itās not the case that āSome research on nuclear winter suggests that 100 Hiroshima- sized nuclear detonations would be enough to destroy the majority of human life on earthā
even the smaller claim that that research does make is contested and some parts of it are based on pretty shoddy methods (especially the reasoning to go from reduced crop yields to famine death)
āthe majority of human life on earthā would in any case be less than āalmost all human life on earthā, and the distinction might matter a lot from an existential risk perspective (since recovery looks much less likely if we have 0.01% of people left than if we have 1%, and much less likely with 1% than with 40%)
To expand on that:
I donāt remember what the specific paper you cite claims about how many deaths would occur, and I canāt quickly find it from skimming, but more recent work by the same authors tends to make the claim that there might be ~2 billion deaths. That is a hell of a lot, but also āonlyā 25% of the worldās population.
And that number is based in part on this non-peer-reviewed report by an advocacy organisation, the methodology in which is very shoddy. (One key issue, from memory, is that it basically argues for ā2 billion people could be food insecureā, and then later implies that it had argued for ā2 billion people could die of famineā, despite the fact that thereās a big difference between those two things.)
And other authors question other parts of the models leading to that estimate.
(I have more detailed notes on all of this, which I could share on request, but theyāre not polished.)
Btw, hereās a relevant section of a post Iām drafting on ā10 mistakes to avoid when thinking about nuclear riskā, which overviews what I see as some key points on nuclear winter etc.. (I could probably share the draft with you if you want.)
Mistake 5 & 6: Ignoring the possibility of major climate and famine effects following nuclear conflictāor overstating the likelihood/āseverity of those effects
When thinking about nuclear risk, people often focus on the immediate harms (e.g., from the blast) and the harms from radioactive fallout. And those harms could indeed be huge! But those harms could be dwarfed by the harms from major cooling of the climateāperhaps a nuclear winter, or perhaps a smaller version of the same effects. That cooling could perhaps cause huge numbers of famine deaths (plausibly in the billions, for some nuclear conflicts). And this seems the most likely way for nuclear war to cause an existential catastrophe.[1]
...or maybe not! The effects depend on factors such as:
how many detonations occur
how much flammable material is in the targeted areas
how much black carbon fires in these areas would produce and would reach high enough in the atmosphere to persist there for years
how severely agricultural production would be reduced by various potential climate effects
how people would respond to expected or occurring agricultural production issues (e.g., how well could they adjust what crops they grow, where, and how; how much would food usage patterns change; would international trade continue)
how likely civilization is to recover from a collapse
And, unfortunately, each of those questions are contested, complex, and under-researched.
Ultimately, I suggest:
Recognising that major climate and famine effects are plausible, but that whether theyāll happen and how bad theyāll be is quite uncertain.
Seeing that as a key consideration when deciding (a) how much to prioritise nuclear risk relative to other problems and (b) which nuclear conflict scenario to prioritise reducing the odds of.
Considering interventions to reduce how bad the climate or famine effects would be (e.g., [[example]])
[1] I still consider myself quite confused on those topics, but here are some of my current bottom-line beliefs, in brief:
Climate effects severe enough to qualify as ānuclear winterā seem likely in scenarios in which there are thousands of nuclear detonations on high-population-density areas (e.g., cities or towns). In contrast, nuclear winter seems unlikely in scenarios with less than a hundred detonations on high-population-density areas. I really wish I had a clearer sense of the probabilities, especially for various scenarios between those extremes. (See, e.g., Toon et al., 2007; Reisner et al., 2018; Robock et al., 2019.)
If nuclear winter occurred, it would probably cause at least hundreds of millions of deaths. Itās also plausiblebut unlikely that itād lead to an existential catastrophe (via the resulting famine combined with other effects, e.g., further conflict triggered by the famine). (See, e.g., Aird, 2020; Beckstead, 2015; Ladish, 2020; Ord, 2020; Rodriguez, 2019; Rodriguez, 2020.)
Climate effects that are similar to but smaller than ānuclear winterā could plausibly cause hundreds of millions or perhaps billions of deaths. But such effects would be much less likely to cause existential catastrophe.
Even a low probability of existential catastrophe is still really terrible and can be well-worth reducing further!
Just coming back to this- thanks for these comments! In light of your and Larkās comments Iād no longer endorse this section: āSome research on nuclear winter...ā. Iāll be very interested to hear your coming
However Iām still very concerned by the precedent this sets for nuclear non-proliferation. This move seems a pretty clear breach of the non-proliferation treaty , and the risks it created of a new nuclear arms race remains the central thing that Iām worried about.
Perhaps more relevant to that since this was published is the new of the new AUKUS security pact. Iām really glad to see that being discussed on the EA Forum here.
(Btw, on non-proliferation and arms racing, there are some relevant forecasts as part of the Nuclear Risk Forecasting Tournament Iām putting together with Metaculus, and there will be more put up next week and then again over the coming months. Iāll also write up some summaries later.)
