them.
Growing latent risk can also be seen in disease. Conditions such as HIV are treatable and manageable with modern medicine, but in a post-collapse world without antiretrovirals, HIV could become an even more dangerous prospect than in the 1980s. The same applies to the many devastating diseases that today have been neutered by improved nutrition, health, and treatments.
As with ecological risk, there are tempting techno-fixes for the next pandemic. For instance, some biosecurity experts have pushed for fitting Far UVC lighting across as many buildings as possible.16 These are essentially lights that sterilize a room without negatively impacting human health. Yet, if overused, a decreased exposure to bacteria could leave our immune systems weak and vulnerable. This could still be an intervention worth exploring, but we do need to think about the downsides and unintended consequences.
New technologies, including techno-fixes, tend to be built on the preexisting stack of technology. AI requires reams of data sourced from the internet, all of which depend on computers, and computers in turn rely on electricity, which is built on an expanse of industrial infrastructure from tankers moving coal to the mass smelting of steel. The constant layering of technologies on top of each other makes the overall technological system more complicated, interconnected, and potentially fragile. If the electricity grid goes down, it would take most technology with it. Technological complexity and dependency carry a cost.17 THE SOMALIA-DENMARK RULE Not all states are alike. Many states today are far more accountable and spend far more on welfare (and far less on the military) than in the past.
The US government (in 2025), which has a much larger military than any other country, still spends only 13 per cent of its budget on its military while allocating 21 per cent for social security. If you are a citizen of Denmark, you’ll get twenty-four weeks of paid leave after having a child (as will your partner), whom you can then send to heavily subsidized childcare. When they leave school they can study for free at a university, before eventually retiring with a decent pension. The benefits offered by these democratic welfare states, and the amount of control citizens have over them, is almost unimaginably high when compared with the Aztec or Akkadian empires.
States of today are also far more robust than those of the past. Sargon and Narmer had to rely on exaggerated intimidation, and most states throughout history didn’t even have a full-time police force. In contrast, rich states today have sophisticated bureaucracies (including specialized emergency services), full-time policing and defence forces, mass surveillance systems, and sophisticated intelligence agencies. There has been an enormous shift from intimidation to administration. In short, most citizens in rich countries have less reason to revolt, and far less likelihood of doing so successfully, than those in pre-modern states.
After five millennia of Goliath evolution, states have reached an unprecedented level of power and stability. More democratic states tend to provide more public goods, maintain a social safety net, and integrate their citizens more thoroughly.
As always, however, the benefits come at a cost. A welfare state is a good thing, but it becomes a liability if the state disappears. One survey of thirty historical regimes found that good governments (more democratic states providing public goods) were more likely than autocracies to experience severe collapses marked by severe, system-wide decline and a significant population loss.18 Today’s inclusive, highly centralized institutions handle crises well, but the strong reliance they generate means that if they do go down, the fall can be harder.
We can think of this as the Somalia-Denmark rule: the more predatory a state, the less harmful (or even more beneficial) its collapse; the more inclusive and benevolent a state, the more costly its collapse. If you are a citizen of Denmark, you depend on a relatively benevolent state. You have much to lose from its collapse. A Somali living under the Barre regime was better off being governed by a local warlord.
Not only are more of us dependent on states, but we are also simply more dependent on globalized infrastructure. You might be reading this while sipping coffee grown in Guatemala, made using a machine powered by a national electricity grid reliant on coal from China or gas from the US, with a phone next to you that is partly composed of rare earth metals from across the horn of Africa. Not only that, but most of us outsource vast amounts of knowledge to the systems around us.
Individually, we know infinitesimally little about how the world works.
Most of us lack the knowledge of how even basic items such as zips work. Scholars have compared us to ants: as we have come together in larger, more hierarchical groups, individuals have become more specialized.19 Individual hunter-gatherers might have had the knowledge to live from the land. Most of us now are hyperspecialized while lacking basic survival skills. Knowledge is increasingly held not by individuals but rather by groups and institutions. Our hyper-specialization and dependence on global supply chains can give us great power, but that dependencey is also a vulnerability.
This might not be an issue if the risk of a global catastrophe or of these systems failing were sufficiently low. Unfortunately, that is not the case.
NOT AS IMPROBABLE AS YOU MIGHT THINK Global catastrophe is often written off as being speculative, long term, and of such a low probability that we don’t need to worry. For instance, some analysts of solar geoengineering have downplayed termination shock, since a calamity large enough to knock the system offline is too rare and humanity would simply recover and reinstall it.20 Most estimates suggest that running a solar geoengineering system would cost less than 1 per cent of global GDP. Even an unprecedented economic catastrophe, something that wipes out 75 per cent of economic activity, would still leave plenty of money to reactivate the system.
Neither of these is a convincing objection. Assuming states will swiftly reinstate solar geoengineering, or another artificial support system, in the wake of calamity means ignoring recent experience. In 2021, creating 3 billion doses of Covid-19 vaccines would have cost $18–120 billion and provided benefits of $17.4 trillion.21 Despite this, vaccine production remained stubbornly low. States and leaders are not rational actors who always act in the wisest, most cost-effective ways during a catastrophe. Relying on them to coolly and quickly reactivate a solar geoengineering system is naïve.
Moreover, global crises are not as improbable as commonly assumed.
The chance of a high-impact volcanic eruption in a century is one in five.
Extreme space weather events are also unnervingly likely. These include solar flares (mass ejections of electromagnetism from the sun) that cause geomagnetic storms in the earth’s magnetic field, disrupting and frying electrical infrastructure. The 1859 Carrington Event is usually used as the threshold for a space weather disaster. This was an intense geomagnetic storm that luckily occurred before most of the world was dependent on electricity. If it occurred today, a similar event would cause widespread electrical failure and disruptions for months. It would probably be enough to knock out a solar geoengineering system.
Estimates for a future event that is as bad as or worse than the Carrington Event, range up to 20.3 per cent per decade. While there is no consensus, and we should place little faith in these numbers, they are not negligible either.
22 The same can be said of nuclear weapons. We don’t – and can’t – know the exact likelihood of a nuclear war, but all the estimates agree that it isn’t low. One study mathematically modelled the likelihood of an inadvertent nuclear war between the US and Russia based on past near misses and false alarms. They calculated an annual probability of between 0.001 per cent and 7 per cent.23 One expert suggested it was closer to 1 per cent per year.
24 In 2022, Harvard Professor Graham Allison put the odds of Russia nuclear bombing the US at between 0.1 and 1 per cent.25 It’s unclear whether that likelihood has since increased. Since Russia invaded Ukraine, superforecasters have calculated the likelihood of Russia using nuclear weapons outside its territories at 2 per cent.26 Pandemics too are not a very low-probability occurrence. Pandemic outbreaks are a matter of when, not if. Using a well-documented track record of outbreaks since at least 1600, scientists have calculated that one occurs roughly every 129 years. That means there is a 38 per cent chance of experiencing a pandemic similar to Covid-19 within an average lifetime.
This risk is based on past experience, but the world is changing. As deforestation, environmental destruction, and climate change worsen, more animals are pushed out of their usual habitats and the chances of another zoonotic infection increase. The probability of a pandemic could double in the coming decade.27 Then there are other, more speculative threats that we can’t place clear numbers on, such as the creation of a powerful AGI or an engineered virus.
All these threats add up. Even just one, like nuclear war, might seem low at 1 per cent per year, but becomes unnervingly high over the space of a century. We are not just facing the threat of nuclear war, but of all these agents of catastrophe together. These drivers also interact. For instance, extreme climate change both heightens the risk of a mass extinction event and inflames geopolitical tensions.
If such a catastrophe does occur it may be far more difficult to recover from than is commonly assumed. That is because the energy ratio of fossil fuels is drying up.
FALLING FOSSIL FUELS On 10 January 1901, the Texas Oil boom began. An oil drill tapped into the Spindletop Hill reservoir, sending a forty-five-metre geyser of oil into the sky.
These ‘gushers’ can now be seen only in black and white photos. The fossil-fuel industry is instead having to resort to increasingly technical, complicated, and costly procedures, such as fracking for shale oil and gas.
They have also had to resort to extracting oil from dirtier, ‘unconventional’ sources like tar sands. Tar sands are seas of loose sand that contain bitumen (a kind of crude oil) mixed with clay, sand, and water. Tar sands are hard to extract oil from and produce around 31 per cent more emissions than conventional oil wells.28 And fossil-fuel companies are also turning to machine learning to locate harder-to-find sources of coal, oil, and gas.
While these techniques have expanded the supply of fossil fuels, they’re also more costly, in both money and energy.
Over the years, it is taking more oil to extract a barrel of oil from the ground. In the US, by 2007 the energy ratio of oil and gas (that is, the number of barrels that can be extracted from the ground using one barrel of oil) had plummeted from 1200:1 to 5:1. That is a 240-fold decrease.
Globally, there is a race between a decline in the energy ratio due to the depletion of fossil fuels and technological efforts to increase it. Right now, depletion is winning.29 The energy ratio of coal is expected to fall similarly by 2040 at the latest.30 One study looked at the ratio at the final stage, once energy reaches the economy, rather than when it is first extracted from the earth. The researchers suggests that we are overestimating the current energy ratio of fossil fuels. By their calculations it is currently just 6:1 worldwide and declining.31 Future projections are no less comforting. By 2050 the energy ratio for oil could fall to 1.5:1.32 The long-term energy ratio of gas is also set to fall dramatically.
33 The trajectory of the energy ratio has led to fears of an ‘energy cliff’ or ‘energy cannibalism’. If the energy ratio drops to such a level then the world can no longer maintain its current production or affluence. One solution would be to move to renewable energy. The downside is that the energy ratio of renewables is not particularly high compared to the fossil fuels extracted during the nineteenth and twentieth centuries. The upside is that they are still reasonable (approximately 10:1 for wind and solar) and increasing owing to technical improvements.34 If we look at the energy ratio for when energy reaches the economy, then renewables may already offer better returns than fossil fuels. All technologies tend to reach a point of diminishing returns and stagnate, but hopefully the improving energy ratio of renewables continues for decades to come.35 It is difficult to overstate how important this trend – and the race between the energy ratio of renewables and the falling energy ratio of fossil fuels – may be. Progress over the past two centuries has been dependent on vast infusions of low-cost fossil fuels. We often don’t see it since we are largely ‘energy blind’ to how much coal, oil, and gas goes into the food we eat and the systems that warm, cool, and illuminate our homes.36 The first wave of industrialization was fuelled by geysers of oil and seemingly bottomless pits of easily accessed coal. These low-cost sources are no more. If the world collapses, it may be impossible to re-industrialize. THE RUNGLESS LADDER Think of our global society as a poorly built ladder. As we climb, the rung we have just used falls away. The higher you climb, the more dangerous the fall. Eventually, any drop from the ladder becomes fatal. The missing rungs represent the heightened severity of collapse due to more powerful weapons, more disease burden, a more deteriorated global environment, and a reliance on industrial agriculture, states, increasingly complex technological systems, and global systems suffering from the Death-Star Syndrome. The first farmers of Europe had only axes and arrows to fight with, had room to disperse, and could slowly transition towards doing more herding and foraging and less farming. Their fall, when it came, was piecemeal, gradual, and from just a few rungs high. It was often even beneficial. Now the fall is likely to be far more bruising.
The global technological and environmental threats we face today along with our hyperconnected and hypercoupled system breed new vulnerabilities, making it easier for big shocks to travel across both borders and sectors, gaining in power as they do. And some of our attempts to ward off disaster build latent risk into the system. The Global Goliath is creating new, unprecedentedly destructive threats, new vulnerabilities through systemic risk, increasing the ability for regional shocks to amplify into global ones, and inflating latent risk. Collapse today looks very different from the liberation of Cahokia or the decay of Rome. We are facing a genuine global catastrophe that could end our industrial world and much of life with it.
In chess, the endgame refers to the stage where only a few last moves are left. The world of Goliath that started around five millennia ago also appears to be reaching its endgame (Figure 43).The endgame is when we face the genuine possibility of existential risk: deep, potentially permanent, global collapse or even human extinction. In the next few decades or centuries, it is highly likely that the system will self-terminate unless it is fundamentally reformed. For the first time, there is the alarmingly high risk that Goliath will destroy itself permanently and may even take our species with it. Such high stakes naturally raise the question of who or what is to blame for this precarious predicament.
20 The Roots of Our Endgame Are We All to Blame? – The Agents of Doom – Crafting Threats in the Dark – Make Your Own Nightmare – Hate the Game and the Players – Root Causes – The Tree of Doom ARE WE ALL TO BLAME?
It’s easy to think that we are all to blame for any future collapse. That we find ourselves in the precipitous position of facing unprecedented climate change, thermonuclear war, systemic risk, and dangerously rapid technological change due to human nature or each of our everyday choices.
This idea of collective blame did not spontaneously appear; it was planted into the public conversation by a few powerful actors. This can be clearly seen in the case of climate change. In 2004 the fossil-fuel giant British Petroleum (BP) hired the public relations firm Ogilvy & Mather to help deflect responsibility away from itself and onto consumers. As part of this ‘beyond petroleum’ media campaign (which cost over $100 million) they popularized the term ‘carbon footprint’, a measure of how much carbon individuals emit going about their everyday activities, whether it be eating breakfast or taking the bus to work. They even developed their own carbon footprint calculator to help everyone estimate just how ashamed they should be of their lifestyle. In 2019 they revamped their efforts through a ‘Know Your Carbon Footprint’ campaign. While individual action is, of course, important, the aim was always to shift responsibility onto the public.1 It was a strategy that was more broadly shared. One analysis of 212 documents from Exxon, another leading fossil-fuel producer, found that they had systematically framed climate change as a problem of individual culpability. For instance, in 1998 Exxon asked the public to combat growing emissions by exercising ‘a little voluntary “can do”’, before merging with Mobil to become the largest oil company in the world, selling 5.4 million barrels of oil per day and reaping an annual revenue of $137.2 billion. For decades, oil producers have framed rising fossil-fuel demand as inevitable and placed the burden of cutting emissions squarely on the shoulders of the masses. It was only in internal memos and articles in littleread scientific journals that they admitted their own culpability.
2 The argument that we are all to blame occurs again and again.
Sometimes we are all to blame because of our desires: people endorse nuclear weapons because they want security or are content with dangerous algorithms because they want convenience, entertainment, and economic growth.
Sometimes we are all to blame because the root of our problems is human nature itself: nuclear arms races and a lack of cooperation on climate change are because of our innately parochial, suspicious, and self-interested nature.3 But blaming the public is a manufactured distraction – a PR myth that dissolves as soon as the hard figures on who produces catastrophic risk are examined. The reality is that a mere handful of giant corporations, countries, and militaries are responsible for creating the great majority of catastrophic risk, shrouding their activities in secrecy to avoid scrutiny, and undermining efforts to regulate them.
Think of this small number of companies and institutions – big tech, the fossil-fuel industry, and militaryindustrial complexes (a collection of militaries and the arms manufacturers they tend to contract to) – who disproportionately contribute to the apocalypse as the ‘Agents of Doom’.
Agents of Doom are in turn shaped and driven by larger incentives. At root, there lie just a few causes that are embedded in both the earth and our skulls. We can trace the causal chain of catastrophic risk from these roots via the primary producers of risk to the endgame we face today (Figure 44). Most catastrophic risk doesn’t emerge randomly but is produced by just a few actors for profit and power. Figure 44: The causal chain of global catastrophic risk We can think of global catastrophic risk as being part of a chain of cause-and-effect. For humanmade threats, this causal chain goes back from catastrophic risk today to the actors producing it, through the structures which shape their behaviour, back to the deepest root causes which have created those structural drivers.
THE AGENTS OF DOOM There’s a good reason that Exxon and BP were so eager to shift the blame onto individuals. Both are part of a cohort of just thirty companies that between them are responsible for a third of global energy emissions.
No matter how we slice up responsibility, these fossil-fuel giants are some of the key drivers of climate change. Over half of the greenhouse gases pumped into the atmosphere from 1854 to 2022 stem from just seventyeight corporate and state-owned companies. Since 2016, with the legal adoption of the Paris Climate Agreement, 80 per cent of the rise in emissions has come from just fifty-seven fossil-fuel and cement companies.4 It’s not due to a lack of knowledge. The greenhouse effect was discovered in 1856 by US scientist Eunice Foote and we’ve had a highresolution understanding of climate change since the publication of the first IPCC report in 1989.5 These companies are well aware of the damage that each tonne of carbon is causing, yet they are choosing to ignore it.
It’s a similar picture regarding countries. Just ten states (counting the EU as one) are responsible for around 80 per cent of cumulative global emissions since the Industrial Revolution. The US alone accounts for about a quarter of greenhouse gases presently in the atmosphere.6 It’s a trend that is consistent over time. Even today the top ten emitters account for roughly three-quarters of global emissions.7 Two countries, the US and Russia, hold 85 per cent of stockpiled nuclear warheads and continue to develop their technological advancement in delivery.
8 The production and maintenance of those weapons is carried out mainly by a small group of twenty-four companies who operate largely out of the public eye but are financed by 287 financial institutions (such as banks and, ironically, pension funds and insurance companies).9 Biological weapons, now officially outlawed, followed a similar historical pattern from the First World War onwards. As with nuclear weapons, it was the most powerful and technologically advanced states – eleven countries such as Japan, Russia, Germany, the US, and the UK – that created biological weapons. Since then, there have been around eighteen incidents from biological weapons, the worst being Japan’s attacks on China during the Second World War. It remains states and large biotech companies who have the necessary resources, equipment, and expertise to create new dangerous pathogens.
This tendency for a few powerful actors to produce new world-damaging technologies holds true for AI as well. The frontier of large text models is largely concentrated in three key companies: Microsoft’s OpenAI, Google DeepMind, and Anthropic. Microsoft and Google rank third and fifth as the richest companies in the world. They are all headquartered on the US west coast (although Google’s DeepMind itself is in London).
That means they all fall under the jurisdiction of one country: the US.
Even if we cast a wider net beyond large text models, we still see just a small number of actors trying to build AGI. In 2020 over seventy projects were explicitly trying to build an AGI, around half of them in the US, as were eight of the nine projects with military connections.10 There are just a handful of powerful actors that are producing the world’s greatest threats. These Agents of Doom are large fossil-fuel companies, bigtech titans, and the military-industrial complexes of a few empires such as the US, China, and Russia. One could, of course, argue that, while the creation of these global threats stems from just a few agents, they are responding to the demands of the masses. It’s another myth that doesn’t hold up to scrutiny. CRAFTING THREATS IN THE DARK Democracy dies in the dark. It is exceedingly difficult to democratically regulate any activity that is kept out of the public eye. Yet, that is exactly what is done when it comes to the production of the world’s greatest threats.
