Nice! I’m keen to see whether the winning methods are more like i) patches for current LLM deficiencies that the AI companies will likely iron out over time by default; or ii) general ways of differentially boosting AIs’ philosophical competence that we’d expect to still make a difference in 5 years. (I predict i, though I of course hope for ii.)
Will Aldred
Karma: 5,329
This is great!
(I got “0 conflicts, 6 bullets bitten: your answers do not settle cleanly onto one of the standard views of population ethics.” I think my biggest bullet bites are allowing incomparability and accepting the Procreation Asymmetry.)
Thanks for raising this. I agree the setup isn’t ideal, but the listed judges are the people who combine sufficiently good judgement + relevant expertise and—crucially—willingness to put in the time. (I did try to enlist other experts, but it didn’t come together.)
the jury is comprised of the author of the work to be criticised and their co-author
To clarify, the essay panel isn’t just the author plus a collaborator—it’s three people: Anthony (who wrote the sequence), Jesse (who collaborated on related work), and Andreas Mogensen, an Oxford philosopher who has previously worked on cluelessness but who had no involvement in the sequence and is independent of it. I’m not sure this fully dissolves the biases/incentives concern, but I hope it goes some way toward doing so.
Hang on, the category/example you cite is listed in the ‘Recommended use of LLMs’ section. So, I’m not sure what you’re disagreeing with?
Indeed, almost half the post is about distinguishing good from bad uses of LLMs, thus I’m struggling to make sense of your last paragraph. Are you referring to discussion (which demonizes all AI use for writing) that has happened elsewhere?
Nod. Plus, another advantage of your ‘consensus ASI’ approach—which is essentially a values handshake—over the deal types outlined by OP is that combined-US-China presents a unified front if and when third-party alien civilizations are encountered.
(A ‘unified front’ is an advantage if military power, and thus bargaining power, scales superlinearly in the deep future. Which seems >50% likely to me.)
my fave is @Duncan Sabien’s ‘Deadly by Default’
Good post, and I strongly agree. My preferred handle for what you’re pointing at is ‘integrity’. Quoting @Habryka (2019):
I think of integrity as a more advanced form of honesty […] Where honesty is the commitment to not speak direct falsehoods, integrity is the commitment to speak truths that actually ring true to yourself, not ones that are just abstractly defensible to other people. It is also a commitment to act on the truths that you do believe, and to communicate to others what your true beliefs are.
(In this frame, What We Owe the Future, for example, was honest but low integrity.)
There was some discussion of this a few months ago: see here.
Although, maybe your main point—which I agree the existing discussion doesn’t really have answers to—is, “How, if at all, should we be getting ahead of things and proactively setting a framing for the social media conversation that will surely follow (as opposed to just forming some hypotheses over what that conversation will be like, but not particularly doing anything yet)? Who within our community should lead these efforts? How high priority is this compared to other forms of improving the EA brand?[1]”
- ^
Fwiw, I personally mostly disagree with efforts to improve EA’s brand/reputation. I endorse embodying our principles and not focusing too much on how that is perceived; IMO, this is an indirect/long-game way of ~optimising for reputation. I’m generally suspicious of actors who try to control their reputations. (Will MacAskill said something similar here.)
- ^
[sorry I’m late to this thread]
@William_MacAskill, I’m curious which (if any) of the following is your position?
1.
“I agree with Wei that an approach of ‘point AI towards these problems’ and ‘listen to the AI-results that are being produced’ has a real (>10%? >50%?) chance of ending in moral catastrophe (because ‘aligned’ AIs will end up (unintentionally) corrupting human values or otherwise leading us into incorrect conclusions).
And if we were living in a sane world, then we’d pause AI development for decades, alongside probably engaging in human intelligence enhancement, in order to solve the deep metaethical and metaphilosophical problems at play here. However, our world isn’t sane, and an AI pause isn’t in the cards: the best we can do is to differentially advance AIs’ philosophical competence,[1] and hope that that’s enough to avoid said catastrophe.”
2.
“I don’t buy the argument that aligned AIs can unintentionally corrupt human values. Furthermore, I’m decently confident that my preferred metaethical theory (e.g., idealising subjectivism) is correct. If intent alignment goes well, then I expect a fairly simple procedure like ‘give everyone a slice of the light cone, within which they can do anything they want (modulo some obvious caveats), and facilitate moral trade’ will result in a near-best future.”
3.
