Intermediate goals for reducing risks from nuclear weapons: A shallow review (part 1/4)

By MichaelA🔸, Will Aldred @ 2023-05-01T15:04 (+35)

This is a linkpost to https://docs.google.com/document/d/1hwFCLtEbkCmi3p9nuHYIe7YgRcOazLes0X7ZFr3B990/edit#

This is a blog post, not a research report, meaning it was produced relatively quickly and is not to Rethink Priorities' typical standards of substantiveness and careful checking for accuracy.

Summary

What is this post? 

This post is the first part of what was intended to be a shallow review of potential “intermediate goals”^[1] one could pursue in order to reduce nuclear risk (focusing especially on the contribution of nuclear weapons to existential risk). The full review would’ve broken intermediate goals down into:

1. goals aimed at reducing the odds of nuclear conflict or other non-test nuclear detonations
2. goals aimed at changing how nuclear conflict plays out if it does occur (in a way that reduces its harms)
3. goals aimed at improving resilience to or ability to recover from the harms of nuclear conflict
4. goals that are cross-cutting, focused on field-building, or otherwise have indirect effects 

This first part of the shallow review focuses just on the first of those categories: goals aimed at reducing the odds of nuclear conflict or other non-test nuclear detonations. We tentatively think that, on the margin, this is the least promising of those four categories of goals^[2], but that there are still some promising interventions in this category. 

Within this category, we review multiple potential goals. For most of those goals, we briefly discuss:

• What we mean by the goal
• Why progress on this goal might reduce or increase nuclear risk^[3]
• Examples of specific interventions or organizations that could advance the goal 
• Our very tentative bottom-line beliefs about:
  • How much progress on the goal would reduce or increase nuclear risk^[4]
  • What resources are most needed to make progress on the goal^[5] 
  • How easy making progress on the goal would be^[6]
  • What key effects making progress on the goal might have on things other than nuclear risk^[7]

Note that, due to time constraints, this post is much less comprehensive and thoroughly researched and reviewed than we’d like. 

The intermediate goals we considered, and our tentative bottom line beliefs on them

This post and table breaks down high-level goals into increasingly granular goals, and shares our current best guesses on the relatively granular goals. Many goals could be pursued for multiple reasons and could hence appear in multiple places in this table, but we generally just showed each goal in the first relevant place anyway. This means in some cases a lot of the benefits of a given goal may be for higher-level goals we haven’t shown it as nested under. 

We unfortunately did even less research on the goals listed from From 1.1.2.6 onwards than the earlier ones, and the bottom-line views for that later set are mostly Will’s especially tentative personal views. 

Some high-level takeaways and observations

• There’s a huge number of goals one could focus on, and, in most cases, a huge number of ways of pursuing each goal. One shouldn’t assume the first goals or interventions they thought of, or the ones that are most often discussed, are the only or best options.
• There’s often major uncertainty about whether progress toward a goal would reduce or actually increase nuclear risk, how easy progress toward the goal would be, and what key effects progress toward this goal might have on things other than nuclear risk.
• That said, we think we can already be moderately confident about prioritizing some goals and deprioritizing others. 
  • In some cases, this involves deviating from the priorities of the mainstream nuclear security community. (Partly precisely because such people’s work makes some goals less neglected, which is a good thing!) 
  • For example, we tentatively think that, on the margin, trying to reduce the odds of nuclear conflict or other non-test nuclear detonations is lower priority than trying to change how nuclear conflict plays out if it does occur, improve resilience to or ability to recover from the harms of nuclear conflict, or do useful cross-cutting / field-building / indirect work.^[8]
• We arrived at our views via a fairly rushed job, and we can see clear ways one could further improve clarity. We think that evaluating a given set of intermediate goals is a well scoped and tractable task which doesn’t require flashes of rare insight.
  • Thus, for junior researchers who want to work on nuclear risk, it may be worth considering trying to (a) come up with intermediate goals ideas not listed in our documents and/or (b) evaluate whether and how best to pursue some of the relatively promising-seeming intermediate goals. It may be best for such work to focus not on reducing the odds of nuclear war but rather how it plays our or society’s resilience and recovery ability (see footnote 9). 
  • See our “Nuclear risk research ideas” post, in particular the “Intermediate goals for nuclear risk reduction” project idea.

