epistemic status: I am fairly confident that the overall point is underrated right now, but am writing quickly and think it's reasonably likely the comments will identify a factual error somewhere in the post.
Risk seems unusually elevated right now of a serious nuclear incident, as a result of Russia badly losing the war in Ukraine. Various markets put the risk at about 5-10%, and various forecasters seem to estimate something similar. The general consensus is that Russia, if they used a nuclear weapon, would probably deploy a tactical nuclear weapon on the battlefield in Ukraine, probably in a way with a small number of direct casualties but profoundly destabilizing effects.
A lot of effective altruists have made plans to leave major cities if Russia uses a nuclear weapon, at least until it becomes clear whether the situation is destabilizing. I think if that happens we'll be in a scary situation, but based on how we as a community collectively reacted to Covid, I predict an overreaction -- that is, I predict that if there's a nuclear use in Ukraine, EAs will incur more costs in avoiding the risk of dying in a nuclear war than the actual expected costs of dying in a nuclear war, more costs than necessary to reduce the risks of dying in a nuclear war, and more costs than we'll endorse in hindsight.
With respect to Covid, I am pretty sure the EA community and related communities incurred more costs in avoiding the risk of dying of Covid than was warranted. In my own social circles, I don't know anyone who died of Covid, but I know of a healthy person in their 20s or 30s who died of failing to seek medical attention because they were scared of Covid. A lot of people incurred hits to their productivity and happiness that were quite large.
This is especially true for people doing EA work they consider directly important: being 10% less impactful at an EA direct work job has a cost measured in many human or animal or future-digital-mind lives, and I think few people explicitly calculated how that cost measured up against the benefit of reduced risk of Covid.
If Russia uses a nuclear weapon in Ukraine, here is what I expect to happen: a lot of people will be terrified (correctly assessing this as a significant change in the equilibrium around nuclear weapon use which makes a further nuclear exchange much more likely.) Many people will flee major cities in the US and Europe. They will spend a lot of money, take a large productivity hit from being somewhere with worse living conditions and worse internet, and spend a ton of their time obsessively monitoring the nuclear situation. A bunch of very talented ops people will work incredibly hard to get reliable fast internet in remote parts of Northern California or northern Britain. There won't be much EAs not already in nuclear policy and national security can do, but there'll be a lot of discussion and a lot of people trying to get up to speed on the situation/feeling a lot of need to know what's going on constantly. The stuff we do is important, and much less of it will get done. It will take a long time for it to become obvious if the situation is stable, but eventually people will mostly go back to cities (possibly leaving again if there are further destabilizing events).
The recent Samotsvety forecast estimates that a person staying in London will lose 3-100 hours to nuclear risk in expectation (edit: which goes up by a factor of 6 in the case of actual tactical nuke use in Ukraine.) I think it is really easy for that person to waste more than 3-100 hours by being panicked, and possible to waste more than 20 - 600 hours on extreme response measures. And that's the life-hour costs of never fleeing; you also have the option of fleeing at a later point if there are further worrying developments, and it's probably a mistake to only model 'flee as soon as there's tactical nuke use' against 'stay no matter what' and not against 'flee slightly later'.
Some degree of costs incurred is quite reasonable. I think that during the Cuban Missile Crisis we were quite close to nuclear war, and probably reasonable people at the time would have (in addition to trying to prevent such a war) tried to leave major cities. I think it might make sense for people whose work is already remote, and who have a non-major-city place to stay, to leave. I think that the fact EAs are weird, and take our beliefs more seriously than most people, and take concrete actions based on expected-value arguments, is a strength. Certainly at some threshold of risk I'll leave with my family. But my overall expectation is that we'll end up causing more disruptions than are justified, at substantial expense to other work which is also about securing a good human future.
Some specific bad tradeoffs that seem easy to avoid:
- going to a very remote area dramatically reduces the risk of being hit in a nuclear exchange, but makes it incredibly inconvenient to work normally/collaborate with others/etc. DC, New York, and San Francisco are among the highest-likelihood-of-being-hit-in-a-full-nuclear-exchange cities in the US, and London obviously the highest in the UK, but if you go from those cities to nearby populous cities you probably get most of the benefit and incur lower costs in productivity/etc. In my opinion Americans who don't live in those cities and don't live next to a base from which we launch our own nukes shouldn't bother leaving (I don't know enough about continental Europe to have opinions there.) For San Franciscans, going to Santa Rosa is probably nearly as good as going to the middle of the mountains, or going to Eugene, and it's much less costly.