“Maybe aligned AIs can unintentionally corrupt human values, but I don’t particularly think this matters since it won’t be average humans making the important decisions. My proposal is that we fully hand off questions re. what to do with the light cone to AIs (rather than have these AIs boost/amplify humans). And I don’t buy that there is a metaphilosophical problem here: If we can train AIs to be at least as good as the best human philosophers at the currently in-distribution ethical+philosophical problems, then I see no reason to think that these AIs will misgeneralise out of distribution any more than humans would. (There’s nothing special about the conclusions human philosophers would reach, and so even if the AIs reach different conclusions, I don’t see that as a problem. Indeed, if anything, the humans are more likely to make random mistakes, thus I’d trust the AIs’ conclusions more.)
(And then, practically, the AIs are much faster than humans, so they will make much more progress over the crucial crunch time months. Moreover, above we were comparing AIs to the best human philosophers / to a well-organised long reflection, but the actual humans calling the shots are far below that bar. For instance, I’d say that today’s Claude has better philosophical reasoning and better starting values than the US president, or Elon Musk, or the general public. All in all, best to hand off philosophical thinking to AIs.)”
@Wei Dai, I understand that your plan A is an AI pause (+ human intelligence enhancement). And I agree with you that this is the best course of action. Nonetheless, I’m interested in what you see as plan B: If we don’t get an AI pause, is there any version of ‘hand off these problems to AIs’/ ‘let ‘er rip’ that you feel optimistic about? or which you at least think will result in lower p(catastrophe) than other versions? If you have $1B to spend on AI labour during crunch time, what do you get the AIs to work on?
(I’m particularly interested in your plan B re. solving (meta)philosophy, since I’m exploring starting a grantmaking programme in this area. Although, I’m also interested if your answer goes in another direction.)
Possible cruxes:
Human-AI safety problems are >25% likely to be real and important
Giving the average person god-like powers (via an intent-aligned ASI) to reshape the universe and themselves is >25% likely to result in the universe becoming optimised for more-or-less random values—which isn’t too dissimilar to misaligned AI takeover
If we attempt to idealise a human-led reflection and hand if off to AIs, then the outcome will be at least as good as a human-led reflection (under various plausible (meta)ethical frameworks, including ones in which getting what humans-in-particular would reflectively endorse is important)
Sufficient (but not necessary) condition: advanced AIs can just perfectly simulate ideal human deliberation
‘Default’ likelihood of getting an AI pause
Tractability of pushing for an AI pause, including/specifically through trying to legiblise currently-illegible problems
Items that I don’t think should be cruxes for the present discussion, but which might be causing us to talk past each other:
In practice, human-led reflection might be kinda rushed and very far away from an ideal long reflection
For the most important decisions happening in crunch time, either it’s ASIs making the decisions, or non-reflective and not-very-smart humans
Political leaders often make bad decisions, and this is likely to get worse when the issues become more complicated (if they’re not leveraging advanced AI)
An advanced AI could be much better than any current political leader along all the character traits we’d want in a political leader (e.g., honesty, non-self-interest, policymaking capability)
- ^
“For instance, via my currently-unpublished ‘AI for philosophical progress’ and ‘Guarding against mind viruses’ proposals.”
Nice!
One quibble: IMO, the most important argument within ‘economic dominance,’ which doesn’t appear in your list (nor really in the body of your text), is Wei Dai’s ‘AGI will drastically increase economies of scale’.
Mod here. It looks like this thread has devolved into a personal dispute with only tangential relevance to EA. I’m therefore locking the thread.
Those involved, please don’t try to resurrect the dispute elsewhere on this forum; we may issue bans if we see that happening.
Richard’s ‘Coercion is an adaptation to scarcity’ post and follow-up comment talk about this (though ofc maybe there’s more to Richard’s view than what’s discussed there). Select quotes:
What if you think, like I do, that we live at the hinge of history, and our actions could have major effects on the far future—and in particular that there’s a significant possibility of existential risk from AGI? I agree that this puts us in more of a position of scarcity and danger than we otherwise would be (although I disagree with those who have very high credence in catastrophe). But the more complex the problems we face, the more counterproductive scarcity mindset is. In particular, AGI safety requires creative paradigm-shifting research, and large-scale coordination; those are both hard to achieve from a scarcity mindset. In other words, coercion at a psychological or community level has strongly diminishing marginal returns when dealing with scarcity at a civilizational level.
…
AI is a danger on a civilizational level; but the best way to deal with danger on a civilizational level is via cultivating abundance at the level of your own community, since that’s the only way it’ll be able to make a difference at that higher level.
FYI readers, here is Habryka’s response to this post over on LessWrong, if you haven’t seen it.