0. Introduction & why this post may be useful   

Nine countries possess a total of roughly 13,000 nuclear weapons (FAS, 2021). The chance of nuclear conflict by 2050 appears to be around 25%.^[9] If nuclear conflict does occur, it might involve the targeting of cities, the use of hundreds or thousands of warheads, immediate fatalities in the tens of millions or higher, and/or severe climate effects resulting in billions of deaths due to famine.^[10] It seems unlikely that even such severe consequences would cause an existential catastrophe, but the odds still seem to be uncomfortably high, and seem sufficient to make nuclear war one of the largest sources of existential risk.^[11] 

But what can be done about this? What “intermediate goals” could a person support in order to reduce nuclear risk? By what mechanisms and to what extent would progress on each goal reduce - or accidentally increase - nuclear risk? What interventions are available for supporting each intermediate goal, what types of resources do they most need, and how much progress should be expected if a given level of those resources is devoted to this goal?

This post is intended to help people think about those questions, which seems crucial when a person is considering (1) whether to focus on reducing nuclear risk or instead on other issues or (2) what to focus on within the broad area of reducing nuclear risk. Here are some ways readers could use this post:

• To gain a better understanding of the key pathways for and factors affecting nuclear risk, as this post is essentially organized around those factors
• To gain a better understanding of the overall picture a given intervention or intermediate goal fits into and which (other) intermediate goals it might affect
• To expand the set of interventions and intermediate goals a reader considers when thinking about how tractable nuclear risk reduction is or which actions the reader should take to reduce nuclear risk
• To inform decisions about which specific interventions, intermediate goals, or organizations to support
• To highlight areas where further research could be valuable
  • See “Research project idea: Intermediate goals for nuclear risk reduction”

Before proceeding, some caveats are in order:     

• This post was intended to be fairly comprehensive and well-organized at the level of potential intermediate goals (though ultimately we ran out of time to really hit that objective). We were not aiming to discuss only a prioritized set of goals, to be comprehensive at the level of interventions, or to go into depth on individual goals or interventions.^[12]
• We mostly ignore possible benefits or harms that are unrelated to nuclear risk (e.g., reducing or increasing fatalities from non-nuclear weapons). Thus, this post doesn’t provide all-things-considered assessments of whether to support each intervention or goal. 

Finally, a few points that may be useful to bear in mind:

• Many of these intermediate goals or interventions could focus either on specific states, specific pairings or small sets or states, or a broader set of states.
• Many of these goals or interventions may be (much) more tractable in democratic rather than non-democratic countries, since the former are easier to influence via things like lobbying and public advocacy. 
• In this post, each time we say “odds of x” or “odds or frequency of x”, the precise thing we have in mind is more like “the expected number of x over the next few decades”. 

Epistemic status

In 2021, Michael did some initial research for this post and wrote an outline and rough notes. But he pivoted away from nuclear risk research before having time to properly research and draft this. We (Michael and Will) finished a rough version of this post in 2022, since that seemed better than it never being published at all, but then didn’t get around to publishing until 2023. As such, this is just a very incomplete starting point, may contain errors, and may be outdated (e.g., it was mostly written before the Russian invasion of Ukraine).

This shallow review gets decreasingly detailed and carefully researched as it proceeds, due to Michael to some extent working on this from top to bottom and then facing time constraints. This doesn’t indicate that the goals that come earlier in the shallow review are more promising or have more possible subgoals. 

How to engage with this post

The full post can be found here. It contains many extensive quotes without added commentary from us, some of which are relevant to multiple sections and hence repeated. There’s a section for each potential goal/subgoal, and each section should make sense by itself. So readers should feel free to just skim, only look at the sections that are of interest to them, and skip repeated quotes.

Acknowledgements

Michael’s work on this post was supported by Rethink Priorities, though he ended up pivoting to other topics before having time to get this up to RP's usual standards. Will helped with the research and editing in a personal capacity. We’re grateful to Eva Siegmann, Matthew Gentzel, and Peter Wildeford for helpful feedback on earlier drafts. Mistakes are our/Michael’s own.

If you are interested in RP’s work, please visit our research database and subscribe to our newsletter. 

Click here to see the full post.

1. ^{^}

   By an intermediate goal, we mean a goal that (1) is more specific and directly actionable than a goal like “reduce nuclear risk”, (2) is of interest because advancing it might be one way to advance a higher-level goal like that, but (3) is less specific and directly actionable than a particular intervention (e.g., “advocate for the US and Russia to renew the INF Treaty”).

   I adopted the term “intermediate goal” from Muehlhauser (2020, 2021), who doesn’t provide a definition but does give examples that illustrate the concept. The definition proposed here is my own.