- we can avoid socially pressuring people towards acting more strongly than they endorse: if you don't care about this, I think you should feel licensed to not care about it and go about your life normally.
- if people do want to leave, I think they should keep in mind the productivity costs of 1) bad internet 2) isolation, and strongly prioritize going somewhere non-isolated with good internet access.
- keep in mind that while expected value sometimes implies reacting strongly to things with only a small chance of being really bad, you have to actually do the expected value calculations -- and your wellbeing, productivity, and impact on the world should be an input.
I think "it's easy to overreact on a personal level" is an important lesson from covid, but much more important is "it's easy to underreact on a policy level". I.e. given the level of foresight that EAs had about covid, I think we had a disappointingly small influence on mitigating it, in part because people focused too much on making sure they didn't get it themselves.
In this case, I've seen a bunch of people posting about how they're likely to leave major cities soon, and basically zero discussion of whether there are things people can do to make nuclear war overall less likely and/or systematically help a lot of other people. I don't think it's bad to be trying to ensure your personal survival as a key priority, and I don't want to discourage people from seriously analysing the risks from that perspective, but I do want to note that the overall effect is a bit odd, and may indicate some kind of community-level blind spot.
I strongly agree with your comment, but I want to point out in defense of this trend that nuclear weapons policy seems to be unusually insulated from public input and unusually likely to be highly sensitive/not good to discuss in public.
I think EAs are broadly too quick to class things as infohazards instead of reasoning them through, but natsec seems like a pretty well defined area where the reasons things are confidential are pretty concrete .
Some examples of information that is pretty relevant to nuclear risk and would not be discussed on this forum, even if known to some participants:
How well-placed are US spies in the Russian government and in Putin's inner circle?
How about Russian spies in the US government? Do the Russians know what the US response would be in the event of various Russian actions?
Does the US know where Russia's nuclear submarines are? Can we track their movements? Do we think we could take them out if we had to? This would require substantial undisclosed tech. If we did know this, it would be a tightly held secret; degrading Russia's second-strike capabilities (which is one effect of knowing where their subs are) might push them towards a first strike.
Relatedly, are we at all worried Russia knows where our submarines are?
In a similar genre, does the US know how to shoot down ICBMs? With 10% accuracy? 50%? 80%? Accuracy would have to be very good to be a game changer in a full exchange... (read more)
I strongly agree with the general point that overreaction can be very costly, and I agree that EAs overreacted to Covid, particularly after it was already clear that the overall infection fatality rate of Covid was under 1%, and roughly 0.02% in young adults.
However, I think it's important to analyze things on a case-by-case basis, and to simply think clearly about the risk we face. Personally, I felt that it was important to react to Covid in January-March 2020 because we didn't understand the nature of the threat yet, and from my perspective, there was a decent chance that it could end up being a global disaster. I don't think the actions I took in that time—mainly stocking up on more food—were that costly, or irrational. After March 2020, the main actions I took were wearing a mask when I went out and avoiding certain social events. This too, was not very costly.
I think nuclear war is a fundamentally different type of risk than Covid, especially when we're comparing the ex-ante risks of nuclear war versus the ex-post consequences of Covid. In my estimation, nuclear war could kill up to billions of people via very severe disruptions to supply chains. Even at the... (read more)
To be clear, I will also leave SF in the event of a strong signal that we're on the brink of nuclear war -- such as US officials saying they believe Russia is preparing for a first launch, or the US using a nuclear weapon ourselves in response to Russian use, or strategic rather than tactical Russian use (for example against Kyiv), or Russia declaring war on NATO or declaring intent to use nuclear weapons outside Russian territory.
I mostly expect overreaction in cases of a weaker signal such as a Russian "test" on territory Russia claims as Russian, or tactical use, or Russia inducing a meltdown at a nuclear power plant -- all of which would be scary, destabilizing, precedent-setting events that dramatically raise the odds of a nuclear war, but which I wouldn't call a "clear and unambiguous signal that a large amount of the world may be utterly destroyed in a matter of hours".