It is an approach that is, despite their efforts to hide it, all too visible in fossil-fuel companies, who have known about climate change and its potential dangers for decades. The largest trade association for the oil and gas industry has known since the 1950s, the coal industry since the 1960s, and fossil-fuel giants such as Total and Exxon since the 1970s.
During the 1970s Exxon made remarkably accurate internal forecasts of future global warming, as did many of their employees in scientific journals. In 1979 an internal Exxon study even admitted fossil fuels were causing climate change and could be responsible for ‘dramatic environmental effects before the year 2050 … the potential problem is great and urgent’. Rather than alert the public, they buried the evidence.
Then they collaborated to cast doubt on the science of climate change.11 Fossil-fuel companies have since at least the 1970s run well-funded disinformation campaigns to undermine action on climate change. They funded conservative think-tanks with millions of dollars, which in turn financed and gave a prominent platform to scientists infamous for falsely challenging the scientific consensus on issues such as the link between smoking and lung cancer, the effects of acid rain, and the existence of the hole in the ozone layer. They included the tobacco lobbyist Frederick Seitz and William Nierenberg, who was both a scientist in the Manhattan Project and co-founded the conservative think-tank the George C. Marshall Institute. Oil, coal, and gas companies didn’t just fund and platform professional contrarians who were sympathetic to their cause; they even helped set up and fund fake grass-roots organizations that pushed against the science of climate change.12 Much of this strategy was coordinated by the International Petroleum Industry Environmental Conservation Association (IPIECA) lobbying group, which has over forty corporate members including BP, Shell, Chevron, and ExxonMobil.13 This deliberate campaign of misinformation successfully set back the public discussion (and action) on climate change by decades. The fossil-fuel companies are not alone in their deception. While microplastics and PFAS may be novel, the dangers were not unforeseen by the companies producing them. The largest industrial producers of PFAS such as 3M and DuPont were running safety studies on these compounds for decades. By 1970 they had established that the chemical C8 (a type of PFAS known as PFOA) was ‘highly toxic when inhaled and moderately toxic when ingested’.
They didn’t publish the findings. Instead, a decade later 3M announced to the public that C8 was ‘about as toxic as table salt’.
Their internal findings included reports on the possible links between PFAS and prostate cancer that were shared between 3M and DuPont, and records of children with eye and facial defects born to workers exposed to PFAS. While the manufacturers shared this information among themselves, they did not share it with the public. Instead, these studies on the negative health impacts were intentionally buried and suppressed.14 They and others continued to insist that these products were safe for decades, sought approval for them from the Environmental Protection Agency (EPA), and denied the adverse health impacts of PFOA. The producers knew these chemicals were potentially toxic forty years before the public health community yet chose to endanger the public in pursuit of profit. Their disinformation campaigns and lobbying efforts would never have been possible if it had not been for their own internal secrecy.
15 Secrecy has also been central to the production of nuclear weapons.
The public has been largely kept in the dark since their very invention.
Both the Manhattan and Castle Bravo projects of the US are now almost synonymous with ‘large, secretive government project’. The initial classification system was so extensive that even basic physics published in textbooks were considered ‘born secret’ (i.e. classified as soon as they were written). The obsession with secrecy even led to the Atomic Energy Commission burning hundreds of copies of an edition of the Scientific American in 1950. Since then, nuclear weapons have remained staunchly overclassified and outside the domain of public oversight.16 A key point of US secrecy was to prevent proliferation. It failed. Eight other countries subsequently developed atomic weapons after the US did, while numerous others could have (and had the knowledge to do so) but chose not to. Similarly, secrecy played no role in having Ukraine give up their inherited stockpile or South Africa dismantling theirs. What secrecy has done instead is handicap public discussion of nuclear weapons. The clandestine approach of the US has been replicated elsewhere. Israel has not admitted to having nuclear weapons, although it is an open secret that they do. Some believe that powerful states still have illegal biological weapons capabilities that have successfully been kept secret.
It’s not just the threats themselves that are being hidden; it’s often also crude corruption, lobbying, and profiteering in the companies and organizations. One of the most common approaches is the creation of ‘revolving doors’ in which individuals move between lucrative positions in industry (or lobbying for industry) and powerful political positions which are supposed to regulate that industry. One illuminating example comes from the US nuclear-defence industry. The former senator for Arizona Jon Kyl was one of the key politicians who pushed for more money to be spent on the multi-trillion-dollar US nuclear modernization programme in 2010.
In 2013 he left office and joined the lobbying group Covington & Burling, working on behalf of the defence contractor Northrop Grumman, the same company that had received the contract to modernize the US arsenal.
Northrop Grumman also spent $1.55 million on political donations in 2020, while its Political Action Committees (PACs) doled out $3.77 million to both Democrats and Republicans. Together, Boeing, Raytheon, Northrop Grumman, and other aeronautical defence companies spent $46.9 million on lobbying.17 Kyl is just one example of dozens of generals and politicians who push for increased military spending while in office and then cycle into handsomely paid positions with defence contractors (who benefited from their actions) afterwards. It is an endemic problem that amounts to legalized bribery: making the right moves while in office can guarantee a career with a six-figure salary in industry later.
It also doesn’t just happen in the US military. One report in the EU found that six fossil-fuel companies (Shell, BP, ENI, Total, Galp, and Equinor) and five of their lobby groups were involved in seventy-one ‘revolving door’ cases over the period 2015–21. During that time, they also averaged six meetings a month with the European Commission and spent €170 million on lobbying.18 In Australia, there is the well-documented case of how oil, coal, and aluminium lobby groups – a coalition who call themselves ‘the Greenhouse Mafia’ – have shaped the country’s energy and environment policies.19 There is also another revolving door between politics and industry. Every resources minister in the Australian government since 2001 who has retired from parliament has taken a position in the mining, consultancy, or fossil-fuel industries after finishing their time in office. Since 1990 more than a quarter of federal or assistant federal ministers have taken jobs with special interest groups after their political careers ended. It’s why Australia, the world’s sunniest and second-windiest continent, has no price on carbon emissions and also has some of the highest emissions per person in the world. Action on nuclear weapons, climate change, and other global threats are the products of intentional efforts to undermine democracy and capture state power.
We see the same revolving-door problem even when it comes to responses to pandemics. In 2021 the Moderna vaccine was approved by Stephen M. Hahn, the head of the US Food and Drug Administration (FDA). Moderna was created by Flagship Pioneering, a bioengineering venture-capital firm that spits out enormous numbers of start-up firms (with the strategy of having the vast majority fail but winning big on one or two).
The vaccine wasn’t even a Moderna invention: it relied on innovations from the US government-owned National Institutes of Health (NIH) as well as Scripps and Dartmouth universities. Moderna was then given $10 billion in taxpayer-funded money before making $36 billion in profit.
Hahn left the FDA a year after authorizing the vaccine and joined Flagship as its chief medical officer. He remains there as CEO-Partner.
This happens not just with Flagship. One of Hahn’s predecessors, Scott Gottlieb, moved from being head of the FDA to being on the board of the pharmaceutical company Pfizer. Between 2001 and 2010 over a quarter of FDA employees who approved cancer and haematology drugs went on to work in or consult for biopharmaceutical companies.20 It is little wonder that the US public has such a deep distrust of public institutions and why there has been such a fertile ground for vaccine conspiracy theories. The revolving door also makes proper regulation of biotechnology far less likely: if you are going to go and work in the industry afterwards then you have far less incentive to burden them with legislation to improve risk management. Those revolving doors and legalized corruption undermine efforts to reduce and respond to global risk. Survey after survey has shown strong public support for reducing catastrophic risk. One survey across six countries found that a strong majority in each one, including Israel and the US, supported the creation of an enforceable international agreement that would eliminate all nuclear weapons. Europeans in particular heavily endorse a ban on nuclear weapons. For years, opinion polls across the US, EU, China, Australia; and others have shown substantial support for greater action on climate change. One poll covering twenty-six countries found that a majority in each country opposed the development and use of lethal autonomous weapons. If anything, these surveys underestimate public support, since they are just gathering the opinion of people who often aren’t fully informed (especially given government and corporate secrecy). Open democratic experiments where citizens are convened in a jury (of 9–12 people) or assembly (of a few hundred), are briefed by experts, and then deliberate, tend to show even stronger support for aggressively tackling global risks such as climate change.21 Whatever the threat, we see the same pattern of a few Agents of Doom profiting from creating global catastrophic risk while cloaking their actions in secrecy and extensively lobbying to prevent regulation.
MAKE YOUR OWN NIGHTMARE In 2023 when a long list of academics and corporate leaders signed a letter warning of the risk that machine intelligence could wipe out the human race, a few names stood out. Sam Altman, Demis Hassabis, and Dario Amodei are the leaders of the three largest, most advanced AI companies in the world, constructing the most dangerous algorithms.
They continue to fret publicly about the machines they are building.
Amodei, CEO of Anthropic, thinks there is a 10–25 per cent chance of AI destroying humanity; Altman, CEO of Open AI, has worried that the ‘development of superhuman machine intelligence is probably the greatest threat to the continued existence of humanity’. There are dozens of other remarks from AI pioneers and producers warning that it could cause human extinction. Yet these tech giants continue to pursue an invention they believe could destroy all humanity, perhaps all life on earth. Even though Amodei thinks there is around a one in four chance that the algorithms he is helping to construct could murder not only all his closest family and friends, but also you and everyone you know, every single one of us, he continues to go to work every day.
22 Not only are these leaders ostensibly making their own nightmare while warning the world about their actions, but they are also actively thwarting attempts to better oversee and regulate the creation of dangerous algorithms. Whistleblowers at OpenAI have warned that employees must gain permission from the company first before they contact the media. That’s the equivalent of asking a local mafia for permission to tell the police about their drug deals. Employees leaving the firm also need to sign highly secretive non-disclosure agreements or else they will lose their shares in the company. They are also bound by ‘non-disparagement agreements’, which essentially mean that employees who have left the company cannot say anything critical of OpenAI, again at risk of losing their shares. It seems to be a widespread practice, with past and present employees at both OpenAI and DeepMind having penned an open letter calling for greater whistleblower protection and safety measures. ‘Open’AI appears to be ironic at best.23 While coders get most of the attention, the role of lobbyists should not be underestimated. From 2022 to 2023 the number of groups lobbying the US government about AI almost tripled. These lobbyists are, perhaps unsurprisingly, not pushing for tighter regulation to prevent human extinction. On the contrary, Google, Meta, OpenAI, and TechNET (an industry-wide lobbying group representing Google, OpenAI, Anthropic, Amazon, and others) have all openly opposed SB 1047, a regulation under the California State Legislature. The bill would require risk assessment of any models that cost more than $100 million to train and are bigger than 1026 FLOPS (a measurement for the number of calculations a computer can perform in a second) before they are deployed to ensure that these programs can’t cause catastrophic harms by, for example, carrying out cyberattacks or helping to build nuclear weapons. It would also provide whistleblower protections. In September 2024, Gavin Newsome, the governor of California, vetoed the bill. He did so largely because of immense industry pressure from the likes of Meta, OpenAI, and Google, all of whom cautioned that the bill would lead to an exodus of AI firms from California. The same companies warning that their machines could threaten human extinction successfully helped veto policy that would better control these machines. They are crying out for regulation in the abstract while killing specific policies which could regulate them.24 Perhaps that’s because people like Altman don’t truly fear their own creations. Warning of an apocalypse is instead a way to build hype about their products while gaining more leverage over regulators. Their plea to governments is that if you want to regulate a doomsday machine then you need the advice of those building it. Their motive is profit and power, not safety. There are two alternative, more charitable explanations, ones that the developers themselves have offered. One is that the technology is so powerful that there is a moral imperative to build it. AI has been declared a potential solution to extreme poverty, energy shortages, climate change, and essentially every other global challenge faced by humanity.
This belief that a technology will create utopia is by no means new.
Proponents of weapons of mass destruction have repeatedly claimed that more advanced weaponry could end all wars by simply making conflict too dangerous and inhumane.
That was said for artillery, dynamite, machine guns, submarines, smokeless powder, and nuclear weapons. Every time, that claim of using more powerful weapons to end war has been proved wrong.25 It’s also an argument that is more about a messiah complex (essentially a declaration that ‘I’ll decide whether the potential benefits justify the risk to humanity’) than altruism.
The other excuse offered by developers is that they need to build the potentially world-ending machine first because if they don’t, someone else will. And they are the (self-proclaimed) good actors who can build it safely and benevolently. It’s a view summarized in a 2024 op-ed by Sam Altman in which he framed AI as a race between democracies and authoritarianism.26 Often the developers use both utopian aspirations and fears of a race simultaneously as excuses. On another occasion Altman said ‘the upside of AGI is so great, we do not believe it is possible or desirable for society to stop its development forever’.27 Similarly, the machinelearning expert Geoffrey Hinton once responded to a question of whether the development of AI could be stopped with, ‘I don’t think that’s possible, and I think we should continue to develop it because it could do wonderful things.’28 The arms-race defence is interesting because it is so common. CEOs have used it to explain why they need to build AI, and to defend environmentally destructive practices; generals have used the same excuse to justify building nuclear weapons and killer robots; and presidents have used it to explain why they can’t cut emissions. The excuse is that if we don’t do it, someone else will, and it’s better if we ‘good guys’ do it first.
We shouldn’t allow this arms-race defence let the Agents of Doom off the hook. Even if you need to compete, no one is forcing chemical companies to bury the evidence on the harms of PFAS, or fossil-fuel companies to run scientific disinformation campaigns, or defence contractors to lobby and use revolving doors to push for bigger defence budgets and nuclear arsenals. That is clearly being done for the benefit of a small number of firms and individuals. Yet there are also deeper incentives at play here.
Unless we address those structural incentives, then regulating or penalizing the Agents of Doom will not be a lasting solution.
HATE THE GAME AND THE PLAYERS In 1928 the polymath John von Neumann wrote a paper called ‘On the Theory of Games of Strategy’ and started the now popular field of game theory. Its aim was to identify what would be the mathematically optimum course of action for a rational agent to take when trying to win a game, whether it be chess or war. Game theory is one of the commonly used explainers of arms races.
Arms races have often been explained in game theory as an example of a ‘security dilemma’. Imagine two neighbours, both self-interested, rational (they act logically to achieve their self-interested goals) individuals who prioritize their own survival.29 They would prefer a situation in which both remain unarmed. Guns are expensive and fatal accidents can occur.
However, they can’t trust their neighbour’s benevolence. The worst-case scenario is you have no weapons when your neighbour attacks. The better alternative would be if you get weapons and your neighbour doesn’t. If your neighbour has also chosen to purchase weapons, then you are of course relieved that you decided to stockpile armaments as well. Hence, both are incentivized to begin stockpiling weapons just in case. A move which is good for each actor individually is bad for the collective.
Globally, it’s often assumed that this security dilemma and subsequent arms races are the result of ‘international anarchy’. In the absence of one overarching arbiter or world government, states can’t help but compete with one another.
A similar game-theoretic explanation is often used to justify environmental destruction. Different competing groups overusing a common resource has often been referred to as a ‘tragedy of the commons’. The tragedy is usually posed as a thought experiment: imagine an openaccess good (a common) such as a sea shared between fishers. If all of them overfish, it will lead to the destruction of the resource.
However, if one of them does so, they can get away with it while getting a larger catch (at least in the short term). A few will always act selfishly, and those who practise restraint will be outcompeted. As with an arms race, individual actors pursue rational short-term concerns, and the result is collective tragedy.
This idea was first proposed by the American ecologist Garrett Hardin in a 1968 essay. Hardin wasn’t interested in environmental commons but was instead trying to argue for population control. Nonetheless, his essay has become a classic explanation for our environmental woes, from regional overfishing to global climate change.30 We can think of the tragedy of the commons as emblematic of a wider problem: a race to the bottom. Imagine a confectionery company that wants to create sweets without the involvement of child labour. Paying adults a fair wage will put altruistic firms at a disadvantage compared with those who opt for the cheapest but most exploitative labour conditions. The company with no scruples and a cheaper product will win out. Both arms races and races to the bottom are underpinned by game-theoretic logic: selfinterested, rational actors pursuing short-term, self-interested goals results in collective harm. It’s easy to see how such dynamics can be used to explain the actions of the Agents of Doom, increasing the likelihood of nuclear war or extreme climate change.
We have already seen the problem of a third type of race that underlies the others: status races. Whether it be between chiefs competing with feasts, rulers through monuments, or billionaires through high-emissions stocks, superyachts, and private jets, status races tend to drive conspicuous consumption and expansion. In the long run, it is driving us into an unpleasant collision with planetary boundaries. As we’ve seen, that race for status also partly underpins arms races and races to the bottom.
While appealing, there are a few issues with viewing the problem through this game-theoretic lens. While it’s mathematically neat, we now have years of studies showing that people aren’t as rational as game theory assumes they are. In laboratory experiments, irrational humans frequently fail to match the predictions of game theory. Rather than being rational and selfinterested, people often show strong signs of reciprocity, a concern for selfimage (strongly related to status), and an aversion to inequity (which often overlaps with an aversion to dominance).31 They often don’t end up in arms races. Similarly, if you put people in a tragedy-of-the-commons-type situation in a laboratory, they rarely end up enacting a tragedy. Simply allowing them to talk to each other prevents the tragedy of the commons from emerging.32 The tragedy of the commons also frequently doesn’t occur in the world outside of lab experiments. For instance, Hardin was quickly disproved by the economists Elinor and Vincent Ostrom, who showed that many communities (including non-state groups) regularly solve the tragedy of the commons. Groups such as the Khoisan and central African hunter-gatherers seem to have managed their environment just fine for hundreds of thousands of years.33 Our deep past also undermines the idea that security dilemmas plague all humans and human groups. Our Palaeolithic fluid civilizations lived in a state of ‘anarchy’, yet appear to have been peaceful. The first signs of organized war and arms races begin only with the rise of dominance hierarchies. There is also a wide array of different ‘peace systems’ throughout world history (including hunter-gatherers) that don’t fall into arms-race-type dynamics. They manage to create peace not just through governments but also through norms and culture.34 Yet, there is a reason that game theory seems convincing. It does seem to explain how many states and corporations today behave, as well as the struggle between elites and empires in the past. Clearly some groups fall prey to this dynamic far more than others.
