New 80k problem profile: Nuclear weapons
By Benjamin Hilton, 80000_Hours @ 2024-07-19T17:17 (+53)
Note: this post is a (minorly) edited version of a new 80,000 Hours problem profile.
As the searingly bright white light of the first nuclear blast faded away, the world entered a new age.
Since July 16, 1945, humanity has had access to technology capable of destroying civilisation.
Amidst rising tensions and the return of war to Europe, we’re potentially seeing the start of a new nuclear arms race.
Meanwhile, the community of brilliant minds who worked throughout the Cold War to prevent nuclear catastrophe has all but disappeared.
And that’s a problem.
It’s a problem because the threat of nuclear destruction is still with us. But that also means that by addressing that threat, we could make humanity more likely to endure.
Summary
It’s very plausible that there will be a nuclear war this century. If that does happen, there’s a reasonable chance that war would cause some kind of nuclear winter, potentially killing billions, and possibly causing an existential catastrophe.
Nuclear security is already a major topic of interest for governments, but has little attention from philanthropists and NGOs, so we think there are likely some neglected opportunities to reduce the risk.
Most opportunities to influence the risk from nuclear weapons seem to be through working in government, researching key questions, working in communications to advocate for changes, or attempting to build the field (for example, by earning to give).
Our overall view
Recommended. Working on this issue seems to be among the best ways of improving the long-term future we know of, but all else equal, we think it’s less pressing than our highest priority areas.
Scale
We believe work to reduce the probability of nuclear war has the potential for a large positive impact, as nuclear war would have devastating effects, both directly and through secondary effects such as nuclear winter. We think the chance of a nuclear war in the next 100 years is something like 20–50%. Estimates of the existential risk from nuclear war within the next 100 years range from 0.005–1%. We think the existential risk from nuclear war is around 0.01%.
Neglectedness
Current spending is around $1 billion per year (quality-adjusted). However, philanthropic spending is only around $30 million per year, and so we’d guess there are high-impact neglected opportunities to reduce the risk.
Solvability
Making progress on nuclear security seems somewhat tractable, and there are a variety of intermediate goals we might aim to achieve which could help. However, there’s a real risk of causing harm.
How likely is a nuclear war?
From 1945 until 1991, much of the world lived in full knowledge that a constant threat of nuclear war loomed. Schoolkids did regular drills to prepare for nuclear attack; hundreds of thousands of fallout shelters were built across Europe and North America.
After the end of the Cold War, fear of nuclear war gradually fell out of the public consciousness. But while the landscape of the threat has changed, the risk remains surprisingly — and scarily — high.
Surveys of experts and superforecasters are one way we can try to put a number on the chances of some kind of nuclear war.
Here’s a table of every estimate we could find since 2000:
Definition | Annualised probability | Probability by 2100 | Source |
---|---|---|---|
Nuclear weapon kills > 1,000 people | 0.54% | 33.34% | Existential Persuasion Tournament (2022), superforecasters |
Nuclear weapon kills > 1,000 people | 0.61% | 37.09% | Existential Persuasion Tournament (2022), domain experts |
Nuclear attack | 2.21% | 81.50% | Lugar expert survey (2005) |
Nuclear war | 1.00% | 53.18% | Applied Physics Laboratory (2021), Chapter 4 |
Nuclear detonation by a state actor causing at least 1 fatality | 0.40% | 26.11% | Good Judgement Inc (2018) |
Nuclear detonation as an act of war | 0.92% | 50.24% | Metaculus (2024) |
Nuclear exchange | 0.28% | 19.11% | Metaculus (2024) |
Nuclear war killing at least 1 million people | 0.39% | 25.55% | Global Catastrophic Risks Survey (2008) |
Depending on who you ask — and the exact definition of nuclear war you use — the typical annual risk of some kind of nuclear war is around 0.25% to 2.5% (although I’d guess that estimates >1% per year seem a little too high for the average year). Extrapolating, that suggests a 20% to 80% chance of nuclear war this century.[1]
Ultimately, the exact numbers aren’t particularly important. The takeaway is that experts think that some kind of nuclear attack could very plausibly happen this century — and some even think it’s more likely than not.
How could a nuclear war come about?
When I hear something like “There’s a 20% or 50% chance of a nuclear war this century,” my first reaction is “What?! That’s way too high!”
The first thing to note is that these estimates are generally looking at a nuclear conflict of any size, not just extremely large ones — we’ll look at the chances of nuclear wars that kill 10% or more of the population later in this article.
But even given that, the numbers are surprisingly high. After all:
- There hasn’t been a nuclear strike since 1945.
- Nuclear states are very aware that the use of strategic nuclear weapons would very likely result in nuclear retaliation.
- Since a retaliatory strike could have devastating consequences for a nuclear state, there’s a strong deterrence effect.[2]
- Nuclear weapons are complex and expensive enough to build that non-state actors seem unlikely to get access to nuclear weapons unless they’re intentionally supplied by a state actor.
It’s true that these are reasons to think that nuclear war is unlikely. But unfortunately, there are lots of ways in which nuclear deterrence could break down.
First, deterrence relies on the possibility of a retaliatory strike strong enough to destroy a nuclear state.[3] But there are ways this could fail. For example, a strong missile defence system might mean that one nuclear state could attack a second nuclear state in the knowledge that their defences mean they’ll likely survive any retaliatory strike. Alternatively, states’ military capabilities could be sufficiently imbalanced such that one nuclear state could destroy a second state’s nuclear capabilities before a retaliatory strike is launched.[4]
Second, decision makers could believe that the results of not launching a nuclear strike would be even worse than launching. For example, this could be because individual decision makers’ positions or power are at risk. Or there could be an escalating large-scale conventional war (especially one between nuclear powers) which threatens the existence of a nuclear state. We’ve argued that there’s about a 1 in 3 chance of a war between great powers by 2050.
Third, and perhaps most importantly, the decision-making process could lead to mistakes in a number of ways, for instance:
- Someone involved in the process makes a mistake (the arguments for deterrence assume that states are rational decision makers) — for example, there has been substantial discussion over whether Putin’s decisions to invade Ukraine, and subsequently make nuclear threats show that he is making these kinds of mistakes.
- False information, such as computer errors or mistaking clouds for nuclear missiles, could interfere with the decision making process.
The calculations are hard and uncertain — if Russia used a tactical nuclear weapon in Ukraine, what would NATO’s response be, and how would Putin know?
Indeed, there’s a long list of historical examples of escalations (both inadvertent and intentional), accidental detonations, and false alarms.[5] On January 24, 1961, two nuclear bombs fell out of a plane over North Carolina. Neither exploded, but one fell apart on impact — and five of the six failsafes on the other failed. On September 19, 1980, the non-nuclear part of a nuclear missile exploded in Arkansas after someone conducting maintenance dropped a wrench above a fuel tank, causing a leak (although this didn’t lead to a nuclear detonation).[6]
We also know of a few very close calls when the world really was on the brink of nuclear war. For example:
- As we’ve written about before, on September 26, 1983, Stanislav Petrov refused to report five incoming US missiles that had been detected by the USSR’s early warning system, suspecting a false alarm.
- During the Cuban Missile Crisis in October 1962, Valentin Savitsky, the captain of a nuclear submarine which had been cut off from radio traffic, decided that a war might have already started and wanted to launch a nuclear torpedo. By chance, Vasili Arkiphov, chief of staff of the USSR flotilla heading to Cuba, was on the submarine and refused to authorise the launch.[7]
- On January 25, 1995, Boris Yeltsin decided not to launch a retaliatory strike against the US on detecting a rocket that radar operators thought was a Trident missile. The rocket was actually studying the aurora over Svalbard and Russia had been notified of the launch, but this information hadn’t filtered down to the nuclear operators.[8]
- Also in October 1962, US missile technician John Bordne claims that four missile sites on Okinawa received orders to fire their nuclear missiles. According to Bordne, Captain William Bassett suspected the order was an error and was nearly forced to shoot a subordinate who insisted on following the order. This account is disputed and no official records have ever been released.[9]
What nuclear weapons are we talking about?
