Effective Altruism Forum
EA Forum

Hide table of contents
by Joe_Carlsmith
6 min read 3

114

AI safetyExistential riskForecastingAI forecasting
Frontpage

(Cross-posted from my website.)

Good Judgment has solicited reviews and forecasts from superforecasters regarding my report “Is power-seeking AI an existential risk,” along with forecasts on three additional questions regarding timelines to AGI, and on one regarding the probability of existential catastrophe from out-of-control AGI. 

A summary of the results is available on the Good Judgment website here, as are links to the individual reviews. Good Judgment has also prepared more detailed summaries of superforecaster comments and forecasts here (re: my report) and here (re: the other timelines and X-risk questions). I’ve copied key graphics below, along with a screenshot of a public spreadsheet of the probabilities from each forecaster and a link to their individual review (where available).[1]

This project was funded by Open Philanthropy, my employer.[2] The superforecasters completed a survey very similar to the one completed by other reviewers of my report (see here for links), except with an additional question (see footnote) about the “multiple stage fallacy.”[3]

Relative to my original report, the May 2023 superforecaster aggregation places higher probabilities on the first three premises – Timelines (80% relative to my 65%), Incentives (90% relative to my 80%), and Alignment Difficulty (58% relative to my 40%) – but substantially lower probabilities on the last three premises – High-Impact Failures (25% relative to my 65%), Disempowerment (5% relative to my 40%), Catastrophe (40% relative to my 95%). And their overall probability on all the premises being true – that is, roughly, on existential catastrophe from power-seeking AI by 2070 – is 1% compared to my 5% in the original report.[4] (Though in the supplemental questions included in the second part of the project, they give a 6% probability to existential catastrophe from out-of-control AGI by 2200, conditional on AGI by 2070; and a 40% chance of AGI by 2070.[5])

To the extent the superforecasters and I disagree, especially re: the overall probability of existential risk from power-seeking AI, I haven’t updated heavily in their direction, at least thus far (though I have updated somewhat).[6] This is centrally because:

  • My engagement thus far with the written arguments in the reviews (which I encourage folks to check out – see links in the spreadsheet) hasn’t moved me much.[7] 
  • I remain unsure how much to defer to raw superforecaster numbers (especially for longer-term questions where their track-record is less proven) absent object-level arguments I find persuasive.[8] 
  • I was pricing in some amount of "I think that AI risk is higher than do many other thoughtful people who've thought about it at least somewhat" already.

In this sense, perhaps, I am similar to some of the domain experts in the Existential Risk Persuasion Tournament, who continued to disagree significantly with superforecasters about the probability of various extinction events even after arguing about it. However, I think it’s an interesting question how to update in the face of disagreement of this type (see e.g. Alexander here for some reflections), and I’d like to think more about it.[9]

Thanks to Good Judgment for conducting this project, and to the superforecaster reviewers for their participation. If you’re interested in more examples of superforecasters weighing in on existential risks, I encourage you to check out the Existential Risk Persuasion Tournament (conducted by the Forecasting Research Institute) as well.

Screenshot of spreadsheet with the probabilities from each forecaster.

  1. ^

    After discussion with Good Judgment, I’ve made a few small adjustments, in the public spreadsheet, to the presentation of the data they originally sent to me. In particular, Superforecaster AI Expert #10 gave two different forecasts based on two different definitions of “strategic awareness,” which were distorting the median in the spreadsheet, so I have averaged them together. And the initial data listed Non-expert #8 (review here) as giving forecasts of “.01%” on various questions where I felt that the text of the review warranted different answers (namely, “No answer,” “Less than .5%”, “<.5% or 5%,” and “Well below .5%).   

    Note that the median in the spreadsheet differs from the “May 2023 aggregation” (and from the headline numbers on the website). This is because the individual forecaster numbers in the spreadsheet were initially produced by Superforecasters in isolation, after which point all the questions were posted on a platform where superforecasters could act as a team, challenge/question each other's views, and make more updates (plus, I’m told, other superforecasters added their judgments). The questions were then re-opened in May 2023 for a final updating round, and the May 2023 aggregation is the median of the last forecast made by everyone at that time.

  2. ^

    It was initially instigated by the FTX Future Fund back in 2022, but Open Philanthropy took over funding after FTX collapsed.

  3. ^

    The new question (included in the final section of the survey) was:  "One concern about the estimation method in the report is that the multi-premise structure biases towards lower numbers (this is sometimes called the "multi-stage fallacy"; see also Soares here). For example, forecasters might fail to adequately condition on all of the previous premises being true and to account for their correlations, or they might be biased away from assigning suitably extreme probabilities to individual premises.  When you multiply through your probabilities on the individual premises, does your estimate differ significantly from the probability you would've given to "existential catastrophe by 2070 from worlds where all of 1-6 are true" when estimating it directly? If so, in what direction?"