Why? ROOT CAUSES There is a reason why arms races often happen across empires of the past and the Agents of Doom today but rarely occur among most individuals.
Empires and the Agents of Doom are not human beings; they are dominance hierarchies. They are artificial agents who act very differently from the average person. Hobbes, when he first spoke of the Leviathan state, defined it as ‘an artificial man, though of greater stature and strength than the natural, for whose protection and defence it was intended; and in which the sovereignty is an artificial soul, as giving life and motion to the whole body’. The courts were its artificial joints, the administration its memory, its economic riches were its strength, and collapse was its death.
Similarly, the corporation is a legal entity imbued with the rights of an individual. Legally, it is an artificial person. Alongside states and corporations, large-scale algorithms are fast becoming a third type of artificial agent, although one largely controlled by states and companies.35 Both states and companies are artificial creations: inventions that both amplify decision-making power and allow for a being that could outlive its creators. A state can enter agreements and take on loans, which it is expected to fulfil, even if the signing head of state has died. ‘Corporation’ is derived from the Latin corpus meaning body (which, ironically, is one thing they lack). The ‘state’ is derived from the Latin status, for a manner of standing or ranking. In perhaps the ultimate Freudian slip, the roots of the state are in status.36 States and companies (and perhaps now algorithms) are not just amplifiers of power and decision-making; they also inflate the worst in us: the darker angels of our nature. Think of a modern state, such as the US or Russia, embodied as a single person stalking the streets. Imagine what they are like. Overwhelmingly self-obsessed and self-interested, constantly boasting of their achievements, continuously suspicious of their neighbours, persistently posturing and hoarding guns and knives, and greedy to the extent of never giving even just 1 per cent of their earnings to those in need.37 That state person, if it took a psychiatry test, would rate disturbingly high on the dark triad (psychopathy, narcissism, and Machiavellianism). Bear in mind that Machiavellianism comes from the Florentine philosopher Niccolò Machiavelli (1469–1527), famous for his work The Prince, which advises leaders to be amoral to achieve political glory. A corporation as a person is much the same: status-seeking, profitfocused, self-interested, Machiavellian, and overly competitive.
When Hobbes thought about the state of nature, he appears to have been thinking of a world populated by these artificial agents, not with actual flesh-andblood humans.38 In both game theory and economics, the model used for humans (often called Homo economicus) resembles these artificial persons much more than the average human. They are self-interested, attempt to act rationally, and pursue material wealth.
Races for arms, resources, or status aren’t innate to human groups.
They are more likely to occur only among groups that exhibit some peculiar characteristics: dominance hierarchies marked by status competition, corruption by power, higher levels of the dark triad, and bureaucracies that aim for rationality (although they rarely completely achieve it).39 While game theory often does a poor job of predicting the behaviour of people as a whole, it does work better for smaller segments of the population. In laboratory experiments, individuals who rank higher on the dark triad (specifically psychopathy and Machiavellianism) are more likely to defect and not cooperate.40 Dominance hierarchies are not just less likely to cooperate and fall into harmful race dynamics; they also select for and amplify the individuals and characteristics that hinder cooperation.
This is not just about races and competition. Militaries and defence companies regularly trade, collaborate, and cooperate. In 2016 the Israeli NSO Group developed a powerful spyware weapon called Pegasus that could covertly hack into phones and laptops. It didn’t share this only with allies but also with ostensible enemies such as Saudi Arabia and the United Arab Emirates. These sales needed formal approval by the Israeli Defense Forces. Today the US accounts for around 42 per cent of global arms exports, including to its rivals. Profit was, and often is, deemed more important than national defence.
Countries and companies don’t just compete and race; they collude and cooperate.41 The same is true of the global ecological crisis. Fossil-fuel companies and PFAS manufacturers are technically competitors. They are fighting each other for a bigger market share. Yet, when necessary, they will lobby and work together to suppress science and undermine policy. This witches’ brew of competition and cooperation is a common feature of racketeering. We saw earlier how the razones gangs of Columbia competed for territory and drug sales, but still governed among themselves and could combine forces against a common enemy, such as the police.
They weren’t just racing against each other to control turf and the drug market; they were part of a wider racket that was trying to maximize extraction. Similarly, emperors and competitors for the throne regularly cooperate to tax the peasants or fight off invaders, all while still feuding.
It is equally accurate to describe this as a racket rather than a race. We can see the modern world in a similar light. As states mainly developed as protection rackets to begin with, so too do Agents of Doom both produce the risk and then benefit by offering the solution. Even OpenAI began with the ostensible goal of creating ‘safe’ and ‘aligned’AI. They are rackets made up not by communities or individuals, but by very strange artificial agents.
THE TREE OF DOOM The root causes of our endgame were put down when the darker angels of our nature (status competition, power corrupting, the dark triad, and the authoritarian impulse) met the new environmental conditions of the Holocene (such as lootable resources). Goliaths began to emerge. At the individual level, those who were the strongest status seekers (and often the most adept at violence and manipulation) were selected for. At the group level, communities that were more effective at conquest tended to outcompete their neighbours and grow. It was multi-level selection. These new groups even invented myths to valorize growth and conquest. From the civilizing mission of European colonizers to the jihad of the Islamic caliphate, dominance hierarchies have had ideologies which justify universal expansion.
These new Goliaths amplified our darker angels. They fell into competitive ‘Goliath traps’: arms races, races to the bottom, and status races using conspicuous consumption. Yet these weren’t just races; they were rackets. Groups who were competing to extract resources from a given population would regularly trade (as with the Late Bronze Age or today) and band together when necessary. Those races/rackets, whether it be for protection or the extraction of data or fossil fuels, are the main drivers of our endgame today (Figures 45 and 46).42
Goliath fuel and our darker angels are the root causes of our endgame.
They spurred the creation of Goliath systems which became stuck in different, interlinked races whether it be for status, security, or a competitive advantage. Today these Goliath traps have branched out into all the contributors to global catastrophic risk: nuclear weapons, AI, killer robots, climate change, the infringement of planetary boundaries, and others. This tree, from the root causes to our unfolding endgame, is still growing.43 Unfortunately, it’s difficult to see how this thousand-year-old process is going to end well.
21 The Fates of Goliath Putting a Date on Doomsday – Our Curse – Evolutionary Suicide – When Ideas Dry Up – Silicon Goliath – Killing Goliath – Choose Your Curse PUTTING A DATE ON DOOMSDAY Table 6: Apocalyptic probabilities Probability Timeline Conditions 49 per cent 2050 Collapse 50 per cent 2100 Collapse or extinction >25 per cent 2100 Existential risk 19 per cent 2100 Extinction 16 per cent (1 in 6) 2100 Existential risk The most common way of trying to calculate when doomsday might arrive has been for experts to guess what our odds are of making it past 2100 (Table 6). The Astronomer Royal Lord Martin Rees says it’s a coin flip between survival and collapse or extinction. Jared Diamond is even more pessimistic, giving us a 49 per cent chance of collapsing by 2050.
University of Oxford researcher Toby Ord believes it is more like Russian roulette, our chances of being wiped out within the next eighty years being one in six. His colleague Nick Bostrom believes it is at least one in four.
Averaging these experts’ opinions gives us a one in three chance of facing a global collapse, or something far worse, by the end of the century.
1 These numbers are frighteningly high. They also tend towards suspiciously round figures and clichés, such as 50/50, a quarter, or Russian roulette. This is just one sign that they aren’t to be trusted.
These are all examples of unreliable forecasts. Superforecasters (the gold standard of forecasters) draw on an exceptional group of experts with diverse information to make predictions about specific, concrete questions. Even superforecasters have little predictive ability beyond a year or two and there is little evidence that they have any accuracy after a decade.2 These apocalyptic predictions are being made about a vague question, looking decades ahead, usually by individuals who aren’t superforecasters. Don’t trust any particular date on doomsday. 3 While we can’t put clear numbers on the likelihood of collapse, what we can do is understand the most likely future scenarios based on the characteristics of Goliath and how they have changed over time.
OUR CURSE The Global Goliath we inhabit has its own unique fuels. The most powerful monopolizable weapons are now nuclear: the US and Russia achieved superpower status by having bigger, more advanced nuclear arsenals (and militaries) than anybody else. More than that, the decision to commit the world to a war that could cost billions of lives lies in the hands of one individual: the US president. It is the most consequential and important decision any leader could make, yet no democratic deliberation or approval is required. It has led some to brand the US a ‘thermonuclear monarchy’.4 The dead-hand in Russia is, of course, no better. These monopolizable weapons have tilted the world away from democracy.
The lootable resources that enable and power the Global Goliath are fossil fuels. Fossil fuels, like grain, are concentrated, defensible, and easily monitored, moved, and stored. Unsurprisingly, they have regularly been appropriated at massive scale by states and corporations. Most of the world’s flows of oil, gas, and coal are controlled by a few companies that are or were often state-owned. Saudi Aramco is Saudi Arabia’s state-owned oil and gas company, the gas titan Gazprom was previously owned by Russia (which maintains a majority share), and Royal Dutch Shell is the merger of Royal Dutch Petroleum and Shell Transport and Trading Company (the former was strongly connected with the Dutch government, and the latter to the UK).
Also, like grain, fossil fuels are energy-dense and have permitted a surge in population. Global population has ballooned from 1 billion prior to the Industrial Revolution to 8 billion today, fed in large part thanks to synthetic fertilizers derived from fossil fuels: grain-based agriculture on petrochemical steroids. Hence, lootable resources tend to build on each other rather than replace each other: fossil fuels enabled larger harvests of grain rather than replacing them. Today, every calorie we eat was produced using around ten calories from fossil fuels.
Land is also now more caged than ever. The world is carved into precise borders that are often hawkishly monitored. Some are physical: Israel, Egypt, India, and Saudi Arabia have all erected border walls and fences covering thousands of kilometres and surpassing the efforts of Rome, China, and Persia.5 We now also have invisible cages: identity cards, passports, and mass surveillance systems have all become new ways to control borders and people.
Most people alive today have been lucky enough to inhabit a world in which many of the most powerful countries at the core of the Goliath were, by historical standards, inclusive and democratic. This was in large part due to the two twentieth-century world wars, which destroyed enormous amounts of capital that was primarily held by the wealthy.
Governments began to tax the rich more aggressively to bankroll the war effort and put in place measures to appease workers (who were now armed, and upon which countries were entirely dependent) such as rent, wage, and price controls.
Unionization rates peaked shortly after the Second World War. Scholars have come to call this great levelling of wealth the ‘Great Compression’.
The world wars didn’t just level wealth. Wealth equality spilled into the other forms of power and made countries less extractive in general. Key welfare-state provisions were passed directly afterwards, and democracy was strengthened. Conscription during the world wars was frequently followed by extending the vote in many countries, including to women in the US and UK directly after the First World War.
6 It would be a mistake to think that this was a lasting and global transformation. The global political and economic system is still grossly unequal. Wealth inequality began to creep back up across many countries in the 1970s. Now the richest 1 per cent of the world own almost half its wealth, while just eighty-one billionaires possess more wealth than the poorest half of humanity. The situation appears to be worsening. The share of global wealth for the top 1 per cent swelled from 25–30 per cent in the 1980s to approximately 40 per cent in 2022. For the US, China, and Europe, the share of the top 1 per cent rose from 28 per cent in 1980 to approximately 33 per cent in 2022. In contrast, the share of the bottom three-quarters has continued to hover around 10 per cent. These numbers are underestimates. They don’t capture the money and capital siphoned into overseas tax havens, for instance (Figure 47).7 Wealth inequality has spilled over into our politics globally. Democracy is backsliding. The number of people living in democracies has been decreasing and the number living in autocracies has been increasing over the last fifteen years. Today 5.7 billion (around 71 per cent of the global population) live in an autocracy. Forty-two countries are currently becoming more autocratic while just eighteen are becoming more democratic by a measurement that includes sixty indicators such as the fairness of elections, independence of the media, and equality of access to resources. Such democracy measurements don’t include workplace democracy or even whether the most pivotal decisions – like whether to launch a nuclear weapon – are made democratically.
The democracy and public services we often take as being the natural condition of states are an historical aberration. One that is slowly being undone as the world drifts back towards the historical default of greater inequality, autocracy, and extraction of wealth and power by elites. There may be a lag (as there was between wealth inequality rising and democratic backsliding), and some reversals, but in the long run we can expect wealth inequality to spill into the other forms of power. It is happening today. Elon Musk is not only one of the world’s richest men but also owns X, wields enormous cultural clout, played a starring role in getting Donald Trump re-elected, and will be instrumental in making policy during Trump’s second term. He is melding together all the different forms of power. Similarly, Amazon founder Jeff Bezos is not just one of the world’s richest men; he also owns the Washington Post (the third largest newspaper by readership in the US), audiobook giant Audible, the streaming service Twitch, and the film company MGM.
Similar patterns have been repeated throughout history, including the gilded age of the late nineteenth century and the rise of autocracies in the 1930s. The record is simply replaying today.
Figure 47: Inequality in key empires across the twentieth century All major countries share a rough trend of inequality peaking before the world wars, plummeting during them (especially during the Second World War), and then rising again from the 1970s onwards. Wealth inequality follows a similar pattern but at a higher rate. Today, the top 10 per cent capture around half of global income, but three-quarters of global wealth.
The curse of our Global Goliath is not necessarily diminishing returns on extraction, though the US does show many signs of this. There remains intense elite capture of politics and factionalism. In 1983 the number of US households worth over $10 million was 66,000. It increased (in real terms) within a few decades to 693,000 in 2019. Since the late 1970s real wages for most of the population have remained largely stagnant (or even declining) despite GDP per capita more than doubling. As Peter Turchin has shown, the role of popular immiseration and inequality in driving populism and political division is far too apparent.8 The enormous military overstretch of the US has helped fuel a growing public debt of $35 trillion.
Yet, this is not the most worrisome curse. The US may decline: a civil war could even break out, although this seems unlikely, given its large full-time, well-equipped police force, military, and intelligence agencies.
Even if it were to happen, the Global Goliath might be shocked for a few years or even a decade, but it is difficult to imagine how it would trigger a global societal collapse. And the US staying as a stable, coherent society is also not an antidote to our global woes. Remember that it was during the 1950s, 60s, and 70s, when elite competition was low and wealth equality higher, that the US built a nuclear stockpile that could destroy the world and had the largest acceleration in its environmental impact. Diminishing returns on extraction remains important and relevant. Yet there is now a new and more dangerous curse: global existential risk. The global system of extraction now creates threats and vulnerabilities which could destroy the whole world.
EVOLUTIONARY SUICIDE The most likely fate of the Global Goliath is self-termination. The pressures that have been selected for over time – larger populations, more disease, greater military capacity, and economic exploitation (categorized earlier as ‘babies, bombs, bacteria, and barbarism’) – may have been evolutionarily advantageous for individual empires, but now underpin the catastrophic risk we face today. Bigger, more interconnected populations mean worse disease outbreaks; stronger militaries mean more damaging arms races; and barbarism is the economic backbone of the systems that produce carbon, weapons, and dangerous technology.
It is a persistent and long-running trend that is difficult to stop. Goliaths have reliably turned economic power and larger populations into worse weapons. Human history is one of growing tail risk. Take the history of projectile weapons. The power and speed of projectile weaponry have grown consistently over time. An average modern-day nuclear weapon is fifty trillion times more powerful than a bow and arrow (as measured in joules of energy released). When we plot the most effective projectile weapons of different ages, we see a steadily rising line, with a sharp uptick in the twentieth century (see Figure 48).9 GDP shows a similar trend over that same period. Increases in economic power have reliably been translated into more potent weapons.
There is no sign of this trend abating. Even today we see the mass development of killer robots and other even more disturbing weapons, including the weaponization of space. States are now competing to create and unleash space-based directed-energy weapons (DEWs), such as ultrashort-pulse lasers. The end goal is to have these space-weapons establish an orbital ‘kill web’ that can identify and eliminate targets anywhere on the planet.10 Figure 48: Energy released by projectiles over time The shape of this graph will be familiar by now. It is similar to what we’ve seen with global GDP, land use change, and energy consumption. More energy and money have been reliably transformed into more destructive weapons.
We see the same trend when we look at environmental impacts over time.
The problem is that the selection pressure for Goliaths has created entities more effective at extraction, whether it be from people or the environment.
Goliaths have evolved new forms of Goliath fuel – such as fossil fuels, data, and nuclear weapons – to persist and grow larger over time. Yet now those new sources of fuel threaten to destroy the entire planet.
Think of this as evolutionary suicide: short-term evolutionary pressures lock the population into a path of self-destruction. Short-term adaptation leads to long-term extinction. We see such a pattern with cancer (which destroys its host and itself in the process) and in laboratory experiments where profligate bacteria change the pH of their environment, slowly rendering it inhospitable and committing themselves to extinction.
Given how deeply locked in are Goliath traps (status races, arms races, and races to the bottom) and other problems, it is hard to see how we can escape without a global collapse unfolding.
Technology is usually mooted as the potential saviour from evolutionary suicide. Bioterrorism can be prevented through mass surveillance and nuclear weapons can be deterred by better nuclear weapons. Many powerful figures now hold up AI as a solution to all our woes. Demis Hassabis explained his motivation for developing AI as to ‘solve the problem of intelligence and then use intelligence to solve everything else’. Some are even willing to take the bet that technological progress will outpace and solve any problems it creates along the way. Ex-Google CEO Eric Schmidt said in 2024, responding to a question about the emissions and energy demand of AI, ‘we’re not going to hit the climate goals anyway because we are not organized to do it and yes the needs in this area [AI] will be a problem. But I’d rather bet on AI solving the problem than constraining it.’ That rhetoric from the producers of AI has proved influential, with AI now almost always framed as not only a positive development, but one that is necessary for economic growth and meeting the challenges of today.
11 For many, technology is a deus ex machina. Unfortunately, there is little reason to believe that betting on technology indefinitely is going to work out.
WHEN IDEAS DRY UP The belief that we can engineer our way out of Goliath’s Curse is widespread. Cleaning up PFAS and other novel entities may be exceedingly expensive, but we can perhaps engineer a new bacterium to eat it away at a lower cost. AI systems might be dangerous, but we just need more technical research to make sure they are faithful to human values and commands.
This approach is what most of the world’s economic system is relying on.