There are many ways of categorising nuclear weapons (by fission/fusion, by delivery mechanism, by yield, etc.). One common distinction is between ‘strategic’ and ‘tactical’ (or ‘non-strategic’) weapons. Today, this distinction usually refers to different categories of explosive yield: ‘tactical’ nuclear weapons are generally smaller (up to around 50 kilotons of TNT equivalent), while ‘strategic’ nuclear weapons generally have yields between 100 and 50,000 kilotons of TNT.[10]
Modern strategic nuclear weapons are usually thermonuclear (hydrogen bombs), meaning they have a second stage of detonation fuelled by the fusion of a small element like hydrogen or lithium. Strategic nuclear weapons can be delivered in a variety of ways:
- Land-based missile silos launching ICBMs (intercontinental ballistic missiles), IRBMs (intermediate-range ballistic missiles), or MRBMs (medium-range ballistic missiles)
- SLBMs (submarine-launched ballistic missiles) or SLCMs (sea-launched cruise missiles)
- Aircraft carrying nuclear bombs or ALCMs (air-launched cruise missiles)
- GLCMs (ground-launched cruise missiles)
In general ballistic missiles are powered initially by a rocket, but then follow an unpowered trajectory arching extremely high before falling back down to earth. Cruise missiles are propelled by jet engines, meaning they fly low to the surface of the earth and are harder to detect, but use much more fuel and move at slower speeds.
How bad could a nuclear war be?
Our guess is that a large-scale nuclear war this century could very plausibly kill billions of people, largely as the result of a famine caused by nuclear winter — and the effects of a nuclear war wouldn’t be limited to these fatalities.
It’s also possible (but unlikely) that a nuclear war could cause an existential catastrophe — which we’ll discuss below.
In this section, we’ll look at how bad a nuclear war could be in more detail. We’ll look at:
- How many weapons would be used?
- How would those weapons be targeted?
- What would the effects of each weapon be?
- How will technological developments affect all of this?
It’s important to bear in mind that we don’t really know the answers to these questions. And if we’re looking at any nuclear war happening this century, rather than one in the next decade or so, it’s even harder to predict — a huge amount can change in technology, geopolitics, and strategy over the course of a century. But we can still try to figure out our best guesses and use these to inform our actions.
How many weapons would be used?
The more warheads there are, the more destructive a nuclear war could be — and, as we’ll see, the more likely it is that there could be a nuclear winter that kills billions.
There are currently around 12,000 nuclear warheads, down from approximately 70,000 in 1986, held by nine nuclear weapons states: Russia, the US, China, France, the UK, India, Pakistan, Israel, and North Korea.
The total number of nuclear weapons has clearly been on a broadly decreasing trend. Nuclear weapons are expensive to build and maintain, and if a state already has enough weapons to ensure second-strike capabilities (and other considerations like the variety of delivery systems[11]), it seems like there’s little additional deterrence effect to building more.[12]
Nevertheless, there are some reasons to think that the number of nuclear warheads might grow:
- The number of non-retired nuclear warheads is increasing for the first time since the Cold War (and has been since around 2017). The total number of warheads on Earth (as in the chart above) has been decreasing, but that’s only because there are thousands of retired warheads from the Cold War still awaiting dismantlement.
- China is driving the new buildup. In 2024, China is estimated to have about 500 nuclear warheads, more than double its arsenal in 2020. US Department of Defense analysts estimate that China may be on track to surpass 1,000 warheads by 2030 and 1,500 warheads by 2035 — putting it on a par with the US and Russia’s current arsenals.
- Russia’s invasion of Ukraine and intensifying conflict with NATO makes it seem more likely that Russia will increase its nuclear weapons stockpiles over the coming decades.
- The only remaining treaty limiting nuclear warheads, New START will expire in February 2026. The treaty, originally negotiated by Obama and Medvedev in 2009, caps US and Russian deployed warheads (i.e. warheads currently on bombers or missiles) at 1,550 each but has no provision allowing for extension. A new treaty would require new negotiations between the US and Russia, and two-thirds approval in the US Senate, which doesn’t seem likely — not least because of declining US-Russia relations.[13]
- Prominent nuclear arms experts have been arguing that the US will need to double their number of deployed warheads in response to China to maintain deterrence.[14] If the US goes down this route, it seems likely that Russia will follow. India is also modernising its nuclear arsenal.
- Within the next decade, we may well see Iran obtain a nuclear weapon.[15] In response, we might expect other countries in the Middle East to start nuclear programmes. In the last five years, officials in Saudi Arabia, Egypt, Turkey, Germany, South Korea, Taiwan and Japan have expressed interest in developing nuclear weapons.
Forecasters on Metaculus think there’s a 20% chance that there will be fewer than 50 nuclear weapons on Earth by 2075; but they also think there’s around a 10% chance that there will be more than 500,000.[16]
In the (unlikely) scenario that China continues to grow its arsenal at the same pace that the Soviet Union did between 1950 and 1985, and the US and Russia respond in kind, we could see total weapons stockpiles exceed 100,000 by 2050.[17]
Overall, the total number of weapons deployed this century is highly uncertain. But it’s important to note that the number of weapons and the chances of nuclear war aren’t independent. In fact, it seems like geopolitical tensions both increase the number of warheads being built and are a key risk factor for nuclear war. So we might expect that if there is a nuclear war this century, there’s a decent chance it follows an arms buildup.
How would nuclear weapons be targeted?
Nuclear weapons could be fired at a range of targets: cities, natural resources, government buildings, nuclear silos, conventional militaries, war-supporting infrastructure and industry, and more.
Roughly, the more nuclear weapons fired at densely populated civilian areas, the more devastating the conflict would be — there would be more people killed in the initial detonations and more smoke released (as we’ll see, more smoke means increased chances of a nuclear winter).
The US’s 2022 Nuclear Posture Review says explicitly that the US does not plan to purposely threaten civilian populations.[18] But this doesn’t rule out attacking military targets (like military factories or Russian Ministry of Defence buildings) in populated areas, and it’s not clear how much this would be adhered to in an actual nuclear war anyway. (Listen to Daniel IElsberg discuss his experiences of poor US decision making).[19] We’re not aware of any similar statements from China or Russia.
In general, the specific targeting strategies of nuclear states are highly ambiguous. We expect that some cities and civilian areas would be hit in a small nuclear conflict — and the chances of this would increase in a larger conflict.
What happens when a nuclear weapon is detonated?
There are roughly four main categories of physical effects of nuclear weapons:[20]
- The initial blast
- Ionising radiation and nuclear fallout
- Electromagnetic effects
- Smoke and nuclear winter
Any nuclear detonation will also have social effects, such as norm-shifting and political or military reactions.
Let’s take a look at the effects of detonating a nuclear bomb with a yield of 800 kilotons (kt) of TNT equivalent (the most common yield size from Russian ICMBs)[21] over New York City.[22]
First, there’s an intense burst of gamma rays and neutron radiation. Everyone within ~2.5km of the explosion receives a likely fatal dose of ionising radiation, although many of them will be killed by something else first.
Next, there’s the fireball of superheated air, ground, buildings,and even the remains of the weapon itself. The fireball is many times hotter and brighter than the Sun and everything in it is instantly vaporised.
The fireball throws off intense thermal radiation. Everyone in the line of sight of the detonation, within around 10 km of the explosion, would receive third-degree burns, and buildings and trees would catch fire. You’d have to be over 20 km away to experience only first-degree burns.
Moving away from the fireball slightly more slowly — at only thousands of miles an hour (although slowing as it spreads) — would be the blast wave. Approximately everyone within 2.5 km of the explosion would die, and most residential buildings within 6.5 km would collapse.
Overall, NUKEMAP estimates up to around 2,000,000 fatalities from an 800 kt nuclear weapon exploding in New York, depending on the exact height the bomb explodes.
Very different effects occur if a nuclear weapon is detonated high up in the atmosphere. In particular, such a detonation could produce an electromagnetic pulse (EMP), which could destroy electronics and electrical power systems over an area of around 100 to 1000 km away from the explosion.[23]
Some time after the initial detonation, radioactive isotopes with relatively long half-lives would be distributed by the wind as fallout. If the nuclear explosion happens near the ground, this fallout will be concentrated and local, with lethal effects lasting days or weeks.
People in the orange (100 rads per hour) contour would likely be killed if they’re not in some kind of shelter.