I think it's worth noting that that I'd expect you would gain a significant relative advantage if you get out of cities before other people, such that acting later would be a lot less effective at furthering your survival & rebuilding goals.
I expect the bulk of the risk of an all out nuclear war to happen in the couple of weeks after the first nuclear use. If I'm right, then the way to avoid the failure mode you're identifying is returning in a few weeks if no new nuclear weapons have been used, or similar.
Beating the traffic perhaps; getting stuck in your car trying to leave SF is worse than sheltering in your SF basement.
I am somewhat surprised that a tactical nuke use by Russia isn't sufficient. A naive fermi I did on Samotsvety's numbers suggest at that point an hour in SF at that point costs you about 2 hours in-expectation, so something about our fermis must be very different, since that seems very likely worth leaving for.
I strongly do not expect full nuclear exchange in immediate response to Russia tac nuke use; the situation that seems plausible to me would involve conventional retaliation against Russian forces in Ukraine, Syria, etc., followed by Russia responding to that. So I think leaving at a further point still means leaving well ahead of a full exchange.
I think my work is much more valuable in worlds without a full nuclear exchange; iirc you are pretty doomy on current trajectories, so maybe you actually think your work is more valuable in worlds with a full nuclear exchange, or at least of comparable value?
I think I'm twice as productive at home, for reasons relating to childcare, disruption associated with fleeing, personal traits, my home being well set up to meet my needs, diet, etc.
Oh, hmm, this might be a big difference. I think my work might be 10x more valuable in worlds with nuclear exchange (since I think the world becomes a lot more malleable as a result of such a crisis, seems like there is a big opportunity to change humanity's relation to existential risk, I have a broad generalist skillset, and if there are fewer people around but I survive, seems like I should have a higher prior that I can influence humanity's future).
I am currently just using a 1x multiplier in my estimates, but I think a 3-5x would more accurately capture my beliefs.
To be... (read more)
Another important consideration that is not often mentioned (here and in our forecast) is how much more/less impact you expect to have after a full-out Russia-NATO nuclear war that destroys London.
I think there's a good chance this basic point is right, but I'm not sure your takeaway from the Samotsvety forecast is correct? I think the 3-100 hours lost in expectation is based on the current information about risk. The Samotsvety forecast is that conditional on a nuclear weapon being used in Ukraine, there is a ~2% chance of London being nuked. I think the mean estimate for expected hours of life loss if one stays in London in that case is ~2000. That's a substantial number of lost hours, and I can see it being rational to get to a safer location if those are the stakes.
Hmm interesting, I got 2000 by just setting rusiaUsesNuclearWeaponsInUkraine to 1 in the squiggle model. Looking at it further, the mean moves around between runs if I just use 1000 samples. Updating to 1000000, it seems to converge on 1700.
I agree that this is a place where forecast aggregation adds a lot of challenges.
This is also tricky because I don't think it lets you compare to the option I'd actually advocate for, which is something like "flee at a slightly later point" - the US has good intel on Russia, and it seems likely that US officials will know if Russia appears to be headed towards nuclear war. If you have to compare "flee the instant a tactical nuke is used in Ukraine" or "stay no matter what", "stay no matter what" doesn't look good, but what you want to compare is "flee the instant a tactical nuke is used in Ukraine" to "flee at some subsequent sign of danger" -- that is, the real question is how many life-hours you get by fleeing early that you don't get by fleeing late (either because we don't get any warning, or because by then many people are panicking and fleeing).
Why would you think that they would transmit this information honestly, rather than managing the crowd and try to have people not panic?
There are a bunch of preparations the US military would want to take in the face of elevated odds of nuclear war (bombers in the air, ships looking for submarines, changes of force concentration) and I don't believe they will sacrifice making those preparations for crowd management reasons. I agree it's possible they'll say something noncommittal or false while visibly changing force deployments to DEFCON 2 or whatever, though this is not what they did during the Cold War and it would be pretty obvious.
Though I should say that I think tac nuke use in Ukraine is also a reasonable trigger to leave, depending on your personal situation, productivity, ease of leaving, where you're going, etc - I really just want people to be sure they are doing the EV calculations and not treating risk-minimization as the sudden controlling priority.