Yet it is a huge gamble that hinges on our ability to endlessly find new ideas to address new problems before they become irreversible.
Most economists seem to think this is a safe bet; that human ingenuity is a bottomless resource. They place particular faith in ‘recombinant growth’: new inventions in one area spilling over into others, sparking more innovation. Think of how computers could be used to increase our knowledge in areas ranging from finance to climate modelling, so enabling even more developments. Innovation should create constant or even accelerating returns.12 Yet innovation appears to be slowing down.
In the US, the average number of patents-per-inventor has decreased since the 1970s. This does not seem to be compensated for by higher-quality inventions. On the contrary, early innovations in a field tend to be the ones that produce the largest improvement.13 The first combustion engines were revelatory compared with the incremental improvement in engines today. Research productivity is also declining across almost all sectors at the rate of around 50 per cent every thirteen years.14 Ideas appear to be getting harder to find.
Why is this? One explanation is that we have picked all the lowesthanging fruit. New inventions require bigger research teams, more money, and more effort. Another is that materials used in technology are becoming more costly (true for some, but certainly not all industries).
There is also a political component: industries often suffocate new competitors, including through ‘defensive’ research and development.
This means investing to protect market share and using tactics such as lobbying, secrecy, buying up potential competitors, and using legal teams to tie up competitors with patent-infringement lawsuits.15 For instance, fossil-fuel companies such as ExxonMobil had a significant hand in the demise of early electrical cars.
Oil companies filed for over twenty patents for electric vehicles since the 1940s as a way to block their production. They simultaneously lobbied against policies that would encourage consumers to adopt electric cars.16 Companies don’t want innovation per se; they want profit. It’s not impossible that there could be a technological revolution that resets the playing field and allows for a new wave of inventions and growth, as happened during the Industrial Revolution. Many see AI as filling this role, while others suggest more bizarre (and disturbing) workarounds such as cloning particularly gifted researchers to boost research productivity.
17 Even if such an AI revolution were to occur, it’s not guaranteed that it would be an effective or sustainable cure. A new age of innovation would come with a host of new problems. Indeed, most of the core problems we have covered thus far have been enabled by advances in technology: splitting the atom created nuclear weapons; the combustion engine spews greenhouse gases into the atmosphere. Outrunning technological problems by developing more technology doesn’t seem like a sensible long-term strategy. Especially if technological development is still being driven largely by Goliath traps: the pursuit of profit, status, and a military edge. Even overcoming the current hurdle of diminishing returns on research productivity might simply worsen our situation. The acceleration of economic growth and innovation may be the sickness rather than the remedy. As we’ve already covered, growth makes staying within planetary boundaries more challenging. Given that this is already a Herculean task, growing for growth’s sake (or because of status competition) is not a prudent course of action.
Economic and technological growth has another downside: legal lag.
Every new technology introduces new problems. Legal and regulatory systems are already struggling to produce new norms, laws, and legislation to keep up with issues such as copyright infringement in large text models, whether armed drones can kill without human supervision, and the regulation of novel entities such as PFAS. Further accelerating the Global Goliath will exponentially increase the number of issues our already slow and overburdened policy systems need to manage. And there is little evidence that we will solve these international problems. A review of 224 studies found that the vast majority of international treaties don’t achieve their intended outcomes. This is especially true of treaties that deal with areas such as environmental protection and human rights.
The few exceptions relate to trade and finance, domains that tend to serve the wealthiest.18 We should have little confidence that our global governance can keep up.
Exponential growth could eventually simply outstrip the ability of our societies to adapt. Using simple yet sound mathematics, the complexity scientist Geoffrey West has shown that exponential economic growth requires increasingly quick paradigm shifts to keep pace. It took around two millennia to go from the Iron Age to the age of navigation, two more centuries to leap to the Industrial Revolution, just over a century to move into the computer age, and just forty years to accelerate into the digital age.
It’s like having to jump from one treadmill to another, each one running faster than the last (Figure 49).19 Figure 49: Accelerating waves of innovation and economic growth across human history The lines tracing energy consumption and GDP per capita are for the US. If the next treadmill is fast enough, we may be thrown off. Our societies might simply not be equipped to adapt to a new internet-level innovation every ten, five, or even two years. Human societies simply cannot keep up with the changes and problems generated by unregulated innovations and the speeding up of life. This is to say nothing of whether the earth system can withstand even further demands created from ballooning growth. A technological singularity could lead to collapse, not utopia.20 Apart from technological stagnation and over-acceleration there is still the slim chance of a deus ex machina that will allow us to innovate our way out of catastrophic risk while growing exponentially, but even this does not necessarily offer a bright future.
SILICON GOLIATH In 2023, 230,000 tonnes of cobalt and 9 million tonnes of silicon were exhumed from the earth. That is 10 million tonnes less than the amount of plastic that is leaked into the environment every year. That exchange of materials – the extraction of silicon and cobalt and the dumping of synthetic materials into the waters of the world – in many ways summarizes our global system of extraction. Cheap materials, fuel, data, and labour go into the Global Goliath; profit, plastic, carbon, novel entities, and new technologies and weapons (as well as many good things, like medical devices) come out.21 It is the unequal exchange of a new type of emerging system: the Silicon Goliath. Silicon is still not the most important or bountiful of our resources: cement, steel, and ammonia are far more vital.22 Silicon is, however, representative of a new Goliath fuel that is reshaping and speeding up the world system.
Data is a crucial new lootable resource. The world’s most powerful intelligence agencies capture vast streams of it to enable mass surveillance. The world’s most powerful companies strip-trawl the internet to gather the data to train new generative AI systems and to better target advertisements. That data is being used to automate cognition, to speed up and scale up all the existing global systems, whether it be the search for fossil fuels, the production of novel entities, mass surveillance, or the organization of robust-yet-fragile global supply chains.
Data and AI are also critical to new monopolizable weapons such as coordinated swarms of killer robots. Imagine a leader who can not only fire a nuclear weapon without democratic oversight (which is already a reality), but also enforce martial law through fleets of armed and loyal killer robots. Then there is the silicon-laced cage of surveillance. Mass surveillance systems are of little use without vast streams of data and sophisticated AI systems to process them. Disturbingly, unless you are protecting your privacy, almost everything you’ve searched on the internet, every email you’ve sent, and every phone call you’ve made or received is likely to be stored in a server in the US where an agent of the National Security Agency (NSA) can access it. This is due to the ‘PRISM’ program, a mass surveillance system that was exposed by the former NSA analyst Edward Snowden in 2013. PRISM taps into the data flowing through big network providers, such as AT&T and Verizon in the US, as well as companies such as Meta, Google, and Microsoft.
It doesn’t just capture the data of US citizens, either, but also that of most people across the world. Around 25 per cent of internet traffic is routed through (and 30–40 per cent of data centres are contained in) the US (where it is tapped by intelligence agencies), while a lot more is captured by tapping big-tech conglomerates such as Google, and sharing data from mass surveillance systems with allies such as the UK.23 PRISM is just one example of a government-led mass surveillance system; Australia, the UK, China, Russia, and many others all have their own. That is in addition to the extensive ‘surveillance capitalism’ programs from which Google and Meta profit handsomely.
24 Searching and organizing all that data wouldn’t be possible without AI.
Surveillance capitalists are accompanied by intelligence communities to create a ‘stalker complex’ that constantly watches the world. The breadth and intrusiveness of this mass surveillance apparatus is the work of a small number of intelligence agencies such as the NSA and its UK counterpart the Government Communications Headquarters (GCHQ).
That stalker complex is constantly seeking to use technology to more closely watch and understand people, whether it be to better target ads or to identify terrorists and political dissidents.25 Such AI-based massive surveillance programs are increasingly being paired with robotics and other technologies. In 1993 the US government’s Defense Advanced Research Project Agency (DARPA) awarded grants for the development of insect-sized drones that could carry out surveillance.26 The US today has a range of sci-fi-like surveillance projects, including the SMART program, which looks to create a planet-wide monitoring system by harvesting data from different sources such as satellites, drones, and planes. China, during the Covid-19 pandemic, used drones employing facial-recognition technology to enforce lockdowns.27 Then there is the sheer size of the rollout of surveillance cameras. The UK alone has somewhere between 4 and 6 million CCTV cameras, or roughly one for every ten citizens. That is just a small taste of things to come.
Goliaths looking to improve ways to control and coordinate their citizens is a long-running trend. As we’ve seen, societies across history have steadily marched from relying on intimidation to increasingly relying on more effective forms of administration and control. It’s a trend reflected in the growing size and effectiveness of bureaucracies across history, as well as an inexorable increase in the amount of information processed by states.
While Narmer and Sargon knew vanishingly little about their subjects, the CEOs and executives in charge of the stalker complex have a wealth of information about citizens of the modern world, including their names, income, addresses, immigration histories, political beliefs, purchases, and even their movements and whereabouts. Rather than relying on a few parttime bureaucrats, they have enormous highly skilled administrations aided by algorithms.28 The modern world is no exception to this trend. Governments everywhere are investing hundreds of millions of dollars to improve a range of surveillance measures, including facial recognition. As new disasters emerge, surveillance will only deepen. The PRISM program was built off the Patriot Act, a piece of legislation that was rushed through Congress just forty-five days after the 9/11 terrorist attacks. It is a classic ploy to use crises as an opportunity to expand surveillance powers, as also occurred during Covid-19, a savvy way in which the powerful capitalize on our authoritarian impulse during a disaster.
A world of growing catastrophic threats will provide ample opportunities to expand the unblinking eyes of the stalker complex.29 Some scholars have already suggested installing a system of ubiquitous surveillance and pre-emptive policing to stop any (for now, imaginary) decentralized weapons of mass destruction like engineered pathogens. Imagine what kind of security and surveillance measures the US or China might respond with if they suffered a case of nuclear terrorism? Simply the spectre of catastrophic terrorism may be enough to scare the world into greater autocracy.
30 Such systems don’t seem to be effective at catching terrorists at the moment, let alone creating a comprehensive, real-time watchful eye over the entire world. Similarly, the production and abilities of killer robots are limited and crude. We should be sceptical of techno-hype, yet the levels of power and precision they could reach as the decades pass is significant.31 For some, this may not be worrisome. Being watched is barely noticeable for the luckiest of us, except for the occasional creepy indication like seeing more tailored ads. But relying on these systems to remain in relatively safe, accountable hands in a world of democratic backsliding is unwise. We are constructing systems of extensive control in a world that is becoming more unequal and extractive.32 The transition between one form of Goliath and another is rarely a clearcut matter. The same is true of the current transition from the Global Goliath to Silicon Goliath (Table 7). As our systems become more dependent on advanced algorithms, and as mass surveillance (and new weapons) curtails the ability to resist through protest, we cross increasingly into the realm of Silicon Goliath. We aren’t quite there yet: many systems around the world can still effectively function without AI, mass surveillance often lacks effectiveness, and new monopolizable weapons such as drone swarms are still limited in their use. Yet we are transforming into a Silicon Goliath at a rapid pace.
Even if the best-case scenario of techno-optimists occurs and we manage to technofix our way out of all the catastrophic risk we face, if we reach some kind of technological singularity, the outcome is unlikely to be appealing to many. Yes, we may have far greater power over nature and technological prowess – perhaps even space flights and robotic slaves – but probably far less freedom, democracy, equality, and human connection. The same rule throughout all history applies to the future too: technology alone won’t give us a civilization. Yet there are actions that will.
KILLING GOLIATH The last path forward is the least likely: that we collectively kill (or perhaps tame) the Global Goliath. Across history Goliaths have collapsed, but wherever there has been Goliath fuel, they have tended to rise again. While some societies have been more democratic, they have tended to be either outcompeted by others or become more extractive over time. There are a few factors in the modern world that could help us to kill Goliath and create a more equal and less exploitative global system. That is, a global civilization.
The first is making resources less lootable. Imagine if the data of today could be gathered only with the full and informed consent of citizens and was fairly remunerated. The world of AI would change dramatically as the development of algorithms slowed and the risk was alleviated.
The second is increasing exit options. Creating more open borders and, importantly, more ways for people to escape mass surveillance, would mean that states would have to negotiate more with their citizens. It’s a trend we’ve seen throughout history: geographies with more exit options favoured democracy. While this would need to be done carefully to address security and cultural concerns, it would be a step towards the fluid civilizations of the past.
A third is using technology to enable both larger-scale democracy and greater oversight of the powerful. The Agents of Doom use secrecy because they know that transparency would lead to greater regulation.
While Rome and other early republics and democracies were limited in how many citizens could attend assemblies, today we can use online platforms and deliberative assemblies to run democracy at a far larger scale more intensely. Technology is expanding the realm of what is democratically possible.
The fourth and final possibility is knowledge. The survivors of the first Goliaths of northern America knew the oppression that accompanied a dominance hierarchy and could organize to avoid it. Yet no society through history understood the darker angels of our nature or Goliath fuel.
Understanding why Goliaths emerge and how they fall into Goliath traps can help us to create interventions and institutions able to stop history from repeating (I’ll highlight several promising interventions in the Epilogue). The final two factors – technology and knowledge – have changed dramatically in the past two decades. They are also the two that give the modern world the best chance of escape from Goliath. But it is a path with few precedents, and there is little sign that we are currently on it.
CHOOSE YOUR CURSE A new Silicon Goliath is emerging. The existential risk we face is in many ways the same risk that always accompanies the rise of a new Goliath.
The first states were littered with blood-covered bronze swords and pits of sacrificed slaves. The creation of empires like the Qin and Rome created seas of fractured skulls. The establishment of a new colonial, capitalist world system was enabled by the sacrifice of the tens of millions made to work under the whip, or die alone drowned in the Atlantic, crushed in mines, or slowly perishing from the pitted skin and scarred lungs of smallpox. The legacy of the Silicon Goliath could be the destruction of the entire planet, already disrupted by new weapons that can annihilate cities in the blink of an eye, and the slow removal of decision-making and autonomy from our hands into the cold and fast machinations of AI. Ironically, the rise of our Silicon Goliath could trigger its fall.
Looking at both the long-run trajectory of Goliath and current trends, our Global Goliath appears destined to hurtle down one of three paths. The first, and most likely, is self-termination. The second is a world in chains.
The third is a world in which we somehow manage to shackle and control it.
Reversing inequality and establishing deep democracy in a lasting fashion has unfortunately been a rarity throughout history. It is a highly unlikely outcome, which makes it worth fighting for even more desperately. While throughout history the collapse of a Goliath was usually temporary and a liberation, in the future it threatens to be permanent and take us and much of the planet with it. We are passengers on a journey that looks likely to end in chains or evolutionary suicide. Our Global Goliath will die explosively unless we kill it first.
The prognosis here may seem dire and speculative. Yet, ask yourself, where do races to the bottom and arms races lead to in the long term but large-scale destruction? We can’t rely on our luck never running out or somehow finding a plucky technical fix to every new technological problem. Similarly, are there any signs of the drive towards mass surveillance slowing or reversing? Even the world’s most ardent democracies are spending at least an order of magnitude more on improving facial recognition and surveillance, and at least a hundred times more on their militaries, than they are on improving democracy (whether that be through initiatives to improve voter turnout or experiments in deliberative democracy). This is what economists call ‘revealed preferences’: follow actions and money to understand the underlying motivations. The preferences of our emerging Silicon Goliath are clear.
This is not to say that these scenarios will come to fruition in a few years, or even a decade. There will be delays and reversals. But over the course of decades and centuries, the deeper trends we’ve seen across history will show through.
If self-termination, some kind of global societal collapse, is the most probable trajectory for our Silicon Goliath, we have to ask: just how bad could it get?
22 After the Fall Hart’s Mystery – From Homo solus to Homo finalis – A Second Industrial Revolution? – A Hellish Earth – Survival of the Richest? – Goliath Rebooted? – A Kaleidoscope of Futures HART’S MYSTERY Are we alone in the universe? Some think that, despite there being no signs of extraterrestrial life, the universe should be teeming with technologically advanced species. This riddle began with an equation formulated by the astrophysicist Frank Drake in 1961. Drake calculated that the number of alien societies we could contact is determined by the number of planets that could carry life, the number of times that life begins and evolves into an intelligent, advanced society, and the length of time for which it has persisted. Given that the universe is around 13.7 billion years old, and contains 100 billion planets, it seems highly unlikely that there are no spacefaring aliens out there. But where are they?
This is often mistakenly termed the ‘Fermi Paradox’. The legend is that during a lunch with colleagues in the 1950s the physicist Enrico Fermi (one of the architects of the first atomic bomb) asked, ‘Where is everybody? Where are all the aliens if there should be so many of them out there?’ The truth is that Fermi was actually complaining about the impossibility of intergalactic space travel. The question of ‘Where is everybody?’ was, in fact, first asked by the astrophysicist Frank Hart in 1975.1 It is also not a paradox. You can end up with an infinitesimal number of potential extraterrestrial cultures if the number of habitable planets or the likelihood of life beginning is tiny. Once you factor in the uncertainty about how likely life is to start, suddenly the paradox dissolves and it becomes mathematically unsurprising that we might be alone in the universe.2 Rather than dealing with Fermi’s Paradox we are dealing with Hart’s Mystery: the mystery of our seemingly solitary existence. When we gaze out to the stars, we realize that for now we are more than likely the only flickering light of sentience in a vast sea of empty blackness.
That raises the stakes of continuing life on this planet: the choices we make now could determine whether the universe is potentially left as a collection of rock and burning stars whirling in a lifeless vacuum for millennia to come. If a global societal collapse does happen, then will it be our species’ final curtain call or just another rebirth?
FROM HOMO SOLUS TO HOMO FINALIS Of all the species that have ever existed, 99.9 per cent are now gone.
It’s almost certain that Homo sapiens will eventually go the same way.
Unless other realities exist that we can access with god-like technology (an enormous if), then when this universe ends – whether it is ripped apart owing to its own expansion or collapses in on itself – we will too.
The more pressing question is whether a collapse in the coming decades or centuries could spiral into human extinction. It’s a difficult question to answer since collapse can take so many forms, from a relatively non-violent decline of our global economy and power structures over decades to an apocalyptic spasm of nuclear-fuelled destruction. Still, it’s important to trace how many people would need to perish for it to result in probable longer-term extinction and what impediments the survivors would face to surviving.3 Humans have survived for at least 200,000 years. If one looks at the extinction rates of other mammals, then statistically that makes our odds of being wiped out in any given year (naïvely not assuming any other catastrophe) one in 14,000. That figure balloons out to one in 870,000 if we consider all our hominid ancestors.4 Yet it is questionable how relevant such calculations are for a species that is spread across the entirety of earth and which can piece cities back together after they have been split apart by atom bombs. There are other, more useful ways of trying to calculate our odds of survival.