If a nuclear bomb is particularly large, or explodes fairly high into the air, these radioactive isotopes would rise into the stratosphere, eventually distributing fallout across the globe. This fallout can contaminate crops and water supplies. We’d guess this is less harmful than the more concentrated fallout from a closer-to-the-ground, lower-yield explosion. That’s because this ‘global fallout’ has happened before — during nuclear weapons testing and as a result of the Chernobyl disaster. These events likely caused an increase in cancer cases and probably resulted in the early deaths of thousands of people at the very least.[24]
What about nuclear winter?
In terms of total fatalities, everything we’ve discussed above might be just the beginning.
If large numbers of nuclear weapons land in civilian areas, it’s plausible that a high proportion of fatalities would result from nuclear winter — soot from fires could rise into the atmosphere and block sunlight from reaching the Earth causing widespread famine.
Nuclear winter is fairly controversial, and researchers disagree about its effects.
Christian Ruhl put together a list of every study on nuclear winter that looks at how much cooling would result from various nuclear war scenarios — we tried to find studies not in his list but we couldn’t find any.
Ruhl lists nine published papers. Six of these nine papers are all authored by at least one of Alan Robock and Brian Toon, all of which are broadly pessimistic about the effects of nuclear winter. Meanwhile, one paper by Reisner et al. comes to a much more optimistic conclusion.
Reisner et al. consider a scenario where 100 15 kt warheads are detonated in an India-Pakistan exchange — the same model scenario considered by Robock and Toon in their first 2007 paper, their 2014 paper, and their 2022 paper.
Robock and Toon’s models show a drop in temperatures of around 1 to 2ºC, lasting 3 to 10 years depending on the model, whereas Reisner et al. find a drop in temperatures of 0 to 0.5ºC, with effects mostly limited to polar regions.
Robock et al. responded to Reisner’s paper, and Resiner et al. responded to that. (See this blog post for a summary of this debate that’s ultimately critical of Robock and Toon.) We think both groups raise plausible criticisms of the other’s models.
The two groups’ funders and backgrounds may be playing a role in their different conclusions. Robock and Toon are often criticised by detractors, not least for their belief that fear of nuclear winter keeps the world safe.[25] On the other hand, Reisner et al. was carried out by the Los Alamos National Lab (founded as the secret site for developing the first atomic bomb) and funded by the US Department of Energy and the US Department of Defence. We’d guess that these organisations have an overall incentive to claim that the risk of nuclear winter is small, as nuclear winter seems primarily to be used as an argument by disarmament activists, and against the US’s plan to win any nuclear war. That said, there are reasons for them to exaggerate the risk of nuclear winter as well, such as improving deterrence.
There are two other papers on Ruhl’s list: One is the 1982 paper which first introduced the concept of nuclear winter, but it doesn’t do any detailed modelling. And there’s a 2016 paper by Pausata et al., which broadly agrees with Robock and Toon. You might think that Pausata et al. provides a neutral third result, but that research was funded by Swedish Physicians against Nuclear Weapons, so it may be biassed for similar reasons.
We’re left pretty unsure about what to expect here — and that’s not hugely surprising given the limited research in the area, and that we’re trying to predict the consequences of a large-scale nuclear war, something we have very little empirical evidence about.
Ultimately, we don’t really know:
- How much soot would be generated by a nuclear attack on a city
- How much soot would get into the stratosphere
- How long that soot would remain in the stratosphere
- How much light that would block
- How much that would affect agriculture and food supplies
- Whether society would be able to respond to that effectively
Overall, we’d guess the right answer is somewhere between the predictions of the Los Alamos team and those of Robock and Toon — and we’re open to either being right. That is, we’d guess that a war including the explosions of 100 15 kt warheads would cause some cooling in a range from negligible (approximately 0ºC) to -2ºC. And a larger nuclear war would have larger cooling effects.
Luisa Rodriguez summarised studies that looked at the relationship between smoke and famine. From looking at these studies, it seems like an amount of smoke that could result in cooling of over 1ºC would plausibly kill around one billion people, and that cooling of more like 4ºC would kill around eight billion people — although these studies are all highly uncertain (and many are again by Robock and Toon).
Putting all this together, we’d guess that up to around two billion people might die as the result of a famine following the explosions of 100 15 kt warheads over cities. Our median guess would be much lower than this upper bound, although a real nuclear war could involve much more than just 100 15 kt warheads.[26]
We’ll look later at whether a nuclear winter could cause human extinction.
How could this all be changed by new technology?
Of course, technology is unlikely to be the same by 2100 as it is today.
Some technological advances we might see:[27]
- Development of hypersonic missiles and increased use of cruise missiles. Cruise missiles get to their target in less time once they’ve been detected, and their targets and payloads are more ambiguous — hypersonic missiles have similar problems. This makes decision making under a possible attack harder and increases risk.
- Changing cybersecurity landscape. Cyber attacks on nuclear weapons systems are highly plausible, and many things — including the development of AI — could change this landscape. It’s currently unclear whether these developments will improve states’ offensive capabilities more than their defensive ones.
- Improvements in remote sensing and weapons accuracy could make it more likely that nuclear forces can be destroyed by conventional weapons or much lower-yield nuclear weapons on a first strike, which could harm deterrence.
- Missile defence systems could improve, preventing one state’s ability to conduct a first strike without substantially increasing their arsenals. This might also mean that any retaliatory strike needs to be launched in a very short period of time. It could also have an effect on the incentives for limited nuclear war by making it unclear how many nuclear weapons are needed to achieve a given result.
- Improved early warning systems could increase the time and information available to states under attack, decreasing the chances of mistakes and inadvertent escalation.
- New kinds of weapons, such as salted bombs (designed to maximise radiation poisoning as a result of their explosion) or pure fusion bombs, could substantially increase the damage caused by a single strike. That said, one expert we spoke to said that over the last 50 years, nuclear weapons have become on average less damaging — nuclear thinkers have become more interested in precision than yield.
- Autonomous weapons and command and control systems could substantially change the landscape, although we’re not sure how.
Ultimately, new technology can be either stabilising or destabilising — this adds another layer of uncertainty to our previous guesses about fatalities. (That said, developing stabilising technology could also be a way to help reduce the risk.)
Other effects of a nuclear war
The effects of a nuclear war wouldn’t be limited to fatalities.
In particular, we might see shifted norms:
- We could see an end to the nuclear taboo — the current stigma that sees nuclear weapons as distinct from conventional weapons, especially that they’re morally worse to use. Alternatively, we could see stronger pushes for disarmament or other moves against nuclear weapons.
- We could see increased or decreased focus on preventing other potential catastrophic risks, like pandemics or artificial intelligence.
- We could see shifts in the global balance of power, which in turn lead to shifts in other norms or moral positions. The world would look like a very different place if, for example, non-autocratic states are substantially weakened (or relatively strengthened). Continuity of government plans tend to be less democratic than our current systems.
- Some authors have claimed that disasters will lead to a general loss of law and order, and an increase in violence.[28]
We’d also see damage to infrastructure and disruptions to water, food, and energy supplies. We’d see disruptions to healthcare and transportation, alongside broader economic changes, such as changes to labour supply. Baum et al. provide a long (but still non-comprehensive) breakdown of these effects, although this is a highly uncertain area overall.
Could a nuclear war cause an existential catastrophe?
We’ve argued that preventing an existential catastrophe — an event that causes human extinction or permanently and drastically curtails humanity’s potential — could be of particular moral importance. We’ll look at three ways nuclear weapons might cause an existential catastrophe based on the most dangerous aspects of nuclear weapons discussed so far:
- Human extinction resulting directly from nuclear detonations
- Human extinction resulting from nuclear winter
- Knock-on effects that lead to an existential catastrophe some other way
Finally, we’ll try to give a rough overall answer on how likely nuclear war is to cause an existential catastrophe.
In short, we don’t think there’s much chance that a nuclear war could cause humanity to go extinct or have a predictably negative impact on humanity’s ultimate long-run potential. (With such a bleak subject, that at least seems like a bit of good news.)
Here’s why.
Are there enough nuclear weapons to kill everyone on earth without nuclear winter?
Right now, no.
Let’s do a rough estimate.