   Karnofsky (2022) also discusses a similar concept:

   How much should one value “transformative AI is first developed in country A” vs. “transformative AI is first developed in country B”, or “transformative AI is first developed by company A vs. company B”, or “transformative AI is developed 5 years sooner/later than it would have been otherwise?”

   If we were ready to make a bet on any particular intermediate outcome in this category being significantly net positive for the expected value of the long-run future, this could unlock a major push toward making that outcome more likely. I’d guess that many of these sorts of “intermediate outcomes” are such that one could spend billions of dollars productively toward increasing the odds of achieving them, but first one would want to feel that doing so was at least a somewhat robustly good bet.

2. ^{^}

   See also Philanthropy to the Right of Boom. Unfortunately we ran out of time to properly write up thoughts on the other categories of goals. Our very rough and incomplete notes on them can be found here.

3. ^{^}

   The area of nuclear risk seems more prone than many other areas to harmful and non-obvious results from well-intentioned and reasonable-sounding goals or interventions. This is due in large part to the important role deterrence plays. For example, limiting the likely scale and harms of nuclear war could weaken deterrence and might therefore make nuclear war more likely.

4. ^{^}

   In reality, “progress” comes in degrees rather than being a binary variable, which makes our statements about the effects and difficulty of “making progress” less meaningful and more fuzzy. A smaller issue is that the usefulness, harmfulness, resources required, and tractability of making progress on a goal could differ depending on how much progress has already been made (e.g., it could be easy and highly impactful to make initial progress, but hard and/or net-negative to get close to “fully achieving” the goal). 

5. ^{^}

   For this post, we divide “resources” and “support” into the following types: 

   -Funding

   -Labor from people with particular skills (e.g., people suited to being diplomats)

   -Advocacy from people with large followings (e.g., celebrities encouraging the public to contact politicians about some policy)

   -Advocacy from people with strong political connections (e.g., people who would be listened to by senior US or Russian officials)

   Of course, in reality, one could also subdivide those types or consider other types of resources. 

6. ^{^}

   By this we mean how easily additional resources could lead to additional progress, relative to what would be the case otherwise. So, in terms of the ITN framework, this accounts for both tractability and neglectedness.

7. ^{^}

   Really this would all depend a lot on the specific version of the goal in question one pursues (e.g., what states were focused on), the specific interventions one supports, the precise resources one can devote to the goal, etc. 

8. ^{^}

   Unfortunately this is based mostly on reasoning that we haven’t written up, rather than what we captured in this post. But for some relevant reasoning, you could see Philanthropy to the Right of Boom and/or our very rough draft of parts 2-4/4 of this shallow review.

9. ^{^}

   This is because the annual chance of nuclear conflict appears to be in the ballpark of 1%, and we would estimate the same annual chance for each year of the coming decades (because we’re very unsure whether it’ll rise, fall, or remain constant). Given that, the chance that there won’t be a single nuclear conflict by 2050 would be 1 - 0.99^28 = 24.5%.

   These estimates are informed by or align with the lines of evidence collected by Rodriguez (2019) and forecasts on the following set of questions on the site Metaculus: 1, 2, 3, 4, 5. Caveats include that many of those pieces of evidence are about outcomes more specific than “nuclear conflict” (suggesting they would be under-estimates), the different pieces of evidence suggest somewhat different chances, and each piece of evidence has notable limitations. But they collectively suggest that the annual chance is quite likely to be somewhere between 0.3% and 3%. 

   See also Michael’s draft Database of nuclear risk estimates [draft]. 

10. ^{^}

    For discussion, see 9 mistakes to avoid when thinking about nuclear risk (a rough draft); Different ways a nuclear conflict could start and play out; Toon et al., 2007; Reisner et al., 2018; Robock et al., 2019; Rodriguez, 2019.

11. ^{^}

    See Michael’s draft Database of nuclear risk estimates and (to compare with other risks) Michael’s earlier Database of existential risk estimates (or similar).

    Additional relevant commentary can be found in Aird (2020), Beckstead (2015), Ladish (2020), Rodriguez (2019), and Rodriguez (2020).

12. ^{^}

     We think pieces that discuss only a prioritized set of goals, are more comprehensive at the level of interventions, or go into depth on some particular goals or interventions are also useful and are complementary to this post. Pieces of those types which we appreciated - and which we’ve drawn on for this post - include Open Philanthropy (2015) and Nuclear Threat Initiative (2020).