Imagine you are one of a group arriving on a lush, deserted island. How big would the group need to be in order to survive and create a selfsustaining society? You need enough to avoid inbreeding and other deleterious effects of a too-small gene pool. You also need a few extra as a buffer to guard against accident, drought, or disease. This is known as the ‘minimum viable population’.
We have some historical examples which can give us a glimpse of what this number is likely to be. The original peopling of both the Americas (at least 23,000 years ago) and Polynesia (around 1000 BCE to 700 CE) may have occurred with a group of around only a hundred. The initial raft of people landing in Polynesia who eventually led to the indigenous cultures of the Māori and the Samoans may have just been a few dozen.
We get other surprisingly low numbers when we look further back in time. The first successful migrations out of Africa into Europe and east Asia may have totalled as few as 1,000–3,000 people. The human groups there still thrived.5 Others have looked to the stars to estimate the minimum viable population. How many people would it take to start an intergalactic colony?
Australian geneticist Ian Franklin and American biologist Michael Soulé in 1980 proposed that fifty individuals would be needed to combat inbreeding, and 500 to prevent genetic drift (small random fluctuations in genes can lead to certain alleles disappearing in a small population, leading to a loss of genetic variation). More recent, refined guesses have turned up larger figures of 14,000–44,000. The big spread in numbers is due to varying estimates of the required buffer against calamity.
Disasters happen, and even if a minimum viable number is enough in theory, it would be precarious in practice. As the Neanderthals found out, sparse dispersed populations can dwindle into extinction with just a bit of bad luck. That larger estimate of 14,000–44,000 would also need to be in close proximity so they could reproduce, protect each other, and form a community. Even a community of 44,000 might be vulnerable if the conditions were dangerous enough.6 While few have explored it, these numbers also depend on the genetic diversity of the group. Even a minimum viable population number of 44,000 is, of course, very low compared with the current global population of 8.2 billion. This makes the possibility of extinction appear far-fetched. The only attempt to examine this issue suggested that even an extreme collapse with extensive infrastructure damage, temporary climate change, and a population loss of 99.99 per cent would have only a 0.1–0.3 per cent chance of causing extinction. There would still be 810,000 survivors, who would have a ‘grace period’ of easy access to remaining supplies (canned goods and produce) and equipment. That might buy them enough time to learn the necessary survival skills such as basic foraging and agriculture. Killing off every single human is an incredibly high bar.
7 That said, there are also reasons to believe that it will be trickier in the future for a population to bounce back post-collapse. The survivors would be living in an environment that is more hostile than settling the warm, bountiful islands of Polynesia. That is due to the latent risk we face: the survivors would not be trying to scavenge and rebuild in the stable climate we’ve had throughout the Holocene. They would be living in a far warmer world, one with fewer survivable habitats and where foraging or farming would be far more difficult. They would also potentially have to cope with more diseases than our ancestors did, and far fewer means of control or cure once the supply of leftovers available in hospitals and health centres ran out.8 Survivors would be affected by the long-term effects of novel entities,9 including declining fertility and sperm counts.10 All this means that the viable population is more likely to be at the 44,000 end of the range, or perhaps even higher.
How we would fare in a global collapse is also dependent on how prepared we are. In the north of Norway in the Svalbard archipelago is the Svalbard Global Seed Vault, also known as the ‘Doomsday Vault’.
There are over 1,700 gene banks around the world which house different seed types (for agricultural research and development) and the Doomsday Vault aims to have a copy of each unique seed. It was intended to be an icy, impregnable fortress sunk into the permafrost, but in a warming world, that permafrost is now melting. In 2017 the vault was flooded, although luckily the meltwater did not reach the seeds.
Nonetheless, it’s hard to predict how well the vault will fare in a 4°C warmer world.11 Whether we survive a catastrophe depends in large part on what food is available afterwards. Most scenarios of global catastrophe involve a largescale climatic disruption, whether that be a post-war nuclear winter, a highimpact volcanic eruption, or extreme climate change. Scientists at the organization ALLFED (the Alliance to Feed Earth in Disasters) have already started to cook up schemes of how to feed the entirety of the world’s population if traditional agriculture fails. Such plans include mass food factories producing mushrooms, insects, algae, and seaweed, or even the straight chemical synthesis of food. However, the research is nascent and more needs to be done at scale. Certainly, we are nowhere close to being able to feed the survivors of a nuclear war. Even if we do scale up these systems, they don’t solve the political problem of distribution. (Today, the world produces more than enough food to provide 2,750 calories per person every day, yet 9 million people starve every year because food isn’t distributed appropriately.) More support is needed for the vital work of organizations like ALLFED if we are to be prepared for the worst-case scenario.12 There are a few lessons we can draw from all this. First, extinction is implausible unless a truly cataclysmic situation involving a combination of catastrophic threats and vulnerabilities occurs. Second, even under these conditions humanity is likely to survive in the short term, at least. Third, we are dramatically underprepared. Fourth, the longer the status quo persists, the worse a future collapse will be. We need to start worrying not just about the fall, but also where we are falling from.
Extinction in the longer term, over a span of thousands of years, might also be more likely if we aren’t able to rebuild a sufficient level of cultural and technological power.
13 A SECOND INDUSTRIAL REVOLUTION?
Almost every collapse has been followed by a ‘technical recovery’: regaining the previously established levels of energy capture and population density. This doesn’t mean recovery in the sense of improving people’s lives. If a future technical recovery is anything like the first bout of industrialization, it will carve deep scars into the body of humanity all over again.
Whether the survivors of a deep global collapse would be able to recover is questionable. We’ve only ever had one industrial revolution, and that was an exceptional one-off that occurred in a stable climate with plentiful, easily accessed coal and oil. If we had a deep collapse and the entire industrial skeleton of the modern world fell to pieces, would we be able to repeat it?
It is a topic that few have dared broach. There have been some pieces of analysis, ranging from presentations and blog posts through to Lewis Dartnell’s The Knowledge: How to Rebuild Our World from Scratch.
14 It is an absurdly underexplored topic, given its importance, partly because it is necessarily speculative. It would depend on how deep the collapse is, and how great is the energy ratio of the fossil fuels that remain in the ground (as well as the renewable energy we have at our disposal). These are largely unknowable factors. Dartnell is the only one to have seriously tried to answer the question.
He took a related but different approach: could the Industrial Revolution have been achieved without fossil fuels? His answer is a tepid maybe.
Charcoal from sustainably harvested trees could provide one source of dense energy, although lack of land would keep the supply limited. Solar PV is durable but difficult to make. It relies on extremely thin slices of pure silicon components. This, in turn, needs technology similar to that required to make semiconductor electronics components. Renewable electricity from wind and water are more promising. We could start with windmills and watermills before progressing to wind turbines and hydroelectric dams. This would be possible only in a few favourable locations. The second industrial revolution would not take place in the UK but more likely in Scandinavia.
Dartnell concluded ambivalently: ‘An industrial revolution without coal would be, at a minimum, very difficult … a combination of renewable electricity and sustainably grown biomass, might be possible after all.
Then again, it might not.’15 Dartnell is right: an industrial revolution without coal would be a momentous challenge at best and physically impossible at worst. We don’t know, but there are reasons to think that the challenge may be easier in a post-collapse world. Even a (near-term) collapse that incurred large-scale infrastructure loss, such as a nuclear war, would still leave plenty of working coal mines, oil rigs, and stored gas. Knowledge of high-efficiency engines and power plants would smooth the transition. If it happened within the coming decades, there would still be many fossil fuels to draw on. There are currently around 1.7 trillion tonnes of coal that could be profitably recovered using existing technology, and 188 trillion cubic metres of gas. If we continue at the current rate of production, we have around 134 years of coal reserves left, 51 of oil, and 44 of gas.
Burning coal, oil, and gas is, of course, eating the future in more ways than one. Using those reserves would be more than enough to cause climate chaos. That is a problem that a second industrial revolution would face: achieving a technical recovery while worsening climate change. This in itself may make the task insurmountable. There is then a second problem.
The main impediment is not the amount of fossil fuels left in the ground but rather their energy ratio.16 We would be approaching industrialization with hard-to-reach and hard-to-use resources. The declining energy ratio of fossil fuels is likely to make a second industrial revolution impractical.
Would post-collapse communities even want to reindustrialize? Studies almost always assume that of course they would want to.17 Yet it’s not a given that a post-collapse population would have the same pro-technology sentiments that many have today. If collapse was triggered by a nuclear war, extreme climate change, or another technological calamity, then the survivors might have little appetite to repeat past mistakes. They might choose to reject the old system, like the people of Cahokia and Tiwanaku, who destroyed elite symbols and monuments as they walked away. In a world ruined by runaway technology, that might be an entirely sensible decision.
The downside is that, although in the short term an unregenerated industrial collapse would almost certainly lower the threat of human extinction, in the very long term we’ll require advanced technology to survive.
A HELLISH EARTH Over tens or hundreds of millennia we are likely to face many worldshattering natural hazards, such as gargantuan volcanic eruptions.
Over millions of years, the chance of a collision with a large asteroid or comet suddenly becomes non-negligible. Survivors living in a non-industrialized world would not have the resources to deflect an asteroid or prevent an eruption. However, the potentially biggest threat is the changes in the earth we are triggering today. Climate change and temperature rise will not magically stop in 2100 (where the models frequently end). The earth will continue to transform.
There were once fears that enough global warming could trigger a runaway greenhouse effect which would boil the water on the surface and sterilize the earth, turning it into a planet essentially like Venus. To everyone’s relief, the IPCC suggested in 2009 that there was virtually no chance of human activities creating a runaway greenhouse effect (although there are a few theoretical models and scenarios that disagree).18 It would also be a process more likely to unfold over hundreds of thousands of years.
A runaway greenhouse effect is not needed to reach human extinction, though. Over tens of thousands of years, the world would be utterly unrecognizable with 6°C of warming. The most deadly and disturbing change would come from the oceans, carbon in the atmosphere making them more acidic while melting ice caps dump freshwater into them, blocking deep-sea circulation and cutting off the circulation of oxygen.
This would create ‘anoxic’ oceans with little oxygen and essentially no marine life. The sea would become a dead zone colonized by purple and green sulphur bacteria. The surface would become a vast expanse of purple with a slick oil-like surface. Hydrogen sulphide gas produced by bacteria would belch from its surface, destroying the ozone layer. Along with lethal heat, mass extinction, and dead oceans, we would also have to contend with more ultraviolet radiation. This would skyrocket skin cancer rates among the already ragged, vulnerable survivors.19 These anoxic oceans and their effects seem to be key to explaining how climate change has triggered mass extinction events in the past.20 Add in the destruction we have wreaked on other planetary boundaries and longterm extinction doesn’t seem implausible. Certain parts of the earth may still be technically habitable, and we are a remarkably ingenious and resilient species that has weathered ice ages and dramatic climate swings before. Yet every species has its limits. That’s true of even our species’ most powerful members. SURVIVAL OF THE RICHEST?
In the past, elites have used their wealth to migrate or change their identities when trouble looms. Power is helpful even during a crisis (unless the holders of it are directly targeted and killed off during the tumult). For now, the mega-rich seem convinced that they will repeat the feats of past elite survivors and escape the apocalypse. The main customers of the most luxurious doomsday bunkers are billionaires. Peter Thiel, the billionaire co-founder of PayPal and AI (or rather, AI for military, security, and policing) company Palantir is a prime example. In 2011, Thiel bought a 477-acre former sheep station in the South Island of New Zealand. He also purchased New Zealand citizenship, secretly lobbying the government and pledging to invest in New Zealand start-ups as part of his application. Citizenship was granted, even though he had spent less than twelve days in the country up until then (the usual criterion is to have lived in New Zealand for 1,350 days to be granted citizenship). OpenAI CEO Sam Altman revealed in 2016 that he and Thiel have a deal by which if a global collapse occurs they will board a private jet and fly to Thiel’s property.
21 Thiel and Altman’s preference for New Zealand is widely shared, since New Zealand will be one of the most resilient countries in the case of a global catastrophe. Its location deep in the southern hemisphere means it will be one of the least affected by the fallout from a nuclear war, and its isolation makes it well protected from pandemics. It is a doomsday safe haven. This has led to it being recommended as a kind of national Armageddon bunker in the libertarian manifesto The Sovereign Individual: How to Survive and Thrive During the Collapse of the Welfare State, a book that has been endorsed by Thiel and other billionaires such as Marc Andreessen.
Then again, why buy a doomsday bunker on a small piece of land when you could buy an entire country instead? That is exactly what the former cryptocurrency mogul Sam Bankman-Fried was considering doing.
Bankman-Fried was head of the cryptocurrency company FTX before being jailed for fraud. He and his brother Gabe (who led the philanthropic FTX Foundation) discussed buying Nauru, the third-smallest state in the world, located in Micronesia. Their plan was to ‘purchase the sovereign nation of Nauru in order to construct a “bunker/shelter”’ that would be used for ‘some event where 50–99.99 per cent of people die [to] ensure that most EAs survive’. (In this case ‘EAs’ refers to the Effective Altruism movement, which Bankman-Fried was part of.)22 Self-preservation is a widespread instinct among the richest and has spawned an entire industry of doomsday bunkers across the world.
Reinforced, luxury bunker-manors with pools, wine vaults, artificial gardens for sunbathing in simulated sunlight, and underground hydroponic farms, can be found from Texas to Czechia. One group of super-rich preppers asked the media theorist and writer Douglas Rushkoff for advice on how best to prepare for a world-ending tragedy of whatever stripe, which they described as ‘the event’. Some have purchased not just bunkers but also their own security forces (in one case a dozen Navy SEALs) to defend their mini-fortresses from mobs and raiders.
Their questions to Rushkoff were tellingly self-interested: ‘How do I maintain authority over my security force after the event?’ The options they had considered included replacing their guards with more loyal robots or ensuring obedience with disciplinary collars (the kind that zap you if don’t follow orders). When Rushkoff suggested that they instead befriend their employees and treat them well now to build a connection that might last through a crisis, they rolled their eyes.23 This bunker mentality is a flawed strategy. It’s unlikely that these bunkers would stand the test of a true worst-case scenario. Even the Svalbard Doomsday Vault was almost compromised by a single flood.
Underground farms might work in laboratory settings but are a different matter in reality.
They would need to function and avoid contamination for years while under duress. All it would take is one rogue bug and the entire food supply could be ruined. It’s also unlikely to work because it does exactly the opposite of what has allowed communities to persist and prosper through disasters of the past and present.
To prepare for a societal breakdown, you are better off stockpiling friends, favours, and skills rather than guns and canned beans. That’s the central message we have from the Palaeolithic and studies of past societal collapses and modern disasters. Communities that respond the best tend to have broad, strong social ties. They quickly cooperate when the wider political structure falls apart, with neighbours and acquaintances providing the role that the state once did, whether that be by sharing resources, organizing medical aid, or protecting each other.
The billionaires of today are pursuing a strategy of investing in places rather than people. Not only is that likely to backfire, it’s also exactly the kind of thinking that is fuelling collapse today.
24 There are even some scenarios in which the poor, not the mega-rich, may be the most resilient. Countries that are most reliant on global industrial infrastructure will be most affected when it is suddenly stripped away.
Imagine a ravenous computer virus shuts down critical infrastructure and we are left without functioning machinery, or the ability to create new pesticides and fertilizers. Europe, the Americas, and parts of India and China could experience losses of up to 75 per cent of their rice, corn, wheat, and soybean crops. Africa, in contrast, would see drops of only around 5 per cent for soybeans and 25 per cent for rice (although almost half of wheat crops could be lost).25 Many African farmers and city-dwellers are also already accustomed to living with unreliable utilities.26 Much will depend on how collapse unfolds. (In the case of a pandemic, New Zealand would be the safer option.) It’s difficult to see through this fog of uncertainty which groups would fare best post-collapse. For now, being poor in wealth but rich in friends is a better strategy than being a suspicious billionaire under siege in a bunker. Not only that, but cultivating bonds of solidarity and mutual aid rather than ruthless self-interest means that the survivors are more likely to help recreate a world that is worth preserving.
This is perhaps the most important and least appreciated question about a post-collapse world: whether it will be a world like today’s – of patriarchy, rich and poor, slaves and masters – or a world of civilization.
GOLIATH REBOOTED?
Ancient Greece looked very different from the rest of the Mediterranean after the Late Bronze Age collapse. The Assyrian Empire continued to have an empire, one that expanded and immiserated its subjects more with each passing decade. Egypt continued to have pharaohs. Greece remained a land without large cities or kings for several centuries. When Classical Greece did emerge, their societies had little connection to the Mycenaean palaces that came before them. Even their writing was entirely different. Linear A and B were lost and the Greeks instead adopted and modified the Phoenician alphabet. The collapse of the Mycenaean kingdoms was unusually deep. And the Greek polities that emerged afterwards were unusually egalitarian.27 City-states such as Athens began to pop up from the ruins. While many sought to restore old hierarchies, other communities remained egalitarian.
Greece became an ecosystem of hundreds of city-states in which much of the organizing was done by citizens, rather than by religious leaders or bureaucrats. They maintained a surprising amount of wealth equality (at least for free men) and their economic growth and well-being far outstripped that of their neighbours. Compared with the empires that surrounded them, Greece was a hotbed of democracy and equality. Why did post-collapse Greece look so different?
The severity of the collapse, Iron Age weapons, and the very geography of Greece all played a role in its democratic recovery. The collapse was exceptionally deep. Wealth was dramatically levelled. Essentially, everyone in Greece became equally poor with few differences in power.
The terrain of Greece also tilted away from autocracy. It is a land of islands, and around 80 per cent of it is mountainous. In other words, there were plenty of exit options. If you didn’t like the edicts of a budding king, you could more easily hop to another island or retreat to the mountains away from the grasp of armies and tax collectors.
New iron weapons also played a role. They were far cheaper and easier to make than bronze, meaning they became more widely dispersed.