  1. ^

    Note that this overall superforecaster probability differs somewhat from what you get if you just multiply through the superforecaster median for each premise. If you do that, the superforecaster numbers imply a ~10% probability that misaligned, power-seeking AI systems will cause at least a trillion dollars of damage by 2070, a ~.5% probability that they will cause full human disempowerment, and a ~.2% probability that they will cause existential catastrophe.

  2. ^

    Technically, the 40% on AGI by 2070 and the 80% on APS-AI systems by 2070 are compatible, given that the two thresholds are defined differently. Though whether the differences warrant this large of a gap is a further question.

  3. ^

    More specifically: partly as a result of this project, and partly as a result of other projects like the Existential Risk Persuasion Tournament (conducted by the Forecasting Research Institute), I now think of it as a data-point that “superforecasters as a whole generally come to lower numbers than I do on AI risk, even after engaging in some depth with the arguments.” And this, along with other sources of “outside view” evidence (for example: the evidence provided by markets, and by the expectations of academic economists, about the probability of explosive near-term growth from advanced AI), has made me somewhat more skeptical of my inside-view take. Indeed, to the extent I would’ve updated towards greater confidence in this take (or towards even higher numbers on p(doom)) had the superforecasters given higher numbers on doom, Bayesianism would mandate that I update at least somewhat downwards given that they didn’t. (Though the strength of the required update depends, also, on the probability I would’ve placed on “the superforecasters come to lower numbers than I do” vs. “the superforecasters come to similar/higher numbers than I do” prior to seeing the results. I didn’t explicitly forecast this ahead of time, but I think I would’ve viewed the former as more likely.) 

  4. ^

    The final premise -- whether permanent disempowerment of ~all humans at the hands of power-seeking AIs constitutes an existential catastrophe (the aggregated superforecaster median here is 40%) -- also strikes me as substantially a matter of philosophy/ethics rather than empirical forecasting. This makes me additionally disinclined to defer. 

  5. ^

    Though I do think that a lot of the game is in “how much weight do you give to various considerations” rather than “what considerations are in play.” The former can be harder to argue about, but it may be the place where successful forecasters have the most advantage. This is related to the more general question of how much of the signal provided by superforecasters (and especially: of an aggregated set of superforecasters) to expect to be evident in the written arguments individual forecasters offer for their numbers.

  6. ^

    I think of it as similar to the question of how much to update on the fact that markets (and also: academic economists) generally do not seem to be expecting extreme, AI-driven economic growth with the same probability I do.

Show all footnotes

114

0
0
1

Reactions

0
0
1

More posts like this

Comments3


Sorted by
New & upvoted
Click to highlight new comments since:
David Mathers🔸27
9
2

Maybe there's just nothing interesting to say (though I doubt it), but I really feel like this should be getting more attention. It's an (at least mostly, plausible some of the supers were EAs) outside check on the views of most big EA orgs about the single best thing to spend EA resources on.

Reply
SummaryBot1
1
4

Executive summary: Superforecasters' aggregated probabilities differ from the author's original report on the premises of power-seeking AI as an existential risk, with higher probabilities on some premises and lower probabilities on others.

Key points:

  1. Superforecasters rated premises on AI timelines and incentives higher but alignment difficulty and impact of failures lower than the original report.
  2. The overall probability of existential catastrophe by 2070 was 1% versus 5% originally.
  3. The author has not substantially updated given unpersuasive arguments and uncertainty about deferring to group probabilities.
  4. Engaging with reasoned disagreements is an open question in forecasting methodology.
  5. The author sees this as similar to determining how much to update based on disagreements from markets and economists.

 

This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.

Reply
Paul_Christiano4
0
0

I think the "alignment difficulty" premise was given higher probability by superforecasters, not lower probability.