We saw earlier that approximately everyone within 2.5 km of the explosion of an 800 kt nuclear weapon would die. That’s an area of around 20 square kilometres. By one estimate, roughly 3.5 million square kilometres of the world are urban areas, so you’d need around 175,000 800 kt nuclear weapons to destroy every urban area (and that’s ignoring the fact that circles don’t perfectly tessellate). That’s a total explosive power of around 140,000 megatonnes (Mt).
Maybe there’s a way to be more efficient – smaller nuclear weapons destroy more area for the same kt of TNT equivalent. Let’s consider the smallest bomb in the current US nuclear arsenal, the B61 nuclear bomb with a yield of 300 tonnes. With those, you could destroy all urban areas with only around 16,000 Mt (although you’d need 52 million bombs).[29]
Here’s a graph of explosive power in the world’s nuclear arsenals over time:
The maximum explosive power we’ve ever had is around 15,800 Mt. That is to say, even if we were to use perfectly-targeted and impractically small nuclear weapons, there have never been enough nuclear weapons to kill everyone in the world’s urban areas.
And that’s before you consider that only 55% of the world’s population live in urban areas.
It’s possible that we could, at some point in the future, end up with an arms race worse than the Cold War. But even then it seems very unlikely that the initial nuclear explosions could kill everyone on earth.
What about radiation?
Our guess is that radiation and fallout are less likely to be able to kill everyone than just the force of the nuclear explosions. As we saw earlier, the area in which fallout is fatal immediately is relatively small (and heavily influenced by factors like the wind), making it very likely that many areas, especially in the southern hemisphere, will receive basically no local fallout. Similarly, we saw that while previous explosions (like Chernobyl and all previous nuclear tests) have caused some global fallout, that’s far from the quantity required to kill everyone. This could change with the development of salted bombs. (Read more about whether radiation could kill everyone.)
Could nuclear winter cause human extinction?
The much more plausible cause of human extinction is nuclear winter.
As we saw above, the extent to which nuclear winter would result in cooling is hugely uncertain — and Robock and Toon’s model is likely an overestimate of the extent of cooling. We’ve also discussed uncertainties about how many nuclear weapons would be launched in a nuclear war and how those weapons would be targeted.
We’re also not sure what the minimum viable population for humans could be, but research by Luisa Rodriguez suggests that even if 99.9% of the population is killed, humanity is fairly likely to survive.
What’s more, nuclear winter models tend to suggest that humans in much of the southern hemisphere would survive even in a very large nuclear war. One paper looks specifically at New Zealand during nuclear winter and finds that food production would likely exceed the required calorie level to survive.
That said, if there’s an arms race followed by a war where many more nuclear weapons than we currently have are fired, the risk of extinction would be higher — especially if the war took place in the southern hemisphere.
But overall human extinction from nuclear winter seems highly unlikely.
It seems like even Robock and Toon wouldn’t disagree with that conclusion.
Luke Oman, a co-author of Robock (2007), said:
“The probability I would estimate for the global human population of zero resulting from the 150 Tg of black carbon scenario in our 2007 paper would be in the range of 1 in 10,000 to 1 in 100,000… I don’t know offhand anyone that would estimate higher but I am sure there might be people who would. [I asked two colleagues] who did respond back to me, saying in general terms ‘very close to 0’ and ‘very low probability.'”
Plausibly, billions of people could be killed as the result of nuclear winter. Any event that deadly would be the worst catastrophe in human history, and we should do much more to avoid and understand the risks. But humanity, as a species, would be very likely to survive.
Could there be knock-on effects that lead to an existential catastrophe some other way?
As we saw above, the effects of a nuclear conflict on society are diverse and difficult to predict — and certainly not limited to fatalities.
But ultimately, we know very little. After all, we’ve only ever had one war which included a nuclear explosion.
One way in which a nuclear war could be an existential risk is if the chances of an existential catastrophe are higher in a post-nuclear-war world than currently. For example, perhaps society after a nuclear war would be more susceptible to a catastrophic pandemic, or perhaps powerful weapons could end up in the hands of dangerous warlords. But there are also reasons to think that a nuclear war could decrease the chances of another catastrophe: perhaps society would learn and be more careful in future, and perhaps a decreased rate of technological progress would make it easier to deal with new threats.
We have so little clear evidence about this question, which means it may well come down to one of priors: is our current society unusually likely to suffer an existential catastrophe? We’d guess the answer to that is yes — but mainly because of the technology we’ve developed, and we’d guess that a society following a major nuclear catastrophe (e.g. a safe refuge in the southern hemisphere) would retain much of our current technological progress. Is our current society unusually likely to suffer an existential catastrophe compared to a similarly technologically developed society? We don’t know, which makes the overall effects of a nuclear war on other existential risks unclear to us.
Another argument is that nuclear war might affect our values and institutions. For example, we might see the destruction of the US and other NATO countries and a world with Russian or Chinese hegemony — and while we of course have no way of being unbiased here, we overall think the world is better off with more rather than less influence of liberal democratic values and institutions. .
If there’s reason to think that our near future values might soon be ‘locked-in,’ for example through the development of transformative AI, changes to our values would have a permanent negative impact on humanity’s ultimate long-run potential, which would constitute an existential catastrophe. For example, we could end up with a situation like stable totalitarianism that locks in suboptimal post-nuclear war values. However, actually predicting the direction of any value changes following a nuclear war seems like a very difficult task — and the chances of ultra long-term lock-in, even from developments like transformative AI, seem very low to me.
Overall, I’d guess that these knock-on effects aren’t any more likely to lead to an existential catastrophe than the direct effects of a nuclear war.
How likely is an existential catastrophe resulting from nuclear war?
There have been various estimates of the chances of existential catastrophe from nuclear war. Existential catastrophe — either killing everyone or otherwise permanently damaging humanity’s potential (e.g. through value lock-in) is an extremely high bar, so the probability of this is probably quite low.
Here’s a table of every estimate we could find:
Definition | Probability | Implied probability from 2024 to 2100 | Source |
---|---|---|---|
Existential catastrophe by 2120 | 0.10% | 0.08% | Toby Ord, The Precipice, 2020 |
Population < 5,000 by 2100 | 0.07% | 0.07% | XPT, superforecasters, median |
Population < 5,000 by 2100 | 0.55% | 0.55% | XPT, domain experts, median |
Population < 5,000 by 2100 | 0.03% | 0.03% | XPT, superforecasters, geometric mean of odds |
Population < 5,000 by 2100 | 0.08% | 0.08% | XPT, domain experts, geometric mean of odds |
Human extinction, annual | 5.93E-12 | 4.57E-10 | Vasco Grilo (2024) |
Human extinction by 2100 | 1.00% | 0.82% | GCR Conference, 2008 |
Existential catastrophe, annual | 0.30% | 20.04% | David Denkenberger, 2018 |
Existential catastrophe, annual | 0.10% | 3.85% | Anders Sandberg, 2018 |
Extinction by 2100 | 1% | 0.82% | Turchin (2008) |
Infinite impact" within the next 100 years | 0.01% | 0.00% | Pamlin, D. & Armstrong, S. (2015) |
Note, many of these sources estimate extinction risk, which don’t include the chances of a non-extinction existential catastrophe (e.g. the sorts of value lock-in considered above).
I think these sources tend to overestimate the risk of extinction by overestimating the risk posed by nuclear winter. Of these estimates, my sense is that the Existential Persuasion Tournament (XPT) estimates are the most informative: they aggregate estimates from a large number of people — but even among candidates in the XPT, I’d tend to agree with those who are more sceptical about nuclear winter.[30]
Overall we’d guess that the existential risk from nuclear war is around 1 in 10,000.
What about catastrophes that aren’t existential?
Nuclear war seems shockingly likely to kill a huge proportion of the population.
As we saw above, the chance of a nuclear war by 2100 seems quite high — many estimates put it at 20–50%.
Some of the sources that tried to estimate the chance of extinction also considered the chances that a nuclear war would kill more than 10% of humans currently alive:
These estimates — a 1 in 10 chance of over 10% of the population being killed as the result of a nuclear catastrophe by 2100 — are similar to our best guess.
That means there’s something like a 1% chance you will die in a nuclear war (although that’ll vary depending on how old you are and where you live). You’re far more likely to be killed in a nuclear war than you are to die in any kind of natural disaster. It’s similar to the chances of being murdered or dying as the result of alcohol or drugs.
And alongside all those deaths, such a war would cause unimaginable suffering.