Communities that could field the largest armies of iron spear- and shieldwielding infantry were the ones that survived. Politically inclusive groups were usually the ones that could mobilize more citizens for war and give them a greater sense of camaraderie and solidarity. This was vital for success in battle. Greeks began to use the phalanx formation: a tightly packed rectangle of troops fighting shoulder to shoulder behind an interlocked wall of shields. Such shared struggle also gave the troops a greater sense of connection beyond the battlefield. Conquered slaves aside, post-collapse Greece selected for democracy and equality.
28 Greece is one example of how deep collapse can lead to a more democratic recovery. The society that emerges from the ashes is freer, more equal, and more prosperous than what came before. It is a far cry from the militant raiders and despotic patriarchs we see in dystopian films like Mad Max or The Road. It’s not only movies that envisage a more autocratic postapocalyptic world. Some scholars have suggested that post-collapse societies will revert to the historical average of predatory autocracies. The current prevalence of modern democracies and republics is not representative of how most of history has looked: it’s an aberration. These scholars argue that in the absence of other evidence, we should assume that we’ll end up looking more like a normal, pre-modern kingdom than a Norwegian welfare state. This might make sense if collapse were a random reset, but it’s not.29 As we’ve seen in the case of Greece or the collapse of the first Goliaths of the Americas, there are some factors that reliably lead to more democratic recoveries. These include more wealth equality, more exit options, fewer lootable resources, and either a rejection of previous hierarchies or a cultural memory of inclusive practices. Less Goliath fuel, a flat playing field, and a cultural distaste for domination all favour democracy. All these should increase after a deep collapse. The more wealth that is levelled, the flatter the playing field, and the more responsible great powers are for the fall, the more Goliath structures will be blamed and discarded. A deep collapse could provide fertile soil for democracy to take root.
However, numerous modern factors could gain momentum, pushing us towards autocracy. There would be cultural memories and technologies of how to watch and control others. Earlier despots didn’t have surveillance equipment or centuries of development on how to bureaucratically track and understand citizens. Parts of the world would still be likely to have an abundance of lootable resources, including fuel and machinery. Then again, the sheer abundance of leftover materials might work against a Goliath reemerging. As in the case of the Black Death, if there is enough population loss then labour may become the limiting factor. This would give people far greater bargaining power over any would-be despots.30 The biggest wildcard determining what kind of society emerges from the ashes is culture. Whether it be in Zomia, the Mississippi, or the south-west US, people passed down stories of evil wizards, overbearing monarchs, and tyrannical trickster gods to warn of the dangers of vesting too much power in leaders. Those were for comparatively less traumatic disasters. Imagine what kinds of tales might be crafted in the wake of the floods and fires of a 4°C world or the sickly glow of mushroom clouds and irradiated rainfall.
Would we speak of mad machine-obsessed priests, myopic cults of an eartheating stock market, and delusional dictators with the power of gods and the restraint of toddlers? It’s easy to see why the survivors of a true cataclysm would not look fondly on the Global Goliath of today and why they might enshrine the myths and practices that would help them avoid its resurrection. As in Greece, the deeper the collapse, the greater likelihood of a democratic recovery. This is, unfortunately, not a good news story, at least not in the long term. Classical Greece was still covered by slave-owning Goliaths, even if there were democratic practices between free adult men.
Eventually it was covered by larger, less democratic empires. North America may have had stories and collective memories to avoid Goliaths, but an indigenous empire would still probably have emerged.
Despite the initial conditions favouring democracy, history is always likely to repeat if Goliath fuel remains. That’s because the darker angels of our nature also remain, and there will be a selection pressure for more militaristic societies over the long run. We can’t naïvely rely on collapse killing Goliath.
A KALEIDOSCOPE OF FUTURES Collapse is not uniform. From a wide view the world system may have collapsed, but once you zoom in, you’ll notice different recovery processes in different regions. Some regions will have better odds of recovering in a politically inclusive manner or kickstarting a low-carbon industrial revolution.
This would mimic the collapses of the past. We see a similar kaleidoscope of outcomes during the Late Bronze Age collapse. The Mycenaeans experienced a near-total collapse followed by a democratic and cultural efflorescence after centuries. The survivors of the Hittite Empire persisted in a few minor, rump states in northern Syria. In contrast, the Assyrians and Egyptians kept their kings and centralized economies.31 The Neo-Assyrians went on a warpath while the Phoenicians focused on trade and cooperation. We would see a similar panoply of outcomes in a postcollapse world, which could be populated by aggressive Australian pirates or democratic sub-Saharan trading cities.
One of the few trends that seems to hold across regions is that the longer current trends persist, the deeper the collapse will be. Latent risk, whether it be antibacterial-resistant microbes or global heating, will continue to build.
The energy ratio of fossil fuels will continue to fall. Unless things change radically, we’ll continue to climb the rungless ladder. The one bright spot is that the deeper the fall, the more likely it is that we’ll rise again as democratic equals. Yet waiting for a cataclysm to rebalance power is an unnecessary, grotesque gamble. We don’t need to roll the dice with evolutionary suicide; we can stop this five-millennia-long trajectory and slay Goliath.
23 Fiat Justitia, Ne Pereat Mundus The Trinity Jury – From Darker Angels to Evolutionary Suicide – Justice Lest the World Perish – Memory Over Myth THE TRINITY JURY In 1945, on the eve of the Trinity atomic bomb test, the scientists behind the Manhattan Project began to place morbid bets. During the construction of the bomb, the physicist Edward Teller had calculated that the fission chainreaction of the bomb they were about to detonate could potentially ignite the atmosphere, extinguishing all life on earth. The probability was incredibly low, but Teller could not rule out the possibility of incinerating the entire biosphere.
The scientists were gambling over not just the end of the world, but potentially the end of all life across the cosmos. Earth contains the only observable signs of life in all the universe. This blue marble floating in the vast ink of space may be the only refuge for the miracle we call life.
The bomb could have left the entirety of existence empty and silent.
It was a huge risk, and also an unnecessary one. The Americans knew that the Nazis were no longer seriously pursuing the creation of their own atomic weapon, and that even the design they had once tried to pursue would never be successful. These were facts that the US government chose not to pass on to the scientists dutifully creating the world’s first atom bomb at Los Alamos. The Trinity Test was no longer about preventing the Nazis from creating a nuclear Third Reich. I want you to imagine that the decision to detonate the bomb – to risk igniting the atmosphere – had rested not with betting scientists or the machinations of wartime generals or politicians but with a jury of randomly selected US citizens fully briefed both on the impacts and risks of the bomb and on the geopolitical and military situation. Now, ask yourself this: would that jury – with their friends, sons, daughters, parents, husbands, and wives waiting at home – have taken the same gamble? What would be a similar jury’s verdict on the other global risks we face today? Would they decide to speed up the development of AI? Would they be happy to continue to increase emissions or search for new sources of oil despite the risk of catastrophic climate change? Would a citizens’ jury sitting in on the board meetings of 3M, DuPont, BP, and Shell have hidden evidence of the global harm caused by their products, or run disinformation campaigns?
The early evidence says they wouldn’t: democracy is risk-averse. In 2020, President Emmanuel Macron of France approved a French Climate Citizens’ Convention. It was a response to the Yellow Vest protests in which citizens protested against rising energy prices, inequality, and perceived corruption. The Convention was designed to address all these concerns by having a group of ordinary members of the French public determine climate policy while informed with the latest science. Macron agreed to adopt all the recommendations advanced by the assembly. They spent the next two years learning and debating. At last, they agreed on a set of 148 recommendations: a progressive policy package to make all products sold in France disclose their carbon content, impose a carbon tax on all imports into the EU, and even make ‘ecocide’ (the destruction of entire ecosystems) a criminal offence.
Macron then faltered and broke his word. Many of the key provisions proposed by the citizens were watered down and weakened. In the end, less than 40 per cent of the proposals were properly adopted and implemented, and the most ambitious were scrapped. It seems unlikely that Macron reneged on his deal because there were fundamental flaws in the citizens’ proposals (remember that the entire process was overseen by large groups of experts). He opposed them because they were too ambitious.1 Imagine that Macron had asked the citizens’ assembly whether France should continue to have nuclear weapons and they had said no. It’s difficult to see him accepting their decision. When citizens are given the reins, they tend not to want to risk the world or the future of their children. So much so that leaders need to seize power back. Currently companies and countries can privately profit from producing public risks. We need to put these actors back under democratic control.2 The use of deliberative assemblies and juries is just one example of ‘open democracy’: accountable, inclusive governance led by the masses with ordinary citizens at its decision-making heart. Open democracy aims to continuously improve and innovate how we make collective decisions. It’s about building more resilient societies, and also wiser, fairer ones. It is a method built on an appreciation of our deep past and evolutionary inclinations.
FROM DARKER ANGELS TO EVOLUTIONARY SUICIDE The biologist E. O. Wilson once identified the fundamental problem of humanity as the fact that we have ‘Palaeolithic emotions, medieval institutions, and godlike technology’. There is some truth to that. Yet it’s not as simple as our outdated emotions meeting superpowered technology. The psychology of most people is not the problem.
On the contrary, we can be immensely sociable, intelligent, altruistic, wonderful at self-organizing, and quick to learn from each other. We see that throughout the ages. For most of human history we lived in fluid civilizations. These were webs of exchange, mutual aid, and reciprocity that were full of mobile, egalitarian, social individuals with a strong aversion to being dominated. Status competition lurked underneath these arrangements but was kept in check through counter-dominance strategies and sheer environmental circumstances. Collapse occurred only when environmental shocks disconnected populations or severed them from the wider civilization.
The fluid civilizations of the Palaeolithic allowed us to survive the ice age and cover regions and continents with shared tools, instruments, and culture. It’s little wonder we make better decisions when we come together, deliberate, and debate. That shared heritage is why communities regularly persevere and even flourish through collapse, and why earthquakes and hurricanes often end up providing tales about the triumph of the human spirit, with even the injured quickly providing mutual aid, food, shelter, and love to those in need. Those same better angels of our nature can explain why we would trust a group of ordinary citizens not to detonate the Trinity bomb.
Open democracy and inclusive institutions are essentially harnessing these better angels of our Palaeolithic psychology. They are using the best of us to make better decisions and closer communities: to harness our collective intelligence and impulses to help each other. They are moving closer to our past as equals in a web of sharing, constantly moving communities: a fluid civilization. Our Palaeolithic emotions are not a curse.
The curse is the institutions and arrangements which bring out the worst in us.
These institutions began to emerge 12,000 years ago, when the climate warmed. People became increasingly dependent on lootable resources, such as grain and fish, and the world got smaller. When weapons became available, and if the landscape was caged, small groups began to seize control of these new lootable resources. We began to slowly, hesitantly adopt more primal forms of organization: dominance hierarchies. It was a case of evolutionary backsliding: we began to give up the arrangements that had made us uniquely human and instead more closely resemble our more hierarchical chimpanzee and gorilla cousins.
As inequalities and hierarchies rose, so too did war. The threat of mass violence caused many to reach for strong, aggressive leaders, an authoritarian impulse that is buried deep in us. They became protection rackets of sorts, and the state was born.
All it took was one or two societies turning into a Goliath to set off a chain reaction. The authoritarian impulse was triggered in neighbours who now had a model to emulate. These were not thoughtfully designed institutions. Those who sought power were the least fit to wield it.
Frequently the most violent and overconfident were selected for and the sheer concentration of power in masters, kings, pharaohs, and emperors led to rulers who were far too often corrupt.
These Goliath racketeers behaved rather differently from most people: they were more intent on growth, status, and aggression than the average flesh-and-blood human. Goliaths competed, and elites fought over the surplus of lootable resources within them. Dominance hierarchies metastasized like a cancer: growing, spreading, and co-opting or conquering the groups and civilizations around them.
Yet they were also self-terminating. They frequently contained the seeds of their own demise: Goliath’s Curse. The driving engine was that wealth inequality tended to increase. That then spilled over into inequalities of power, whether that be in autocracy (inequality in the control of decisionmaking), theocracy (inequality in the control of information), or rule by the military (inequality in the control of violence). Societies became more extractive as time passed. The social landscape was increasingly tilted towards the interests of the most powerful, be they plutocrats, priest-kings, military dictators, or entrenched bureaucrats.
The benefits that Goliaths could provide began to dry up while the costs increased. Early on, this frequently engendered a strong reaction from people with fresh memories of egalitarianism. In weak Goliaths with few mechanisms of control, people often easily rebelled or fled. This was the Primal Curse of Goliath and could lead to completely abandoned cities or a boom-and-bust cycle among early farmers and kingdoms.
Larger empires had a different curse. They could control their citizens more effectively, but inequality and societal top-heaviness drove corruption, elite factionalism, poor decision-making, overexpansion, environmental degradation, and immiseration of the masses. The state was slowly hollowed out to the stage where shocks – drought, floods, war, disease – could split it apart. This diminishing return on extraction was the next iteration of Goliath’s Curse.
Eventually, far larger colonial empires that could cross the oceans developed. They seem to have largely escaped the curse. Their technological and economic growth was so rapid that even declines rarely spiralled into collapse. Instead, the curse was the damage they inflicted on other societies across the world. They constructed a global system of extraction built on slavery, dispossession, genocide, collapse, and shatterzones. It created a rimless wheel of global empire where resources were extracted from a periphery which often experienced societal collapse, while the core received resources and grew increasingly resilient. We see a similar pattern in the world today. State failure is largely limited to states that have a particularly harsh history of colonization, which left behind broken, extractive institutions. This Global Goliath now has a different curse. The rimless wheel of empire now produces catastrophic risk which is capable of destroying the entire world system. The production of catastrophic risk is largely limited to a few powerful, secretive actors. These are the Agents of Doom: the fossilfuel industry, big tech, and military-industrial complexes. They are rackets that both cooperate and compete with each other to produce global risk for private profit and power. The Global Goliath has also incentivized the creation of ever faster and more interconnected systems. Whether you are worried about climate change, nuclear weapons, or overextended just-intime supply chains, the answer is the same. It is the Goliaths and Goliath traps that began evolving millennia ago.
Throughout history, collapse has usually been a good thing for most citizens. There could be costs, especially when it came to increased violence. This was due less to collapse than to status seekers battling to create a new state. The bloodiest and most disturbing processes throughout history have been due to the birth and expansion of a Goliath, not its fall.
This was especially true of the creation of today’s Global Goliath through the marriage of colonization and capitalism.
We now live in a world system that is transforming into a Silicon Goliath.
We have entered the endgame. Either we kill Goliath or we potentially face a final collapse that might succeed where the ice age failed and turn us from Homo solus into the final footnote of our species’ history.
To rephrase E. O. Wilson, the problem of humanity is that our Bronze Age institutions suppress the best and bring out the worst of our Palaeolithic psychology to produce humanity-killing technology. It’s less catchy, but more accurate. Or we can drop Wilson’s formulation and just say that the darker angels of our nature are flying us towards evolutionary suicide.
JUSTICE LEST THE WORLD PERISH There is an old Latin maxim Fiat justitia, et pereat mundus, which roughly translates to ‘Let justice be done, though the world perish’. It hints at what many see as an underlying tension between justice and resilience.3 It’s common to hear similar sentiments today. We often hear that democracy must be sacrificed to cut emissions, unleash technology, or prevent terrorism. That we need to sacrifice privacy on the altar of national security. The we need to restrict freedom and democracy to make it out of the twenty-first century alive.4 Yet all the evidence we do have suggests that more inclusive institutions tend to make better, more future-focused decisions in the long run. We’ve already come across evidence from fields ranging from computer science to forecasting suggesting that bigger, more diverse groups that share information reach better decisions.5 Similarly, one of the few studies to look at what types of societies make the best long-term policies found that democracies significantly outperformed autocracies.6 Giving in to the impulse to embrace despotism is not wise. It is simply repeating a mistake from thousands of years ago, on a global scale. It is a decision built on following a darker angel of our nature rather than evidence. It is turning to the authoritarian impulse when we should be turning towards each other.
We live in a remarkably lucky world. Preventing collapse and escaping the endgame requires building a better one: lower carbon emissions and fewer deaths from air pollution; fewer weapons and more money for schools and hospitals; more involvement of regular citizens in governance; less corruption; fewer plutocrats and more democracy.
These are frequently what populists from both the left and right are calling for. Supporters of both Donald Trump and Bernie Sanders would agree on the basic idea that politics and the economic system no longer represent or fairly support most people and that democracy is fraying and needs to be resuscitated.
The challenges of populism, avoiding collapse, and building a better world all have the same solution: the redistribution of power. We must use technology to help us run open democracy at scale (with widespread use of deliberative assemblies and juries) and keep inequality in check.
Those who are worried about the rise of populism need to start listening to the anger of the people around them rather than vilifying them. They need to offer a path that is more appealing than an authoritarian impulse fed by scapegoating minorities, protestors, and immigrants. That is not one based on bland pleas for hope or growth. It requires real, concrete change: introduce wealth taxes and progressive taxation; crack down on corruption (including banning legalized bribery like the revolving door between politics and industry); and give people a genuine direct say in government and workplaces through open democracy.
Such changes are eventually likely to be embraced and celebrated.
When people speak of ‘making America great again’ they are thinking of the 1950s, a time when the tax on the highest income bracket was over 90 per cent, unionization was high, and wealth inequality was low. When people think of the glory of Classical Greece, they are usually thinking of Athenian democracy and its egalitarian culture. Now we have the opportunity to build societies without the patriarchy and racism of the 1950s US, or the slavery of Athens, and with far greater material comfort and well-being. It is not a matter of trying to achieve justice even if it destroys the world.
It is fiat justitia, ne pereat mundus: let justice be done lest the world perish.
MEMORY OVER MYTH Escaping Goliath’s Curse will be a Herculean task. We need to reroute the path of human history. Making it out of the endgame will require deeprooted and systemic changes.
Such reforms will clash with myths that have evolved over thousands of years. The most entrenched is that subjugation is a necessity. We now live in a world that would have been unthinkable to our ancestors: one filled with bosses, managers, masters, and dictators. Each has some kind of dominating power over those they control: bosses can fire their employees, landlords can evict their tenants, dictators can jail or execute their citizens, and gods can eternally punish and torture their creations.
Many would claim that we intrinsically need elites and hierarchies. That the rich and rulers (and even domineering gods), despite the problems they create, are necessary to any well-functioning large-scale society.
The problem is that the term ‘elites’ has always been a misnomer. What we need is real elites: people who are exceptionally skilled at particular tasks and jobs, be they plumbers, cleaners, caretakers, researchers, doctors, or teachers. We want and need people to train the next generation, take care of the most vulnerable, and use human knowledge to improve our freedom and well-being, whether that be creating new antibiotics or helping forge new forms of democracy. We have little evidence that what we traditionally call elites – the richest and most powerful throughout history – have the same proven skill as athletes, cooks, artists, and writers. We have no reason to believe that the pharaohs and kings of the past were particularly good decision-makers.