Reply
More from Joe_Carlsmith
75
Fake thinking and real thinking
Joe_Carlsmith
·
3
3
34
AI for AI safety
Joe_Carlsmith
·
1
1
72
Takes on "Alignment Faking in Large Language Models"
Joe_Carlsmith
·
1
1
View more
Curated and popular this week
 ·  · 5m read
 · 
[Cross-posted from my Substack here] If you spend time with people trying to change the world, you’ll come to an interesting conundrum: Various advocacy groups reference previous successful social movements as to why their chosen strategy is the most important one. Yet, these groups often follow wildly different strategies from each other to achieve social change. So, which one of them is right? The answer is all of them and none of them. This is because many people use research and historical movements to justify their pre-existing beliefs about how social change happens. Simply, you can find a case study to fit most plausible theories of how social change happens. For example, the groups might say: * Repeated nonviolent disruption is the key to social change, citing the Freedom Riders from the civil rights Movement or Act Up! from the gay rights movement. * Technological progress is what drives improvements in the human condition if you consider the development of the contraceptive pill funded by Katharine McCormick. * Organising and base-building is how change happens, as inspired by Ella Baker, the NAACP or Cesar Chavez from the United Workers Movement. * Insider advocacy is the real secret of social movements – look no further than how influential the Leadership Conference on Civil Rights was in passing the Civil Rights Acts of 1960 & 1964. * Democratic participation is the backbone of social change – just look at how Ireland lifted a ban on abortion via a Citizen’s Assembly. * And so on… To paint this picture, we can see this in action below: Source: Just Stop Oil which focuses on…civil resistance and disruption Source: The Civic Power Fund which focuses on… local organising What do we take away from all this? In my mind, a few key things: 1. Many different approaches have worked in changing the world so we should be humble and not assume we are doing The Most Important Thing 2. The case studies we focus on are likely confirmation bias, where
 ·  · 2m read
 · 
I speak to many entrepreneurial people trying to do a large amount of good by starting a nonprofit organisation. I think this is often an error for four main reasons. 1. Scalability 2. Capital counterfactuals 3. Standards 4. Learning potential 5. Earning to give potential These arguments are most applicable to starting high-growth organisations, such as startups.[1] Scalability There is a lot of capital available for startups, and established mechanisms exist to continue raising funds if the ROI appears high. It seems extremely difficult to operate a nonprofit with a budget of more than $30M per year (e.g., with approximately 150 people), but this is not particularly unusual for for-profit organisations. Capital Counterfactuals I generally believe that value-aligned funders are spending their money reasonably well, while for-profit investors are spending theirs extremely poorly (on altruistic grounds). If you can redirect that funding towards high-altruism value work, you could potentially create a much larger delta between your use of funding and the counterfactual of someone else receiving those funds. You also won’t be reliant on constantly convincing donors to give you money, once you’re generating revenue. Standards Nonprofits have significantly weaker feedback mechanisms compared to for-profits. They are often difficult to evaluate and lack a natural kill function. Few people are going to complain that you provided bad service when it didn’t cost them anything. Most nonprofits are not very ambitious, despite having large moral ambitions. It’s challenging to find talented people willing to accept a substantial pay cut to work with you. For-profits are considerably more likely to create something that people actually want. Learning Potential Most people should be trying to put themselves in a better position to do useful work later on. People often report learning a great deal from working at high-growth companies, building interesting connection
 ·  · 31m read
 · 
James Özden and Sam Glover at Social Change Lab wrote a literature review on protest outcomes[1] as part of a broader investigation[2] on protest effectiveness. The report covers multiple lines of evidence and addresses many relevant questions, but does not say much about the methodological quality of the research. So that's what I'm going to do today. I reviewed the evidence on protest outcomes, focusing only on the highest-quality research, to answer two questions: 1. Do protests work? 2. Are Social Change Lab's conclusions consistent with the highest-quality evidence? Here's what I found: Do protests work? Highly likely (credence: 90%) in certain contexts, although it's unclear how well the results generalize. [More] Are Social Change Lab's conclusions consistent with the highest-quality evidence? Yes—the report's core claims are well-supported, although it overstates the strength of some of the evidence. [More] Cross-posted from my website. Introduction This article serves two purposes: First, it analyzes the evidence on protest outcomes. Second, it critically reviews the Social Change Lab literature review. Social Change Lab is not the only group that has reviewed protest effectiveness. I was able to find four literature reviews: 1. Animal Charity Evaluators (2018), Protest Intervention Report. 2. Orazani et al. (2021), Social movement strategy (nonviolent vs. violent) and the garnering of third-party support: A meta-analysis. 3. Social Change Lab – Ozden & Glover (2022), Literature Review: Protest Outcomes. 4. Shuman et al. (2024), When Are Social Protests Effective? The Animal Charity Evaluators review did not include many studies, and did not cite any natural experiments (only one had been published as of 2018). Orazani et al. (2021)[3] is a nice meta-analysis—it finds that when you show people news articles about nonviolent protests, they are more likely to express support for the protesters' cause. But what people say in a lab setting mig
Recent opportunities in AI safety
170
Forethought: A new AI macrostrategy group
5 authors
· · 4m read
6
6
122
AI Moral Alignment: The Most Important Goal of Our Generation
Ronen Bar
· · 10m read
31
31
44
Center on Long-Term Risk: Summer Research Fellowship 2025
Center on Long-Term Risk
· · 1m read
0
0
View more