As a result, we think that — even if we set aside the chances of an existential catastrophe — nuclear war is still one of the most pressing problems facing the world today.
Nuclear risk is increasingly neglected
You’d think that reducing the risk posed by nuclear weapons would be a major focus for humanity — after all, the chances we face a nuclear conflict aren’t low, and everyone’s heard of the issue.
But since the end of the Cold War, attention to this area has substantially decreased, especially attention focused on preventing the risk of nuclear conflict between existing nuclear powers (which is the most likely cause of the sorts of large-scale destruction we discussed above).
We’ll mainly look at funding and focus on two main sources: governments and philanthropic organisations. We think the private sector isn’t a major source of funding for nuclear risk reduction.
Philanthropic spending
In 2020, the MacArthur Foundation, the biggest philanthropic funder of nuclear security, stopped funding the field, with its final funds being spent in 2023.
According to a Founder’s Pledge report, annual philanthropic funding fell from around $50 million a year to around $35 million.
For comparison, Oppenheimer’s budget (the film that is, not the Manhattan project) was around $100 million. $30 million is also less than philanthropic spending on our other top problems, like the at least $50 million philanthropic spending on preventing an AI-related catastrophe.
Government spending
All that said, nuclear risk receives lots of attention and funding from governments.
For example, the National Nuclear Security Administration (NNSA) in the US had a budget of around $25 billion. Other parts of the US government also focus on nuclear-related risks, like the Bureau of Arms Control, Verification and Compliance (with a budget of around $8 million) and the Defense Threat Reduction Agency (with a budget of around $1 billion).
The finances of China and Russia are much more opaque. We’d guess they spend less than the US, but we wouldn’t be surprised if it wasn’t that much less.
We’re not sure what the effect is of government spending. The vast majority of the NNSA’s budget ($19.8 billion) is going towards weapons modernisation, whose effect on risk is unclear — although around $2.5 billion is spent on nonproliferation, emergency response, and counterterrorism programs.
You might think it’d be reasonable to expect that it’s in no one’s interest to have a nuclear war, so this spending would on average help reduce risk. Unfortunately, we think that there are consistent problems with the way the US budget is allocated. For example, we spoke to Jeffrey Lewis — a political scientist with a background in arms control — who told us that US military spending is affected far more by the competition between the Navy, Army, and Air Force than broader strategic considerations.
So it doesn’t seem like this US spending is primarily aimed at reducing risk and, as a result, we don’t think it makes the problem non-neglected.
Many governments also fund important international organisations, such as the Comprehensive Nuclear-Test-Ban Treaty Organization (with a budget of around $125 million). The IAEA budget for Nuclear Safety, Security and Verification is around €200 million, although much of that will be spent on safety related to nuclear power. The United Nations Office of Disarmament Affairs has a budget of around $10 million a year, split across nuclear, biological, chemical, and conventional weapons.
Overall, we’d guess that governments are spending something like $10 billion on reducing nuclear risk. There’s a major focus on counter-terrorism and non-proliferation (which makes sense as nuclear states can all agree they don’t want other states or groups getting nuclear weapons) and little focus on preventing the sorts of large-scale nuclear catastrophe we’ve discussed in this article. Let’s say this amounts to something like $1 billion (quality-adjusted). We’re extremely uncertain, but $1 billion sounds reasonable to us and puts it in a similar ballpark to our guess of quality-adjusted pandemic prevention spending.
How many people are working on this problem?
We’re not sure.
Crudely dividing our $1 billion government-spending estimate by a salary of $50,000 gives around 20,000 total people working on the area overall.
Since philanthropic organisations work on slightly different (and plausibly more cost-effective) areas, it’s worth looking specifically at how many people are working on this in non-governmental organisations.
Three of the largest non-governmental organisations working on the area — the Arms Control Association, Bulletin of the Atomic Scientists, and Ploughshares Fund have around 70 total staff. We’d guess that around 30 people in the effective altruism community are also working on this issue.[31]
How neglected is this issue overall?
These estimates show a mixed picture.
There’s a large amount of government attention — similar to the amount given to preventing catastrophic pandemics, although much less than the hundreds of billions spent on climate change.
But this spending isn’t particularly well-targeted. What’s more, around 40% of the NNSA’s employees are eligible for retirement,[32] and many in the field warn about the lack of new, younger people entering the area. So there may be more opportunities for talented people to enter the area and have an impact than the total spending implies.
And non-governmental spending and personnel on this issue are shockingly small.
Overall, we think that reducing the risk of a large-scale nuclear catastrophe is very neglected — probably less neglected than reducing risks from AI, but more neglected than many other areas that many prioritise, such as global health or climate change.
There are ways we can tackle nuclear risk
Ultimately, decisions around the deployment and use of nuclear weapons are in the hands of the nuclear-armed states: the US, the UK, France, Russia, China, India, Pakistan, Israel, and North Korea.
Most plausible paths to reducing nuclear risk involve changing the actions of these countries and their allies.
We’ll focus on the US (and NATO countries) here because those are the countries most of our readers are well-placed to work in, but we’d expect many of these policies to be useful across the world. (We’ve written elsewhere about working on policy in an emerging power.)
So, which actions would be most beneficial to pursue?
Overall, after talking to experts in the area, we think there’s substantial value in attempting to:
- Reduce the risk of accident or miscalculation. A significant portion of nuclear risk comes from accidental use of nuclear weapons — especially during a crisis or a period of heightened tensions. There have been several close calls over the past 80 years. These risks could be mitigated by improved technology and procedures, for example by improving the reliability of early warning systems.
- Prevent deployment of the most escalatory weapons. For example, in early 2023, the US decided not to develop a new nuclear-armed sea-launched cruise missile (SLCM-N). SLCM-Ns are considered destabilising because of their target and payload ambiguity, so this decision may have been a success for risk reduction.[33] Similarly, there may be policies that could disincentivize Russia and China from deploying some of the destabilising weapons they are currently pursuing.
- Introduce checks and balances on the use of nuclear weapons. In the US, the President has the sole authority to launch nuclear weapons. Having this authority in the hands of just one person increases the chances that the US will launch nuclear weapons when it would not be in the interests of the world to do so. There are several proposed pieces of legislation which would limit presidential authority, as could an executive order.
- Anticipate and limit potentially dangerous technological developments. We could see new risks arising from the integration of artificial intelligence into nuclear command and control or the development and use of space-based missile defences — although we’re uncertain about whether these technologies would be good or bad on net. Identifying and then prohibiting any dangerous technologies before they are deployed — even if that’s done unilaterally (as proposed in a recent bipartisan US senate bill) — could lower the risk of catastrophe. And it might lead to reciprocal commitments by other nuclear powers, including Russia and China.
- Achieve resilient deterrence while avoiding a new arms race. We’re seeing some countries (in particular China) increase their nuclear arsenals. Ideally, a US / NATO response to this would maintain both nuclear and non-nuclear capabilities that, when combined, make it clear to adversaries they will gain no advantage by striking first or by continuing further arms buildups.
- Promote arms control dialogue between the US, China, and Russia. Unfortunately, there is virtually no chance for extension of New START, the existing nuclear arms reduction treaty between the US and Russia which is set to expire on February 5, 2026. But similar dialogue in the future could lead to substantial breakthroughs. Also, other dialogue measures can help improve relations, especially during periods of high tension, for example: codes of conduct, hotlines, transparency measures, and information exchanges.
- Strengthen international norms and laws against nuclear possession and use. For example, nuclear states could ensure their nuclear targeting is consistent with the Law of Armed Conflict. It would also be great to end up in a situation where nuclear states recognise arms control as a crucial means to achieving their own military goals.
- Develop a better understanding of the possible pathways to inadvertent escalation, especially among senior decision makers.
Another option is to work on this problem indirectly by reducing the risk of great power conflict.
We’ll look below at what you can do to help achieve these goals.
What are the major arguments against this problem being pressing?
Here are some of the most compelling reasons we haven’t yet discussed against working in this areas:
- There’s a real risk of accidentally causing harm. For example, discouraging countries from maintaining an aggressive posture in which they say they will use nuclear weapons if attacked could make the world less safe by encouraging conventional conflict.