On the contrary, societies that overthrew their initial dominators remembered them very differently: Native Americans recalled the rulers of Cahokia, the old Puebloan societies, and others as evil wizards and gamblers, not great leaders of decorated civilizations. Most of our history is viewed through the lens of the 1 per cent. The Native Americans and the people of Zomia glimpsed history through the eyes of wary survivors.
Today we have similarly little evidence that CEOs or investment bankers are particularly skilful at their jobs. Companies may be successful, but it is difficult to say whether they owe their success to the individual brilliance of a leader or founder (usually the preferred option), the grit and skill of their workers, or simply pure dumb luck. We are all fooled by randomness. Let a random group of gamblers haggle long enough on the stock market and eventually a billionaire will emerge. We’ll attribute their success to their personal philosophy, diet, or morning routine, but in reality it was just fortuitous. It is far more difficult to disentangle individual skill from sheer luck in finance or commerce than it is with a firefighter, an athlete, or a chef.7 We rarely have good evidence that those in elite positions with enormous salaries are giving an elite performance. (For instance, financial advisors’ stock picks appear to perform no better than random chance.8 ) In general, the higher up the ladder of a corporation or a hierarchy someone ascends, the less evidence there is of personal skill contributing to performance.
Statistician Nassim Nicholas Taleb calls this ‘the inverse rule’.9 That inverse rule is built upon the 1 per cent view of history and the countless myths of elite superiority: from the idea that kings have a divine right to rule to the notion that the world’s biggest companies are dependent on the genius of one or two individuals rather than the tireless labour and courage of underpaid workers in schools, hospitals, mines, and factories.
The algorithms and skyscrapers of today, like the pyramids and great walls of yesteryear, are built on the sweat of tens of thousands of ordinary labourers. The large text models of today are not founded on the intellect of Sam Altman but on the work of thousands of academics who pioneered machine learning, the countless workers who dig up and process the materials, and the myriad underpaid and unpaid individuals who provide the data and train the machines. Our greatest achievements are not because of the 1 per cent; they are collective endeavours. As George Eliot once wrote, ‘The growing good of the world is partly dependent on unhistoric acts … half owing to the number who lived faithfully a hidden life, and rest in unvisited tombs.’10 It’s time to drop the 1 per cent view and start paying attention to the unmarked tombs and unknown heroes upon whose shoulders we’ve built our world.
Another myth we’ll need to overcome is that we can’t reshape society and upend dominance hierarchies. We live in a peculiar world in which many believe that we can split atoms and even recreate intelligence on silicon, yet it is impossible to do democracy at scale or resolve international arms races. This bafflingly selective failure of imagination is not because of evidence. We’ve controlled arms races before and have the technology and know-how to have large-scale, intense democracy. It is instead just another useful story of subjugation: we can hope for technological singularities but not social emancipation. The truth is that we can change our social structures and escape Goliath traps. And we’ll need to. Reshaping the world to avoid races (or rackets) for status, arms, and advantages is our great challenge. It won’t be easy, but it is possible. And it could determine whether this is our finest century or our final century.
Dispelling these myths will be difficult. It’s difficult to find a utopian idea among the Greeks or Romans that did not involve a hierarchical world.
The thinkers of antiquity were far more capable of imagining the abolition of marriage than war, patriarchy, or class domination. Similarly, today, the feverish techno-dreams of billionaires involve reproducing human intelligence on silicon, colonizing far-flung icy planets, and harvesting stars, but tell us little of whether we have equality between the sexes.
Our world and our myths are the product of thousands of years of Goliath evolution.
A central myth we’ll need to unravel is the myth of progress. History is a not a story of improvement from the stirring of civilization and agriculture through to the rise of empires to the skyscrapers and satellites of the modern world. Agriculture and sedentism took a wrecking ball to our physical and mental health. The rise and fall of empires did result in bloodshed and hardship, but it also often meant the fall and rise of our heights and freedoms. Technology did not automatically improve the overall human condition; the benefits were usually eaten up by a select elite class.
Technology can still be pivotal and beneficial: vaccines, plumbing, anaesthetics, antibiotics, central heating, the printing press, and so much more have helped to improve the standards of living of billions. Yet we need to realize that shared progress in health and well-being came only when technological developments were paired with collective struggle; that to have the benefits of technology without the catastrophic risk of climate change, nuclear weapons, killer robots, and novel entities, we’ll also need to struggle for democratic control of how technology is produced and used. We’ll need to fight to escape Goliath traps. And that fight will need to be waged every day.
Without that fight, our Silicon Goliath is likely to take us on a path to either an unfree world or a global collapse. A potential, final curse.
It won’t be easy, but anything worth fighting for rarely is. Even if there seems to be little hope, take comfort in defiance. It is worth doing the right thing even if the odds are stacked against us.
The world is not a Greek tragedy.
11 We can’t rely on a deus ex machina like AI solving our problems.
Real progress and an escape from Goliath’s Curse will come only from the battle to rebalance power. We can write a better story together. A story not about collapse or rise and fall. Rather, it would be a story of redemption. A story about rebuilding civilization.
Epilogue: Slaying Goliath Escaping the Endgame – Recreating Civilization – Becoming a David ESCAPING THE ENDGAME Collapse may seem inescapable. This is an illusion. Most of the challenges we face are entirely solvable. If our world falls apart in the cold of a nuclear winter or the unending blaze of climate change, it won’t be because there was nothing we could have done. There are many ways to reduce the risk of global collapse, to defeat Goliath, and even actions that individuals can take.
Staying Within Planetary Boundaries Every large country now has a ‘pathway to decarbonization’: a detailed road map made by large groups of experts detailing how to decarbonize their society and reach net zero emissions. Most of these plans depend on a few pillars: switching to renewable energy for electricity production, electrifying transport, adopting aggressive energy efficiency measures, and reducing any remaining emissions through reforestation and direct air capture (machines which directly suck greenhouse gases out of the atmosphere and sequester them).1 Reducing our consumption of meat and dairy would also help cut emissions, as well as enormously reducing the amount of suffering in the world.
One key way to slash emissions is to stop supporting fossil fuels.
Currently governments around the world reduce the price of fossil fuels through measures such as tax breaks. More than that, the price of fossil fuels doesn’t include their costs to society, such as the roughly 5–10 million deaths caused annually by fossil-fuel air pollution.2 These are costs that we all pay. All of these are ‘subsidies’ on the price of fossil fuels. Currently fossil fuels are subsidized by around $7 trillion per year, or around 7 per cent of global GDP.
3 Scrapping subsidies and properly pricing carbon could both dramatically cut emissions and save lives. The money saved by no longer handing out tax breaks for fossil fuels, as well as the revenue from a carbon tax, could generate a stream of funding worth around 3 per cent of global GDP. That is enough money, according to some estimates, to eradicate extreme poverty, provide universal access to sanitation and clean water, end hunger, and still have over $1 trillion to spare.4 Decarbonizing the world will be a momentous effort, but it won’t necessarily be an expensive one (unless you are a fossil-fuel company).
The IPCC in its latest review suggests that limiting global warming to 2°C will mean global GDP losses by 2050 of 1.3–2.7 per cent, or 2.6–4.2 per cent to restrict warming to 1.5°C.5 Understanding the world purely through GDP misses many things, including less volatile energy prices and millions of lives saved from air-pollution-related deaths. Once you factor these in, decarbonization could well be a net economic benefit.6 (Although quibbling over whether saving the world will be a net economic cost is perhaps the height of stupidity.) Climate change is just one of many planetary boundaries we face.
Staying within planetary boundaries will mean giving up the elite dream of endless economic growth. There are many ways we can progress as a society while remaining within planetary boundaries. These include reducing the working week to three or four days, dropping GDP as an economic target in favour of more holistic measures like the Genuine Progress Indicator (GPI) – which factors in poverty, inequality, pollution, and resource depletion – and scaling down destructive forms of production such as fossil fuels, advertising, and factory farming.7 An even more sweeping action is to ensure that the effects of products on the environment and human health – whether it be cancer and lost fertility from PFAS, or lost biodiversity – are reflected in their market price. This would be a seismic shift in our economic system built on just one small common-sense change: companies should have to pay for the economic and social damage they do. It is simply making our economies more truthful and realistic. This would not leave us in shared poverty. We can provide decent living standards for 8.5 billion people using just 30 per cent of existing global energy and resource use.8 Controlling AI Controlling AI is just as doable as cutting emissions. We can – and should – ban the development of killer robots and automated general cognition (AGI). Technology bans may seem unrealistic, but there are hundreds of examples of technologies throughout history that have been developed and rescinded or intentionally never developed.9 There are leverage points in every part of the supply chain for building advanced algorithms. One way to both create a fairer economic system and indirectly slow AI is by paying workers properly along the entire supply chain, from the miners in the Congo, to the factory workers in Taiwan and China, through to the cleaners in San Francisco.
AI is dependent not just on cheap labour but also on stolen data. Giving citizens control of their data and properly compensating them for it would not only be the fair thing to do, but would also substantially slow the development of AI by forcing AI companies to actually pay for the resources they currently use for free. Rather than it being stolen through the digital trawl-nets of OpenAI and others or having it silently (and often unknowingly) taken as part of archaic user-service agreements that no one reads, people could instead sell their data to companies who want it through ‘data unions’ that would negotiate a fair price (rather than every individual having to bargain with Google or Apple).10 Three key ingredients for AI are computing power (in the form of microchips, also frequently called ‘compute’), energy, and coders. Compute is concentrated in a few supply chains that flow primarily through Taiwan.
There are multiple ways that a coalition of countries could cooperate to restrict and slow the supply of compute. One is to put a cap on the size of a compute cluster, or how much can be used in a single model.11 Pricing carbon to reflect the true cost of electricity or setting a cap on the maximum energy usage of data centres or models would also act in a similar way to slow AI. Whistleblower protections and stronger unions could also help budding tech workers organize against companies and programs that lack scruples.12 Once models are released, legislation could hold companies responsible for the harms they cause (as the vetoed SB 1047 bill in California attempted to do) and limit their use in dangerous domains, whether that be banning facial-recognition algorithms or the incorporation of AI into the miliary.
13 A World Without Nuclear Weapons Dismantling nuclear weapons stockpiles is also eminently possible. The US was typically dismantling 1,000 or more nuclear weapons every year during the 1990s. With about 10,000 weapons in the world today, a global effort could easily see stockpiles eliminated within a decade.
Studies have shown that many countries could plausibly rid themselves of nuclear weapons within ninety days.14 That could be done through an international treaty, and indeed we already have one: the 2017 Treaty on the Prohibition of Nuclear Weapons (which binds countries to not use, test, possess, acquire, or develop nuclear weapons), a treaty that no nuclear power has joined but seventy-three others have.
A ban is needed because no level of nuclear weapons is safe. As a minimum initial step, countries could reduce their stockpiles to a low enough level that it would provide a ‘deterrent’ against war. That could be around 300 low-yield weapons per country.
15 Yet even a regional war between Pakistan (170 weapons) and India (172 weapons) would still have global impacts, dropping temperatures by 4°C.16 That is to say nothing of the millions of lives lost and the knock-on effects. Weapons that can kill millions indiscriminately will never be safe and any society that has them can never call itself civilized. Our species can never truly be worthy of the mantle ‘sapiens’ (‘wise’ in Latin) until this scourge is lifted from the planet. The Problem Is Power The natural rebuttal to each of these measures, such as banning AGI, killer robots, and nuclear weapons, is the defence of political feasibility. It is seemingly a golden rule of the politics of collapse: what is effective is unfeasible and what is feasible won’t be effective.
Political feasibility may seem like an immovable force, but it can change rapidly. During the spread of Covid-19 governments enacted far-reaching measures that previously seemed unthinkable. When countries do act decisively, miraculous change is possible.17 For many it’s somehow easier to imagine a nuclear winter or even an AGI apocalypse than to imagine a world without abhorrent weapons of mass destruction. It’s easier to imagine the end of the world than the end of Goliath. That is because we are still stuck in Goliath traps: status races, arms races, and races to the bottom (where firms or states try to outcompete each other by exploiting people and environments more intensely).
Decreasing nuclear stockpiles, slashing carbon emissions, and making our societies more democratic are all completely feasible, and doing so in the long term will mean escaping these traps and digging up the root causes of existential risk. We’ll need to do what few societies have successfully done: kill Goliath.18 RECREATING CIVILIZATION Today we have measures that can level each source of power, reduce Goliath fuel, and tame the darker angels of our nature.
Democratizing Political Power We need to introduce open democracy, with deliberative juries and assemblies creating national policies for governments and providing oversight of corporations. If decision-makers are randomly selected from society, we will no longer be selecting for those who crave status and power or who rank higher in the dark triad. The constant cycling through new citizens to make decisions will also help ensure no one is in authority long enough (or holds enough power) to be corrupted. But we don’t just need to rely on physical assemblies of randomly selected citizens as in ancient Athens. New technologies allow for online platforms to aggregate the opinions of enormous numbers of people when making critical decisions about how we live.19 We’re often told that open democracy is not practically possible, but that has always been an argument for hoarding political power.
This is just the tip of the iceberg. There are many other options, such as banning the revolving door between regulators and industry. If we start to properly fund experiments in democracy, and reward them like we reward start-ups in tech, then who knows in what profound new ways we’ll be able to harness the wisdom of crowds.20 Democracy is not just for governments. If you believe that democracies work because they result in better decision-making, and that all of us deserve a say in collective endeavours, then there is no reason to accept workplace autocracy. The problem throughout history has been not states (some states have been, and many are, important providers of social support, public services, and public goods) but concentrations of unaccountable power. Corporations such as Exxon or the East India Company have been just as destructive as empires, and firms today are less accountable and less democratic than most states. Corporations could similarly be democratized by reforming them into large-scale worker cooperatives governed by workers alongside deliberative juries, and with an overriding legal goal of providing social and environmental benefits, not short-term returns to investors.
Make the World Equal Again Levelling political power won’t be enough. Even a deep democracy that uses citizens’ assemblies and juries, as well as digital technology to carry out regular direct votes, will eventually be undermined if some people have billions of dollars with which to rig any system we come up with. One could argue that we have democracy and massive inequality in the world today, and that the two are not in tension. It’s a mistaken idea. What we ordinarily call democracy – systems in which a subset of people (who are aggressively propagandized by political marketers and billionaire-owned media empires) vote every four to five years for a tiny number of (usually rich) representatives who are funded and lobbied by corporations (for whom they frequently work afterwards), who then enact policies which usually better represent elite interests than popular opinion – is better described as an oligarchy with democratic furnishings.
It is far more inclusive than most governments throughout history, but that is a low bar compared to what is possible. Even this threadbare democracy is being frayed by increasing wealth inequality. It is no coincidence that (with a slight time lag) democracy started backsliding after inequality began rising across the world in the 1970s. Today, the grinning face of many ailing democracies is that of tech and property billionaires. Inequality in one form of power or another will eventually spill into others: the rich buy elections, overly powerful generals launch coups, and autocrats amass fortunes. The original and most powerful inequality was found in wealth: lootable resources provided a form of power that could be easily translated into other forms of power and passed down through generations. Political democracy will die without economic equality.
Wealth becoming more unequal may be close to an iron law throughout history, but reversing it is surprisingly easy. The simplest way is through taxation. The US had an income tax of above 90 per cent on the highest earners from 1944 through to 1963 (the highest rate in the US is now 37 per cent). It didn’t lead to an economic bust. It actually helped to usher in an economic golden age for the US. A highly progressive taxation system should be combined with an even higher tax on wealth such as land and stocks.
Other measures include placing a cap on wealth, $10 million for example (a level that is well beyond what any individual needs), or capping the income of the highest earners within companies at five times that of the lowest-paid worker. While tax dodging and tax evasion may seem like the inescapable consequence of such proposals, there are a raft of ways to prevent this, including establishing a coalition of countries that set a minimum tax rate on large multinational corporations and penalizing tax havens that charge less.21 There is also an abundance of ways to reduce global inequality. One option is a global minimum wage, which could be half the median wage of a country to take account of local conditions. Another is a comprehensive debt jubilee – in other words, writing off all debts – a policy that has been surprisingly common (and successful) throughout history.
22 Democratic Control of Information and the Military One of the best ways to distribute the control of information is to break up existing monopolies, specifically in big tech and the media. This could mean creating new legislation as well as rigorously implementing existing anti-trust tools, such as the Sherman Act and Clayton Act in the US, which have already been used to block big mergers.
Another is to make power structures more transparent. Governments’ classification and secrecy practices, particularly for military projects, are excessive, lack structure, and are often used to hide incompetence and corruption rather than important state secrets.23 Corporate non-disclosure agreements are little better, as we’ve seen with OpenAI and others cracking down on safety-concerned whistleblowers. The foundation of any democracy is a well-informed electorate, and thus transparency should be the rule, not the exception. Secrecy should be permitted only if it has been signed off by an independent arbiter or a citizens’ jury. Another key step will be empowering and protecting whistleblowers, whether it be in the military or in big tech. That means providing stronger legal protection for those who have the courage to report misconduct and repealing archaic laws that allow governments and corporations to hunt, silence, and imprison whistleblowers, such as the Espionage Act (1917) in the US.
Rebalancing power will inevitably require increased protest and activism in the coming decades. Yet governments have made a concerted recent effort to curtail the right to protest, such as through the UK’s Police, Crime, Sentencing, and Courts Act 2022, or the 323 different bills that have been introduced in the US since 2017 that restrict the right to protest, and even encourage violence against peaceful protesters.24 Finding ways to roll these back and unshackle concerned citizens will be critical to levelling power and holding the Agents of Doom to account.
Democratically Controlling Technology and Production We can also slow – or even halt – the production of dangerous new technologies and weapons by placing them back under democratic control.
Monopolizable weapons and technologies under the control of the few have been the norm for just the last 1 per cent of human history. This is not a call to halt technological progress; it is a call to democratically control technology to ensure it is progress. AI can be used to better spot cancers, vaccines have prevented the suffering of billions, space exploration ignites the enthusiasm and imagination of millions, and we’ll need new technologies to capture carbon from the atmosphere and stabilize melting ice sheets. It is a question of better directing technological development and production.