I totally agree that:
However, my model is that any use of nuclear weapons "in anger" is a Big Red Line. It's become almost unthinkable through eight decades of tradition, through dozens of high-tension periods, regional wars and international crises. I don't think the public will really distinguish between "tactical" use and "strategic" use. The exact yields of Russian "tactical" devices seem to be secret, but I don't think it'd be much less powerful than Hiroshima, and given Ukraine's geography it'd almost certainly hit civilian homes somewhere. Because of that, my p(nukes used in Ukraine) is pretty low (maybe 6%?), but p(SF gets nuked|nukes used in Ukraine) is fairly high (maybe 25%); there would be intense escalation pressure on both sides.
Comment from General Hyten, former head of STRATCOM:
"In 2017, then-Air Force Gen. John E. Hyten took exception to the idea that tactical nuclear weapons are really in a different category than strategic nuclear weapons. Hyten, who was at that point overseeing U.S. nuclear weapons as the chief of U.S. Strategic Command, described how the United States could respond if another country used them.
“It’s not a tactical effect, and if somebody employs what is a nonstrategic or tactical nuclear weapon, the United States will respond strategically, not tactically, because they have now crossed a line, a line that has not been crossed since 1945,” Hyten said."
"Mr. Biden also challenged Russian nuclear doctrine, warning that the use of a lower-yield tactical weapon could quickly spiral out of control into global destruction.
"I don't think there is any such a thing as the ability to easily use a tactical nuclear weapon and not end up with Armageddon," Mr. Biden said."
I think this pretty true for the San Francisco-based and rationalist strands of the community. But in New York/London/Oxford, people were more balanced - being cautious for an initial several months, then reverting back to normal at a similar time to the rest of society.
As someone who helped raise the alarm about Covid (and is trying to do so for this one as well), I have wondered if my actions were actually harmful. I posted an update on Facebook in I think May 2020 something to the effect that more harm may come from EAs losing productivity than from the actual disease. I consider myself a pretty good updater for these situations but a lot of people are subject to information cascades. I do think some people remained, frankly, way too fucking neurotic about this longer than was reasonable. I wish more people grokked the coordination cost of imposing more friction along their collaboration surface area. As an example, there was a post I think last autumn that was like "what is EAG doing about Covid?" and I considered that annoying and felt sorry for EAG people.
One argument in favor of your viewpoint is that if global nuclear war happens, there's really not much EA work left to do in the aftermath besides help a few people around you if you survived. That might be comparable to global health and development relief? Maybe global poverty people who live in these cities and think they have alpha on what lifesaving efforts they could somehow participa... (read more)
Thanks for this thoughtful reflection. I do want to register that I think I disagree there wouldn't be much EA to do post- a nuclear exchange between the US and Russia - it would be a scary hard world to live in, and one where many of our previous priorities are no longer relevant, but it's work I think we could do and could improve the trajectory of civilization by doing.
If there is a loss of civilization, I agree there would not be that much that EAs could do. However, I think there is a lot we could do to try to prevent the loss of civilization after nuclear war, which is part of what these posts were about. Yes, ALLFED is working on solutions that could be scaled up quickly in the case of nuclear war.
Re: COVID, the correct course of action (unless one was psychic) was to be extremely paranoid at the start (trying for total bubbling, sterilizing outside objects, etc) because the EV was very downside-skewed—but as more information came in, to stop worrying about surfaces, start being fine with spacious outdoor gatherings, get a good mask and be comfortable doing some things inside, etc.
That is, a good EA would have been faster than the experts on taking costly preventative acts and faster than the experts on relaxing those where warranted.
Some actual EAs seemed to do this well, and others missed in one direction or the other (there was a lot of rapid group house self-sorting in March/April 2020 over this, and then a slower process afterward).
Yep, agree - I think it was warranted to be extremely cautious in February/March, and then the ideal behavior would have been to become much less cautious as more information came in. In practice, I think many people remained extremely cautious for a full year (including my family) out of some combination of inertia and exhaustion about renegotiating what had been strenuously negotiated in the first place.