- You might think that because most people want to avoid a nuclear war, the area would become less neglected if there was a clear arms race or increase in risk.. We think that’s plausible — although we’d note that while there’s been an increase in risk in the last few years, funding in the area has actually decreased.
- You might think our estimate of the existential risk from nuclear war is too high. For arguments along these lines, see this post by Vasco Grilo and this post by Jeffrey Ladish.
- You might think that reducing the risk of existential catastrophe is even more important than we do. Ultimately, we think that the impacts of nuclear war on current society are a comparable reason to prioritise nuclear war (alongside the approximately 1 in 10,000 existential risk). If you disagreed with that, you’d probably want to work on an area that’s a more plausible existential risk.
What can you do to help?
We think there are four main ways you can help. You could work in:
- Governments and international organisations to help them to reduce risk
- Research to either answer currently unanswered questions on nuclear risk or to build a platform for advocacy
- Communications and advocacy to push for changes and policies that are needed
- Field-building, for example earning to give to increase the amount of funding available, or helping get more young people into the field
We discuss the options in more detail here:
Nuclear career reviewYou might also consider donating to the Founders Pledge Global Catastrophic Risks Fund.
What are the key organisations in this space
In general, we’d recommend working in government organisations, media companies, political parties, and international organisations.
Within the US government in particular, we’d suggest you consider working in:
- Congress
- The Department of Defense
- The Department of Energy (including the National Nuclear Security Administration and the National Labs)
- The State Department
Here are a few organisations specifically working on reducing nuclear risk:
- The Nuclear Threat Initiative is a US nonpartisan think tank that works to prevent catastrophic attacks and accidents with nuclear, biological, radiological, chemical, and cyberweapons of mass destruction and disruption. NTI also sponsors the William J. Perry Project, an initiative from former US Secretary of Defense William Perry to educate the public about the continuing threat of nuclear weapons. See current vacancies.
- The Carnegie Endowment for International Peace is also a US nonpartisan think tank focusing on peace and international cooperation. One of its 10 programs focuses on nuclear policy.
- Longview Philanthropy is an organisation that recommends grants focused on improving the lives of future generations. They have a nuclear security grantmaking programme that aims to grant $10 million to opportunities focused on reducing existential risks posed by nuclear weapons. See current vacancies.
- Founders Pledge is a nonprofit that encourages startup founders and investors to make a legally binding pledge to donate a portion of their personal proceeds from exit to charity. They have a team working on grantmaking in nuclear security.
- Ploughshares Fund is the largest US philanthropic foundation focused exclusively on peace and security grantmaking. It supports initiatives to reduce current nuclear arsenals and to limit the likelihood of nuclear war (and to a lesser extent, risks from chemical and biological weapons). See current vacancies.
- The Future of Life Institute works to reduce the risks from a number of areas, in particular nuclear war, pandemics, and advanced AI. See current vacancies.
- Global Catastrophic Risk Institute is an independent research institute that investigates how to minimise the risks of large-scale catastrophes, such as from nuclear war. See current vacancies.
- Alliance to Feed the Earth in Disasters (ALLFED) researches and develops alternative foods that could be quickly scaled up to feed everyone in a nuclear winter and coordinates preparedness and planning. Contact ALLFED to volunteer or apply to an open position.
- The Berkeley Risk and Security Lab does academic research at the intersection of nuclear risk and emerging technology.
Learn more
80,000 Hours podcasts relating to nuclear war:
- Andy Weber on rendering bioweapons obsolete and ending the new nuclear arms race
- Joan Rohlfing on how to avoid catastrophic nuclear blunders
- Daniel Ellsberg on the creation of nuclear doomsday machines, the institutional insanity that maintains them, and a practical plan for dismantling them
- Annie Jacobsen on what would happen if North Korea launched a nuclear weapon at the US
- Samuel Charap on the 2022 Russian invasion of Ukraine and the effects of this on nuclear security (published March 2022)
- Luisa Rodriguez on why global catastrophes seem unlikely to kill us all
- David Denkenberger on using paper mills and seaweed to feed everyone in a catastrophe, ft Sahil Shah
- Lewis Dartnell on getting humanity to bounce back faster in a post-apocalyptic world
- Christian Ruhl on why we’re entering a new nuclear age — and how to reduce the risks
Other content on nuclear security:
- Luisa Rodriguez’s research on risks from nuclear weapons
- Global catastrophic nuclear risk: A guide for philanthropists and Philanthropy to the right of boom, both by Christian Ruhl
- Open Philanthropy on nuclear weapons policy
- Chapter 18: “The continuing threat of nuclear war” by Joseph Cirincione in Global Catastrophic Risks, edited by Nick Bostrom and Milan M. Cirkovic
- How the Global Catastrophic Risks Institute is tackling the problem
- ALLFED publications on food system resilience
- Podcast: Samantha Pitts-Kiefer: “It’s my job to worry about any way nukes could get used”
- Podcast: The Logic of Doomsday with William J. Perry and Sam Harris
- Podcast series: At the Brink with William J. Perry
Other related content:
- ^
The probabilities used in this table were estimating the chances of a nuclear war-like event over a variety of time periods:
- 2022 to 2030 (XPT)
- 2004 to 2014 (Lugar)
- Annualised “current” probability in 2021 (Applied Physics Laboratory)
- 2018 to 2021 (Good Judgment Inc)
- 2024 to 2050 (Metaculus — nuclear detonation)
- 2024 to 2070 (Metaculus — nuclear exchange)
- 2008 to 2100 (Global Catastrophic Risks Survey)
To create the probabilities in the table, we assumed that the annual probability of nuclear war is independent. This is a decent but imperfect assumption, and may lead to inflated probabilities for predictions originally made for shorter time periods.
- ^
When discussing nuclear retaliation, people often talk about the doctrine of mutually assured destruction — the claim that the large-scale use of nuclear weapons would result in the complete destruction of all parties involved in the nuclear exchange. But when we spoke to Jeffrey Lewis, professor at the James Martin Center for Nonproliferation Studies, he argued that the US’s current strategy is in fact to attempt to win a nuclear war, not to assure mutual destruction.
But it’s still the case that one significant impact of having nuclear weapons is that a nuclear weapons state can deter attacks (including nuclear attacks) through the threat of retaliation. As a result, we’ve decided to talk about deterrence in this article, rather than use the term mutually assured destruction.
- ^
There are possible versions of deterrence that focus on states’ leadership: these don’t entail destroying all of a state’s productive capacity, though they do end the current government of the state. But, in practice, we think a large portion of nuclear deterrence relies on retaliatory nuclear strikes, and that appears to be the primary strategy employed by existing nuclear states.
- ^
These capabilities don’t necessarily have to be nuclear. As a hypothetical example, the US plans to buy 100 B-21 stealth bombers, which could look as small as an insect on radar detection systems. The US is also building the stand-in attack weapon, a small hypersonic missile which could potentially destroy hardened missile launchers. If each B-21 could carry 20 stand-in attack weapons, that would suggest the US could destroy around 2,000 targets using around $6 billion worth of conventional weapons.
- ^
One expert we spoke to thinks that accidental nuclear weapon use is becoming more likely due to the entanglement problem: the mixing of conventional and nuclear tools and military forces. For example, radars, satellites and communications systems are used to support both conventional and nuclear operations, and could be subject to non-nuclear or cyberattacks. But this could damage nuclear early warning systems, which could be interpreted as preparations for a nuclear war. That said, we’d guess there are also reasons to think it might be becoming less likely (for example, technological improvements to the accuracy of early warning systems).
- ^
For a list of similar such events, see section 4 of Baum, Seth et al., A Model For The Probability Of Nuclear War, Global Catastrophic Risk Institute Working Paper 18-1.
- ^
There are a few details and caveats that seem worth noting about the Arkhipov incident:
- The crew of the Soviet submarine may have believed they were under attack at more than one point in time. They surfaced due to US signalling depth charges that may have been misperceived as real depth charges. Upon surfacing, they claim to have faced blinding searchlights and numerous warning shots.
- The submarine was designed for much colder temperatures than the tropics. Going to Cuba, the crew faced temperatures from 50 to 65°C and high levels of carbon dioxide.
- Savitsky likely had an incentive to act more aggressively in the presence of the Chief of Staff, knowing he didn’t have the actual decision-making power.