To add to what Larks said, I would also say that:
itās not the case that āSome research on nuclear winter suggests that 100 Hiroshima- sized nuclear detonations would be enough to destroy the majority of human life on earthā
even the smaller claim that that research does make is contested and some parts of it are based on pretty shoddy methods (especially the reasoning to go from reduced crop yields to famine death)
āthe majority of human life on earthā would in any case be less than āalmost all human life on earthā, and the distinction might matter a lot from an existential risk perspective (since recovery looks much less likely if we have 0.01% of people left than if we have 1%, and much less likely with 1% than with 40%)
To expand on that:
I donāt remember what the specific paper you cite claims about how many deaths would occur, and I canāt quickly find it from skimming, but more recent work by the same authors tends to make the claim that there might be ~2 billion deaths. That is a hell of a lot, but also āonlyā 25% of the worldās population.
And that number is based in part on this non-peer-reviewed report by an advocacy organisation, the methodology in which is very shoddy. (One key issue, from memory, is that it basically argues for ā2 billion people could be food insecureā, and then later implies that it had argued for ā2 billion people could die of famineā, despite the fact that thereās a big difference between those two things.)
And other authors question other parts of the models leading to that estimate.
(I have more detailed notes on all of this, which I could share on request, but theyāre not polished.)
Btw, hereās a relevant section of a post Iām drafting on ā10 mistakes to avoid when thinking about nuclear riskā, which overviews what I see as some key points on nuclear winter etc.. (I could probably share the draft with you if you want.)
Mistake 5 & 6: Ignoring the possibility of major climate and famine effects following nuclear conflictāor overstating the likelihood/āseverity of those effects
When thinking about nuclear risk, people often focus on the immediate harms (e.g., from the blast) and the harms from radioactive fallout. And those harms could indeed be huge! But those harms could be dwarfed by the harms from major cooling of the climateāperhaps a nuclear winter, or perhaps a smaller version of the same effects. That cooling could perhaps cause huge numbers of famine deaths (plausibly in the billions, for some nuclear conflicts). And this seems the most likely way for nuclear war to cause an existential catastrophe.[1]
...or maybe not! The effects depend on factors such as:
how many detonations occur
how much flammable material is in the targeted areas
how much black carbon fires in these areas would produce and would reach high enough in the atmosphere to persist there for years
how severely agricultural production would be reduced by various potential climate effects
how people would respond to expected or occurring agricultural production issues (e.g., how well could they adjust what crops they grow, where, and how; how much would food usage patterns change; would international trade continue)
how likely civilization is to recover from a collapse
And, unfortunately, each of those questions are contested, complex, and under-researched.
Ultimately, I suggest:
Recognising that major climate and famine effects are plausible, but that whether theyāll happen and how bad theyāll be is quite uncertain.
Seeing that as a key consideration when deciding (a) how much to prioritise nuclear risk relative to other problems and (b) which nuclear conflict scenario to prioritise reducing the odds of.
Considering interventions to reduce how bad the climate or famine effects would be (e.g., [[example]])
[1] I still consider myself quite confused on those topics, but here are some of my current bottom-line beliefs, in brief:
Climate effects severe enough to qualify as ānuclear winterā seem likely in scenarios in which there are thousands of nuclear detonations on high-population-density areas (e.g., cities or towns). In contrast, nuclear winter seems unlikely in scenarios with less than a hundred detonations on high-population-density areas. I really wish I had a clearer sense of the probabilities, especially for various scenarios between those extremes. (See, e.g., Toon et al., 2007; Reisner et al., 2018; Robock et al., 2019.)
If nuclear winter occurred, it would probably cause at least hundreds of millions of deaths. Itās also plausible but unlikely that itād lead to an existential catastrophe (via the resulting famine combined with other effects, e.g., further conflict triggered by the famine). (See, e.g., Aird, 2020; Beckstead, 2015; Ladish, 2020; Ord, 2020; Rodriguez, 2019; Rodriguez, 2020.)
Climate effects that are similar to but smaller than ānuclear winterā could plausibly cause hundreds of millions or perhaps billions of deaths. But such effects would be much less likely to cause existential catastrophe.
Even a low probability of existential catastrophe is still really terrible and can be well-worth reducing further!
Just coming back to this- thanks for these comments! In light of your and Larkās comments Iād no longer endorse this section: āSome research on nuclear winter...ā. Iāll be very interested to hear your coming
However Iām still very concerned by the precedent this sets for nuclear non-proliferation. This move seems a pretty clear breach of the non-proliferation treaty , and the risks it created of a new nuclear arms race remains the central thing that Iām worried about.
Perhaps more relevant to that since this was published is the new of the new AUKUS security pact. Iām really glad to see that being discussed on the EA Forum here.
(Btw, on non-proliferation and arms racing, there are some relevant forecasts as part of the Nuclear Risk Forecasting Tournament Iām putting together with Metaculus, and there will be more put up next week and then again over the coming months. Iāll also write up some summaries later.)