Some people furthermore tried very aggressively to apply social pressure against fully vaccinated people holding events and returning to normalcy in spring of 2021, which I think was an even more clear-cut mistake given the incredibly high pre-omicron vaccine efficacy. I am not actually sure I know anyone who I believe missed in the incautious direction, and if we'd had equal misses in both directions I'd feel a lot better about our community decisionmaking.
Can someone share their credence for why Russia would use only a few warheads against America, attacking only a few cities? If Russia uses many warheads, most US cities that most of us would feel at ease in are a no-go. If we expect Russia to use many if they use them, it is even less worth leaving (unless you want to exit America and NATO regions).
What is America's public-facing policy on retaliation? Do they fire number of warheads proportional to how many Russia fires at us/our allies? If so, it incentivizes Russia using fewer missiles. But I have the vibe that we have told them we would use everything at our disposal, so they don't have a reason to only target a handful of cities.
[[EDIT: Found the source of the picture. It is from 2015 at latest, so would look (very?) different today. And don't think of the purple triangles as first priority to be launched by Russi... (read more)
The black dots assumes Russia has 2000 functional missiles that they successfully launch against the US and that successfully detonate, and that the US is unable to shoot many of them down/destroy missile launch sites before launch. My understanding is, concretely, that even if all Russian missiles currently reported ready for launch are launched, there's 1500 of them not 2000, and that one would expect many to be used against non-US targets (in Ukraine and Europe). The 500 scenario (purple triangles) seems likelier to me for how many targets Russia would try to hit.
Further, my impression of the competence of the Russian military, the readiness of their forces, the state of upkeep on their nukes and missiles, the willingness of individual commanders ordered to launch to do so, etc. is quite low. In many cases they have had an incredibly embarrassingly low success rate at firing missiles at Ukraine, which is an easier task than launching on short notice in a nuclear war. They seem to be using un-upgraded Soviet technology that is often degrading and failing, and the theft of parts for sale on the black market isn't uncommon.
For each nuclear missile, lots of things ... (read more)
I appreciate you making this post, and agree with many of your points. One thing I'd add is that the situation of strategic nuclear warfare is unprecedented and would be extremely chaotic. In my mind this significantly raises the value of more traditional (in the sense of having worked fairly well across human history) hedges against uncertain situations. For example, while travelling to New Zealand or South America might be a good hedge against worst case scenarios of a full strategic exchange, being near friends + family or being somewhere you are a citizen might be much better than being in New Zealand in the case of massive civil unrest or less-than-expected weather disruption.
I think it's worth mentioning that it isn't obvious you should use linear EV maximisation when probabilities are very low (say, less than 1%).
When probabilities of events are high, EV maximisation makes sense because you only need a few events before law of large numbers kicks in and people not doing EV maximization are much worse off than people doing it in nearly all worlds.
When probabilities of events are very low you need a much larger number of events before you see the same effect.
What's the rationale behind New York and San Francisco?
If it's because they are major population centers, what are reasonable estimates + rationale for P(target population | nuclear war)? I would have guessed attacking military sites (and in particular nuclear-related sites) would be much more likely, and I don't think there are major military targets very close to San Francisco.
How irrational was it to be concerned about long covid pre-vaccination? That was my mistake. I presume I should've done something differently. But I don't have much medical knowledge or a strong intuition for how to analyze a study.
One thing that would be really useful in terms of personal planning, and maybe would be a good idea to have a top level post on, is something like:
What is P(I survive | I am in location X when a nuclear war breaks out)
for different values of X such as:
(A) a big NATO city like NYC
(B) a small town in the USA away from any nuclear targets
(C) somewhere outside the US/NATO but still in the northern hemisphere, like Mexico. (I chose Mexico because that's probably the easiest non-NATO country for Americans to get to)
(D) somewhere like Argentina or Australia, the ... (read more)
Overall I disagree with the general thrust of this post on the grounds that the risk of death from nuclear war (conditional on nuclear use in Ukraine) seems so much higher than from covid for the typical EA demographic, but I would like to +1 the point that people may be able to eliminate most of their risk without moving to completely the middle of nowhere.
Russia has enough active nuclear weapons to completely destroy something like 1-500,000 km^2*. The UK alone has a land area of 250,000 km^2, the US is around 10 million km^2 and NATO as a whole is 25 mi... (read more)