Researchers have already suggested ways in which companies and countries could use experts to map out the most likely development pathways of different technologies (such as AGI) and then use deliberative juries or assemblies to decide whether to pursue them and what regulation to put in place. This would help us democratically and pre-emptively regulate technology, as opposed to the current losing strategy of rushing haphazardly to make legislation after a new technology is released.25 We could use a similar democratic process to navigate difficult trade-offs on the path ahead, such as whether we should use emergency options (these might include negative emissions, short-term geoengineering, or interventions to stabilize ice sheets) to avoid the worst effects of climate change: examine the potential worst-case effects of both the warming we are facing and the potential emergency solutions, and then let a group of expert-informed citizens decide.26 This is not a complete blueprint for solving our problems.
Trying to foist a vision of utopia on the world has never ended well.
Instead, we should democratize the world, level power, reduce Goliath fuel, reduce existential risk, and see what world blossoms from this new civilization. BECOMING A DAVID In the biblical tale, Goliath is slain by a single slingshot-toting future king by the name of David. Defeating Goliaths, in reality, is not done alone.
Instead, it requires actions that leverage the people and communities around us.
Don’t Be a Dick First, I propose a simple pledge to not be a dick. This is a pledge to not work for, invest in, or support any firm, institute, or individual that significantly contributes to global catastrophic risk. Don’t work for an Agent of Doom, whether it is an AGI lab, a fossil-fuel company, or an arms manufacturer. The idea that you are going to change an Agent of Doom from the inside is nonsense, and the excuse of ‘if I don’t do it someone else will’ is the sort of bullshit used by guards at concentration camps. Hold yourself to a higher standard. Stopping arms races, status races, and races to the bottom begins when each of us refuses to get trapped in them.
More than that, it is a pledge to not become a Goliath yourself. All of us rank somewhere on the dark triad. All of us are capable of being corrupted by power. All of us crave status to some degree. Don’t let the darker angels of your nature win. If you find yourself atop a hierarchy, democratize it and distribute your power among your peers. Every year, try to make sure that your actions leave the world a little freer than it otherwise would have been.
This same pledge can be adopted not just by individuals but also by institutions. While Google did a poor job of fulfilling its credo of ‘don’t be evil’, all is not lost. Corporations can and should reorganize themselves to be more internally democratic and minimize the pay discrepancies between the CEO and interns (who should be paid). Firms can also use their money and portfolios to fight against doomsday. Large institutional investors like pension funds and sovereign wealth funds can and should pledge to use their investments to minimize global catastrophic risk (not just moving their investments away from fossil-fuel companies, big tech, and arms producers, but moving them into sectors that help make the world more resilient).27 No institution should be financing our global descent, and no one should support one which does.
Be a Democrat Second, practise democracy. Democracy is not just a form of government; it is a culture and a way of life. One that all of us need to recapture. Be a democrat: don’t just vote every few years, but join a union, join an activist group (while there is no group fighting against all the sources of catastrophic risk, there hopefully will be one soon), advocate for workplace democracy, and discuss political matters in a productive way – whether it be nuclear weapons or mass surveillance or climate collapse – with friends and family members.
Vote Against the Apocalypse Third is voting. You are first and foremost a citizen, not a consumer.
While democracy is more than voting, voting is a fundamental part of democratic practice in the modern world. If you know about climate change, a nuclear winter, and the other real risks to the survival of our species but base your vote on who promises to cut your taxes, then you are culpable for our current path towards self-destruction. It is a travesty that no election to date has been decided by the candidates’ positions on nuclear weapons and climate change. There is no reason to contribute to that trend in the future.
Choose the party and candidate that offer genuine, practical ways to empower citizens through deliberative juries and assemblies, not those who use empty rhetoric about ‘draining the swamp’, ‘protecting the nation’, or providing ‘change we can believe in’. Choose those who have concrete plans to adopt a high carbon price, to price environmental impacts in general, to ban fossil-fuel subsidies, to join the 2017 Treaty on the Prohibition of Nuclear Weapons, and to place a ban on the pursuit of AGI and killer robots. If there are none, then contact your local representative and lobby for it, or join a grassroots movement that is fighting for real change. Voting may seem inconsequential, but it can change things extremely quickly. If everyone in Australia (or the US, or the UK) began switching off their lights prudently and chose the lowest-carbon travel options, then it would barely make a dent in national emissions. If they changed their vote to those offering the strongest decarbonization plans, then suddenly emissions could be on track to be eliminated within decades. You of course only have one vote, but talking to family and friends can multiply your impact. Debate and discussion are the lifeblood of democracy.
Don’t Be Dominated Fourth is to re-cultivate the counter-dominance intuitions that guided us through the Palaeolithic. Oppose domination in all your relationships, whether they be personal, family, or workplace. Whenever you come across a hierarchy, whether it is based on wealth, gender, or age, ask whether it is legitimate and whether it justifies domination. If it doesn’t, then try to overturn it.
One of the first and most pernicious stories justifying subjugation was that of the saviour leader. High priests, god-kings, and budding rulers claimed to predict the future by reading bones, ensure the harvest through ritual sacrifice, and provide protection from unseen spirits and neighbouring groups. Now, we have billionaires and oligarchs offering salvation by building walls between us, or inventing a superintelligent AI god to solve our problems, or building colonies on Mars as an escape plan. These have always been – and still are – glorified parlour tricks to keep people distracted while the bosses pick their pockets.
It is time to say ‘enough’. It is time to realize the bright and terrifying truth: no gods, kings, heroes, or masters are going to save us. Slaying Goliath and avoiding evolutionary suicide is, like all great achievements, going to be a collective action. It is on us. It is going to take the thankless, unglamorous work of having difficult conversations with your friends, giving up power, trusting your fellow citizens rather than a strongman leader when you’re scared, going into the streets even when you are comfortable at home, and taking an ethical stand even if it might cost you your job. Each of those little acts is another stone flung at Goliath, and each crack in its skull is a doorway to freedom. The stand will be worth it. A world free of nuclear weapons, carbon emissions, killer robots, and the threat of collapse, one full of genuine democracy and economic equality, is no utopia. It is within reach.
Revolutions begin where myths end; Goliath ends when civilization begins.
Acknowledgements A book is never a solo endeavour. I’m grateful to all those who helped bring this one to fruition. While the manuscript has been improved by the comments of this small army of scholars, friends, and editors, any mistakes that remain are solely my own.
A special thanks to my editors Greg Clowes and Todd Portnowitz, as well as my agent Chris Wellbelove, who all read through multiple versions of the manuscript and provided hundreds of comments in response. Greg was a particularly prolific, insightful, and challenging source of feedback. It is immeasurably better due to their efforts. I’m grateful for the exceptional copy-editing skills of Trevor Horwood, who did the hard work of trawling through hundreds of pages to sharpen the prose and clarify the messages. I would also like to thank Robert Drew and Anthony Hippisley for their careful and meticulous proofreading of the final manuscript.
My eternal gratitude to my ever-reliable personal freelance graphic designer David Kemp. He helped design over twenty of the figures in this book, including all of the timelines. I’d also like to extend my warm thanks to the other graphic designers who helped to redraw and finalize the figures.
This book wouldn’t exist (at least not for many years) had it not been for Richard Fisher and Tom Killingbeck. Tom, an editor at Penguin at the time, unexpectedly reached out to me in October 2019 to ask if I would consider writing a book. He walked me through the process of writing a proposal and shepherded it through the approval process. Thanks for taking a gamble on a young, little-known academic, Tom. That initial deal was based on an article for the BBC that my editor Richard Fisher had commissioned me to do. Since then, Richard has been a consistent source of advice in the world of writing and also reviewed the entire book. Thank you, Richard.
I am an outsider to many of the topics covered in these pages. I have luckily been able to rely on the deft advice of several native guides over different disciplines, to whom I am deeply indebted. The first is the historian of empire John Haldon. John read the manuscript in its entirety, was one of the first to hear the full argument of the book, and has been a continuous source of advice, including on how to have a healthy sense of scepticism about collapse. Another is the archaeologist Eric Cline, who also read the entire book and was my primary guide to the Late Bronze Age collapse. A third is the polymath and historian Walter Scheidel, whose ideas have left fingerprints across the pages you’ve read. Walter provided some early words of encouragement about my theory of collapse in 2021, which helped give me the confidence to stay the course with this sprawling project. A fourth is Benoît Pelopidas, whom I’m lucky to have had as my primary guide to all things relating to nuclear weapons, as a reviewer for all of Part Three of the book, and as a friend. I owe a deep thanks to each of them.
I also owe a thank you to all those who read through different parts of the manuscript and provided useful comments and edits: Dan Lawrence, Cecilia Padilla-Iglesias, Marilyn Masson, Douglas Fry, Lalitha Sundaram, Nate Hagens, Gideon Futerman, Thomas Homer-Dixon, Makella (Makky B) Brems, Scott Janzwood, Michael Lawrence, Jennifer Birch, Ajay Gambhir, Jonathan Neale, Carla Zoe Cremer, Rachel Ainsworth, Cristina Jauset Gonzalez, Elizabeth (Coups) Cooper, Georgiana Gilgallon, Haydn Belfield, Marten Scheffer, Adrian Hindes, Tom Matthews, Megan Shipman, Aaron Tang, Daniel Zimmer, and Constantin Arnscheidt. Over the years many others have contributed to the research which underpins this book, whether it be through written pieces or just illuminating discussion. These include Tim Lenton, Xu Chi, Lara Mani, Tom Hobbes, and my other colleagues past and present at the Centre for the Study of Existential Risk (CSER), the Cascade Institute, the Notre Dame Institute for Advanced Studies (now the Institute for Ethics and the Common Good, or the ECG), and Notre Dame Environmental Humanities Initiative (EHUM). Thanks to Oscar Rousham for helping to put together the first version of the MOROS database, and to Tillman Hartley for helping to compose and refine the latest versions.
Dan Hoyer, Guy Middleton, and Tim Kohler deserve a special thanks as they all read through almost the entire manuscript and provided a series of short essays in response. I’m indebted to them for their attention to detail and the time they spent on improving this manuscript. Florian Jehn provided particularly copious and insightful comments for the entire book.
His Substack ‘Existential Crunch’ is also an invaluable source of knowledge on collapse and existential risk, one I would highly recommend to anyone who wants further reading. The award for most excruciatingly detailed and helpful comments undoubtedly goes to Matthijs Maas. Matthijs left myriad comments on the manuscript (including in the form of hours of voice notes), talked through some of the biggest ideas (and even coined the term ‘Primal Curse’ for the initial iteration of Goliath’s Curse), and has developed the helpful habit of consistently sending me relevant articles, chapters, and books over the years. It’s hard to imagine a better friend and colleague.
Two of my closest friends and colleagues, Mel Cowans and Nathaniel Cooke, reviewed the entire book meticulously. They, along with Melissa May, were ever reliable sources of support and advice throughout the years and even housed me for portions of the writing. Alongside Olivia Liang and Raph Gall (as well as Arthur Liang-Gall) they became my adopted UK family. I’m also grateful to my Australian family, who have been pillars of support over the years: Ellie, Kristy, Tammy, Mum, Swaran, and David (as well as our furry family members Bo Bo, Bailey, and Bahni). Thanks to Emily Arnold, who made the final months of finishing this manuscript far brighter and far more fun than they should have been.
This book has been written across several years and many homes, and I’m grateful for each one. These include CSER, the Meridian Office in Cambridge, the Global Systemic Risk group at Princeton University (where I undertook a sabbatical in summer 2022 thanks to the hospitality and guidance of Miguel Centeno and Thayer Patterson), as well as the ECG and EHUM at the University of Notre Dame. Notre Dame housed me for the most intensive writing periods and I’m appreciative of the help and (typically Midwestern) hospitality of Roy Scranton, Meghan O’Sullivan, the ECG staff, and 2023–24 faculty fellows (especially my fellow deep history enthusiast Marcus Kreuzer and fellow freedom fan Makella Brems).
This book wouldn’t be possible without the brilliant ideas and tireless work of others. In particular it is heavily indebted to the work of Peter Turchin, especially through his research into structural demographic theory and state formation using the Seshat database. Many of the later arguments of moving beyond growth and the travesties of the global Goliath are inspired by studies led by Jason Hickel. Much of the work on abandoned experiments draws from the excellent scholarship of Justin Jennings. Much of the spirit of this book, as well as the central concepts – such as lootable resources and the art of not being governed – were inspired by the work of James C. Scott. Tragically Jim passed away shortly before this book was completed. I hope he would be pleased with how it has carried forward his ideas.
The greatest thanks and the dedication of this book are reserved for my brother. Not just because of the poetry of devoting a book called Goliath’s Curse to someone called David. David has been over the course of my life a brother, mentor, inspiration, and even a best friend.
He helped me through the most difficult times while writing this manuscript, endured the punishment of being the first reader for every draft, and was a patient sounding-board for many of the ideas. This book wouldn’t exist if it wasn’t for him. Thank you, brother.
Glossary Authoritarian impulse The psychological tendency for individuals to be more accepting and supporting of authoritarian structures and leaders when faced with a threat to their safety and security.
Caged land Territory with few exit options. This is a form of Goliath fuel that tends to lead towards the establishment of dominance hierarchies.
Caged land can include geography (such as Egypt, which is cordoned off by desert and sea); a population tied to certain resources (such as a stream of salmon or a field of wheat); or technological changes (such as the construction of walls). All of these can – and do – interact and change over time.
Chaos A social system in which everyone is trying relentlessly and violently to dominate, but no one ever succeeds.
Civilization A social structure characterized mainly by non-hierarchical, democratic cooperation.
Collapse The relatively rapid contraction and fragmentation of different power systems. This results in large-scale losses in hierarchy, population density, and energy capture.
(Cumulative) cultural evolution Knowledge, practices, and technologies that are selected for, passed down across generations, and accumulated over time. Dark triad Three antisocial personality traits: psychopathy (callousness, and a lack of empathy and remorse), narcissism (an inflated sense of entitlement and self-importance), and Machiavellianism (manipulating others for personal gain).
Darker angels of our nature A combination of psychological traits that have led to Goliaths: status competition through domination; the dark triad (psychopathy, narcissism, and Machiavellianism); embracing strong authority structures when threatened (the authoritarian impulse); and becoming corrupted by power.
Death-Star Syndrome Large, robust systems that are surprisingly fragile if hit in the right spot with a sufficient amount of force.
Democracy A spectrum of (more or less) equal and inclusive political practices.
Democracies can be hierarchical (ancient Athens for adult males) or nonhierarchical (egalitarian foragers such as the Hadza).
Diminishing returns on extraction As states and elites try to extract more from their citizens, resources, and neighbours, this causes increased inequality, corruption, oligarchy (poor decision-making), overstretch, popular immiseration, elite factionalism, and a declining energy ratio.
These different effects hollow out the state until it falls apart or is broken apart by different shocks. The second iteration of Goliath’s Curse.
Dominance hierarchy A social-ranking system in which one group or individual is placed above others owing to their ability to impose penalties, including violence.
Empire Essentially synonymous with a state. For ease, we should think of these as particularly big states, or states that have conquered other states.
Endgame A situation in which there is a significant risk of global catastrophic or existential risk. In short, a time when Goliath is likely to self-terminate permanently or be fundamentally reformed. Evolutionary backsliding The movement from egalitarian fluid civilizations into dominance hierarchies (Goliaths) which more closely resemble the social structure of gorillas and chimpanzees.
Evolutionary suicide Short-term evolutionary pressures lock the population into a path of selfdestruction. Short-term adaptation leads to long-term extinction. Existential risk The likelihood of a deep, lasting, global collapse or human extinction.
Extractive institutions Great concentrations of power leading to institutions (whether it be an authoritarian government, an oppressive theocracy, or a bloated bureaucracy controlled by the rich) in which elites extract more resources, labour, and energy through unequal exchange.
Global catastrophic risk The likelihood of an unprecedented loss of more than 10 per cent of global population and disruption of the critical systems on which the world depends.
Global decimation risk The likelihood of an event or process that leads to a loss of around 10 per cent of global population within a year.
Global Goliath A collection of interconnected hierarchies across the entire world in which some individuals dominate others to control energy and labour. In our modern world this includes hierarchies such as the state, patriarchy, rich and poor (based on a capitalist global economy), and myths about the need for hierarchy.
Goliath A collection of interconnected hierarchies in which some individuals dominate others to control energy and labour.
Goliath evolution The evolution of dominance hierarchies. Violence selects for Goliaths that are better at warfare, and usually larger and more centralized. Goliath fuel Ecological and technological conditions that allow for the establishment of a Goliath. These include lootable resources, monopolizable weapons, and caged land.
Goliath traps The destructive competitive dynamics that dominance hierarchies are liable to fall into. The pursuit of short-term gain by status seekers leads to everyone losing out in the long term. There are three main Goliath traps: races for status, arms races, and races to the bottom.
Lootable resources Resources that can easily be seen, stolen, and stored. A form of Goliath fuel.
Monopolizable weapons Weapons or tactics that can be concentrated (or even monopolized) and give one group an advantage over their competitors.
Primal Curse The first iteration of Goliath’s Curse. This involves early Goliaths falling apart owing to weak institutions meeting strong counter-dominance intuitions, as well as rising inequality (and different shocks). Rungless ladder A metaphor for how more powerful Goliaths tend to have deeper and more widespread collapses. For the modern world this heightened severity of collapse is due to more powerful weapons, more deteriorated environments, and a reliance on states and complex technological systems.
Silicon Goliath A Goliath that covers the entire world where the populace is controlled through mass surveillance and dominance hierarchies, which are in turn overwhelmingly dependent on advanced algorithmic systems. The main forms of Goliath fuel for this form of Goliath are data (lootable resource), killer robots and nuclear weapons (monopolizable weapons), and mass surveillance systems (caged land).
Somalia-Denmark rule The more predatory a state, the less harmful (or even the more beneficial) its collapse; and the more inclusive and benevolent a state, the more costly its collapse.
State(s) A set of centralized institutions that impose rules on and extract resources from a population in a territory. States can also provide public goods, whether that be roads, bridges, or (in later cases) education and healthcare.
Stories of subjugation Myths, ideologies, and religions that justify the domination of the many by the few.
Systemic risk The potential for individual disruptions or failures to cascade into a system-wide failure.
Tree of Doom The causal chain from root causes through Goliaths and Goliath traps to existential risk.
Unequal exchange The unfair exchange of value between individuals or groups owing to power imbalances.