- Use of the nuclear torpedo would have been directly suicidal for the crew given their range to the US ships. This likely helped create hesitation. It could also have provided an incentive for Savitsky — signalling his patriotic selflessness.
Thanks to Matthew Gentzel for these points.
- ^
One expert we spoke to said this incident was extremely unlikely to have escalated into a war. It was peacetime with too little force and the wrong trajectory for a full-scale attack. The perceived threat was really an EMP attack (which we discuss in the section on what happens when a nuclear bomb is detonated), and Russia was, as a result, likely waiting for detonation to confirm it was such an attack.
- ^
It’s important to note that this list only contains declassified examples of close calls. It’s plausible to us that the worst examples of close calls are more likely to remain classified.
- ^
The precise distinctions between ‘strategic’ and ‘tactical’ nuclear weapons are somewhat arbitrary. Beyond yield, these words also refer to the intended use of these weapons, with ‘tactical’ weapons intended for battlefield use and ‘strategic’ weapons intended for use against cities, industry, and infrastructure. They can also be used to refer to the delivery system, where a ‘strategic’ system can be delivered over long distances while a ‘tactical’ one has a more limited range. But in principle, weapons with ‘strategic’ yield can be used for ‘tactical’ purposes and vice versa.
The Origins of Limited Nuclear War Theory, chapter 2:
“The distinction between tactical and strategic, whether regarded as legitimate or not, could not be but highly context dependent… The distinction between ‘tactical’ and ‘strategic’ … was notoriously imprecise.”
- ^
We’ve been told that many people in the US nuclear community have various arguments for why it can still be beneficial to build more weapons when you already have a second-strike capability. Some of these include:
- A greater variety of delivery systems would allow for a better tailoring of nuclear use to different scenarios (e.g. a new system might allow you to target a moving ship that you couldn’t before, or a new system might make it so that you don’t have to shoot a nuclear missile across a neutral country’s territory to target something in the adversary country). Adding this new system might take you above the warhead count that is absolutely necessary for “minimum deterrence” but still provide a benefit to nuclear deterrence and nuclear risk that some people think is valuable.
- Some people think that it improves deterrence for a country to be able to significantly limit the damage that another country could impose on it through a nuclear attack. This might require a large number of nuclear weapons to be able to target a significant portion of the other country’s nuclear forces. A classic essay arguing for this is Victory is Possible by Colin Gray and Keith Payne. One expert we spoke to told us this way of thinking is very influential in DC.
- Other people think that warhead count and total summed yield meaningfully affects the ability of a country to achieve its goals in the middle of a crisis with another nuclear-armed country, even beyond the concrete capability benefits that these weapons could actually provide.
We don’t have clear views on the validity of these arguments.
- ^
We attempted to look into how many weapons are required for minimal deterrence. We found claims that hundreds of warheads is approximately right for minimal deterrence, but we’re not sure where this number comes from.
Our guess is that the answer here is situation dependent: if the US had more nuclear weapons, these could more effectively wipe out the Russian nuclear arsenal, making a retaliatory strike less likely — and as a result Russia would need more nuclear weapons to retain their second strike capabilities. Technological developments like improved missiles or better missile defence systems could also change this balance.
Also, it’s clear that during the Cold War, there were very large numbers of nuclear weapons, plausibly more than required for minimal deterrence. So the concept of minimal deterrence might be missing large parts of the picture (for example, irrational decision making or perverse incentives in governments causing more weapons to be built).
- ^
If New START collapses, the number of deployed nuclear warheads in the US and Russia could approximately double in size in the course of a few months by adding additional existing warheads onto additional existing missiles, according to the Federation of American Scientists.
- ^
Franklin Miller, a highly regarded nuclear defence policy expert, argued in the Wall Street Journal:
“If the U.S.-Russian arms-control dialogue survives Mr. Putin’s invasion of Ukraine—a big if—and assuming Mr. Putin doesn’t detonate a nuclear weapon, the administration could propose a new U.S.-Russian treaty with a ceiling of 3,000 to 3,500 total nuclear weapons for each side. This would limit the threats to our allies and homeland and also permit a U.S. strategic nuclear capability that would deter both Russia and China. (Including China in a trilateral nuclear arms-control accord is unrealistic. China has rejected participation in such talks as well as the transparency and verification vital to a successful treaty.)
If a new arms-control dialogue is politically unacceptable, the Biden administration should exit New Start after a year and begin building toward the 3,000 to 3,500 force levels to maintain a credible deterrent against Moscow and Beijing. Many members of the Western arms-control community would complain of a “new arms race.”” - ^
Metaculus, a forecasting website, asks, “Will Iran possess a nuclear weapon before 2030?”
“This question will resolve as Yes if the Iranian Government credibly states that it has a nuclear weapon or has tested a nuclear weapon at any time between January 1, 2020 to January 1, 2030, according to credible media reports.”
With 962 predictions, the community prediction is at 51% (as of June 2024).
- ^
We think you should be sceptical about these exact numbers — making predictions is harder the further into the future the prediction is made and only 19 forecasters have put down a prediction as of May 2024. But we think this is a good illustration of the range of possible outcomes.
- ^
For more, see the section “What went down can come back up” in Founder’s Pledge’s report on nuclear risk and philanthropy.
- ^
The 2022 Nuclear Posture Review, while discussing what the US’s strategy would be if deterrence fails, says:
“We will maintain a safe, secure, and effective nuclear deterrent and flexible nuclear capabilities to achieve our objectives should the President conclude that the employment of nuclear weapons is necessary… Longstanding DoD policy is to comply with LOAC in all armed conflicts, however characterized, and the DoD Law of War Manual recognizes that “[t]he law of war governs the use of nuclear weapons, just as it governs the use of conventional weapons.” In addition, longstanding U.S. policy is to not purposely threaten civilian populations or objects, and the United States will not intentionally target civilian populations or objects in violation of LOAC.”
- ^
The US has many relatively low-yield and accurately-targetable nuclear weapons (rather high-yield ones), which suggests that they primarily plan to attack specific military and strategic targets — and the same is true for Russian nuclear forces. This makes sense, because one key objective of these nuclear arsenals is likely to survive a first strike and/or to prevent a retaliatory strike. (That said, each of these low-yield weapons are still hundreds of TNT equivalent, so if they went off in a city it would be worse than Hiroshima or Nagasaki in most cases.)
- ^
This categorisation is based on the breakdown in Baum, Seth and Barrett, Anthony, A Model For The Impacts Of Nuclear War, Global Catastrophic Risk Institute Working Paper 18-1.
- ^
Table 1 in the Bulletin of the Atomic Scientists’ Nuclear Notebook’s review of Russia’s nuclear arsenal lists the number of warheads by ICBM type. The most common ballistic missile warhead size is ambiguous, as the SS-18 M6 Satan is listed as carrying 10 500 or 800 warheads (for a total of 460 warheads). But there are an additional ~140 800 warheads on other ballistic missiles, so we think it’s reasonable to pick 800 for the purposes of this exercise.
- ^
This section was written using information from NUKEMAP and The Devastating Effects of Nuclear Weapons, MIT Press. Note that the ranges of each effect depend on the height at which the explosion occurs, which would be chosen by an attacking force and would vary depending on what they were trying to achieve. We’ve focused on the maximum ranges for an 800 kt warhead. That is, we’ve assumed the warhead explodes at the optimal height for each effect. In reality, you couldn’t achieve all these effects at once — you could increase the range of the high-blast-damage pressure wave, but that might decrease the range of the low-blast-damage wave, etc.
- ^
The US Congress asked a commission to assess the risk of an EMP attack on the US, with their report published in 2004. Figure 2 (p. 6) and figure 3 (p.7) illustrate the range of the EMP pulses.
- ^
From the report Ten years after Chernobyl: What do we really know
“Compared with other nuclear events: The Chernobyl explosion put 400 times more radioactive material into the Earth’s atmosphere than the atomic bomb dropped on Hiroshima; atomic weapons tests conducted in the 1950s and 1960s all together are estimated to have put some 100 to 1,000 times more radioactive material into the atmosphere than the Chernobyl accident.”
- ^
For further criticism, see Singer, 1985; Seitz, 2011; Reisner et al., 2019; Pausata et al., 2016; Reisner et al., 2018
- ^
For a detailed estimate, see this.
- ^
These advances were selected from Technological developments that could increase risks from nuclear weapons by Rethink Priorities.
- ^
“The idea here is that as people fall into
desperate circumstances, they will tend to take desperate measures, including resorting to
violence and committing other crimes. This idea has wide appeal but is highly contested in the scholarly literature. Studies of natural and human-caused disasters find that people do sometimes panic or misbehave, but they tend to act more cooperatively than they do under normal circumstances (e.g., Tierney et al. 2006; Rao et al. 2011). Studies of the effect of climate change
and other environmental disturbances on violent conflict are inconclusive (e.g., Gleditsch 2012).”From A model for the impacts of nuclear war — a working paper from the Global Catastrophic Risk Institute.
- ^
See our calculations in this spreadsheet. It’s important to note that by “efficient” we mean efficiency with fixed explosive power, but not efficiency with fixed costs. Smaller bombs can be more expensive per kt.
- ^
The Existential Persuasion Tournament was a forecasting event which brought together domain experts and forecasters into a multi-stage tournament where they could learn from each other and improve their estimates over time. Read more.
- ^
All numbers in this section are given as FTE (full-time equivalents). The estimate on the effective altruism community is based on this unpublished blogpost from Rethink Priorities.
- ^
Sarah Bigood, Undergraduate disarmament and nonproliferation education: gaps, opportunities, and new approaches, The Nonproliferation Review (2019).
- ^
Payload ambiguity means defenders wouldn’t know if the incoming missiles were nuclear or not and wouldn’t know where they are going, increasing the risk of escalation.
That said, after this article was first published, an expert got in touch to point out that there are some reasons to think SLCM-Ns are stabilising:
- SLCM-Ns have less target and payload ambiguity than the air leg of the nuclear triad that is already in the US arsenal. (For example, B-52 bombers are used for conventional missiles but can also hold between 1 and 20 nuclear-armed missiles, and those missiles could hit a wide-range of targets, leading to a fair amount of target and payload ambiguity.)
- It’s been argued that SLCM-Ns makes it less likely that another country would use a low-yield weapon because they provide a more effective US ability to respond in kind to that attack without having to escalate to a larger-scale attack.
Overall, we’d guess these SLCM-Ns are destabilising, but we’re unsure.
Owen Cotton-Barratt @ 2024-07-19T22:46 (+9)
I'm interested if you have views on how this intersects with advancing AI. It felt a little striking that it wasn't mentioned more, given that in other places 80k talks about a reasonable chance of transformative AI within a small number of decades.
(I think that accelerating AI progress could increase the risk of nuclear war, and I weakly guess that AI-linked risks might account for the majority of nuclear war risk over the next 30 years or so; but I'm conscious that I don't have a great basis for trying to think quantitatively about this, and feel very interested in others' takes.)
Matt Boyd @ 2024-07-20T21:50 (+3)
Similarly to Owen's comment, I also think that AI and nuclear interact in important ways (various pathways to destabilisation that do not necessarily depend on AGI). It seems that many (most?) pathways from AI risk to extinction lead via other GCRs eg pandemic, nuclear war, great power war, global infrastructure failure, catastrophic food production failure, etc. So I'd suggest quite a bit more hedging with focus on these risks, rather than putting all resources into 'solving AI' in case that fails and we need to deal with these other risks.
Vasco Grilo🔸 @ 2024-07-24T15:22 (+5)
Great points, Matt.
I think essentially all (not just many) pathways from AI risk will have to flow through other more concrete pathways. AI is a general purpose technology, so I feel like directly comparing AI risk with other lower level pathways of risk, as 80 k seems to be doing somewhat when they describe the scale of their problems, is a little confusing. To be fair, 80 k tries to account for this talking about the indirect risk of specific risks, which they often set to 10 times the direct risk, but these adjustments seem very arbitrary to me.
In general, one can get higher risk estimates by describing risk at a higher level. So the existential risk from LLMs is smaller than the risk from AI, which is smaller than the risk from computers, which is smaller than the risk from e.g. subatomic particles. However, this should only update one towards e.g. prioritise "computer risk" over "LLM risk" to the extent the ratio between the cost-effectiveness of "computer risk interventions" and "LLM risk interventions" is proportional to the ratio between the scale of "computer risk" and "LLM risk", which is quite unclear given the ambiguity and vagueness of the 4 terms involved[1].
To get more clarity, I believe it is be better to prioritise at a lower level, assessing the cost-effectiveness of specific classes of interventions, as Ambitious Impact (AIM), Animal Charity Evaluators (ACE), the Centre for Exploratory Altruism Research (CEARCH), and GiveWell do.
- ^
"Computer risk", "LLM risk", "computer risk interventions" and "LLM risk interventions".
Denkenberger @ 2024-07-20T21:51 (+6)
Existential catastrophe, annual | 0.30% | 20.04% | David Denkenberger, 2018 |
Existential catastrophe, annual | 0.10% | 3.85% | Anders Sandberg, 2018 |
You mentioned how some of the risks in the table were for extinction, rather than existential risk. However, the above two were for the reduction in long-term future potential, which could include trajectory changes that do not qualify as existential risk, such as slightly worse values ending up in locked-in AI. Also another source by this definition was the 30% reduction in long-term potential from 80,000 Hours' earlier version of this profile. By the way, the source attributed to me was based on a poll of GCR researchers - my own estimate is lower.
Vasco Grilo🔸 @ 2024-07-24T15:48 (+2)
Hi David,
Existential catastrophe, annual | 0.30% | 20.04% | David Denkenberger, 2018 |
Existential catastrophe, annual | 0.10% | 3.85% | Anders Sandberg, 2018 |
Based on my adjustments to CEARCH's analysis of nuclear and volcanic winter, the expected annual mortality of nuclear winter as a fraction of the global population is 7.32*10^-6. I estimated the deaths from the climatic effects would be 1.16 times as large as the ones from direct effects. In this case, the expected annual mortality of nuclear war as a fraction of the global population would be 1.86 (= 1 + 1/1.16) times the expected annual mortality of nuclear winter as a fraction of the global population, i.e. 0.00136 %(= 1.86*7.32*10^-6). So the annual losses in future potential mentioned in the table above are 221 (= 0.0030/(1.36*10^-5)) and 73.5 (= 0.0010/(1.36*10^-5)) times my expected annual death toll, whereas I would have expected the annual loss in future potential to be much lower than the expected annual death toll.
Matt Boyd @ 2024-07-20T21:50 (+5)
Thanks for posting this. I'll comment on the bit about New Zealand's food production in nuclear winter conditions. Although the paper cited concludes there is potential for production to feed NZ's population, this depends on there being sufficient liquid fuel to run agricultural equipment and NZ imports 100% of it's refined fuels. Trade in fuel would almost certainly collapse in a major nuclear war. Without diesel, or imported fertiliser and agrichemicals, yield would be much lower. Distribution would be difficult too. See this paper: https://onlinelibrary.wiley.com/doi/abs/10.1111/risa.14297 Ideally, places like NZ would establish the potential to produce fuel locally, eg biofuels, in case of this scenario. If restricted to use in agriculture and food transport, with optimised cropping, surprisingly little biofuel would be needed. This kind of contingency planning could avert famine, and any associated disease and potential conflict. I agree that the existential risk is very low. But it is probably slightly higher when considering these factors.
Radical Empath Ismam @ 2024-07-21T01:42 (+4)
I also want to say I've been really impressed by the work of the Quincy Institute (https://quincyinst.org/). They advocate for an anti-war & anti-interventionist foreign policy, which is very neglected amongst think tanks in this area due to the high levels of funding by weapons-makers and the military-industrial complex. They're work on the Iran nuclear deal was highly lauded.
Nuclear weapons are of course a specific area they focus on, but their more holistic Anti-war & Anti-interventionist advocacy has noticeably improved the foreign policy debate and challenged hawkish perspectives that dominate this space and I suspect is a very effective way to reduce X-Risk.
Would love to see EAs engaged in Anti-war & Anti-interventionist spaces as they seem very neglected and a chance to make an outsized impact.
The nobel-winning ICANW also does really excellent work and spoke at EAGx Australia.
Zoé Roy-Stang @ 2024-07-21T01:23 (+3)
Who can I talk to about a nuclear war prevention idea? (And is there a public repository of ideas?)