Relatedly, @MichaelDickens shallow-reviewed Horizon just under a year ago—see here.[1] Tl;dr: Michael finds that Horizon’s work isn’t very relevant to x-risk reduction; Michael believes Horizon is net-negative for the world (credence: 55%).
(On the other hand, it was Eth, Perez and Greenblatt—i.e., people whose judgement I respect—who recommended donating to Horizon in that post Mikhail originally commented on. So, I overall feel unsure about what to think.)
Fyi, the Forum team has experimented with LLMs for tagging posts (and for automating some other tasks, like reviewing new users), but so far none have been accurate enough to rely on. Nonetheless, I appreciate your comment, since we weren’t really tracking the transparency/auditing upside of using LLMs.
Beyond the specifics (which Vasco goes into in his reply): These tweets are clearly not serious/principled/good-faith criticisms. If we are constantly moderating what we say to try to make sure that we don’t possibly give trolls any ammunition, then our discourse is forever at the mercy of those most hostile to the idea of doing good better. That’s not a good situation to be in. Far better, I say, to ignore the trolling.
I agree with ‘within dedicated discussions and not on every single animal welfare post,’ and I think Vasco should probably take note, here.
However, I’m not really sure what you mean by reputational risk—whose reputation is at risk?
Generally speaking, I very much want people to be saying what they honestly believe, both on this forum and elsewhere. Vasco honestly believes that soil animal welfare outweighs farmed animal welfare, and he has considered arguments for why he believes this, and so I think it’s valuable for him to say the things he says (so long as he’s not being spammy about it). If people are constantly self-censoring out of fear of reputational risks, and the like, then it’s ~impossible for us to collectively figure out what’s true, and we will thus fail to rectify moral atrocities.
And, like, the core of Vasco’s argument—that if soil animals are conscious, then, given how numerous they are, their total moral weight must be very high—is really quite straightforward. I’m skeptical that readers will go away feeling confused, or thinking that Vasco (and, by extension, animal welfare folks in general?) is crazy, such that they somehow end up caring less about animals?
Just want to quickly flag that you seem to have far more faith in superforecasters’ long-range predictions than do most people who have worked full-time in forecasting, such as myself.
@MichaelDickens’ ‘Is It So Much to Ask?’ is the best public writeup I’ve seen on this (specifically, on the problems with Metaculus’ and FRI XPT’s x-risk/extinction forecasts, which are cited in the main post above). I also very much agree with:
Excellent forecasters and Superforecasters™ have an imperfect fit for long-term questions
Here are some reasons why we might expect longer-term predictions to be more difficult:
No fast feedback loops for long-term questions. You can’t get that many predict/check/improve cycles, because questions many years into the future, tautologically, take many years to resolve. There are shortcuts, like this past-casting app, but they are imperfect.
It’s possible that short-term forecasters might acquire habits and intuitions that are good for forecasting short-term events, but bad for forecasting longer-term outcomes. For example, “things will change more slowly than you think” is a good heuristic to acquire for short-term predictions, but might be a bad heuristic for longer-term predictions, in the same sense that “people overestimate what they can do in a week, but underestimate what they can do in ten years”. This might be particularly insidious to the extent that forecasters acquire intuitions which they can see are useful, but can’t tell where they come from. In general, it seems unclear to what extent short-term forecasting skills would generalize to skill at longer-term predictions.
“Predict no change” in particular might do well, until it doesn’t. Consider a world which has a 2% probability of seeing a worldwide pandemic, or some other large catastrophe. Then on average it will take 50 years for one to occur. But at that point, those predicting a 2% will have a poorer track record compared to those who are predicting a ~0%.
In general, we have been in a period of comparative technological stagnation, and forecasters might be adapted to that, in the same way that e.g., startups adapted to low interest rates.
Sub-sampling artifacts within good short-term forecasters are tricky. For example, my forecasting group Samotsvety is relatively bullish on transformative technological change from AI, whereas the Forecasting Research Institute’s pick of forecasters for their existential risk survey was more bearish.
How much weight should we give to these aggregates?
My personal tier list for how much weight I give to AI x-risk forecasts to the extent I defer:
Individual forecasts from people who seem to generally have great judgment, and have spent a ton of time thinking about AI x-risk forecasting e.g. Cotra, Carlsmith
Samotsvety aggregates presented here
A superforecaster aggregate (I’m biased re: quality of Samotsvety vs. superforecasters, but I’m pretty confident based on personal experience)
Individual forecasts from AI domain experts who seem to generally have great judgment, but haven’t spent a ton of time thinking about AI x-risk forecasting (this is the one I’m most uncertain about, could see anywhere from 2-4)
Everything else I can think of I would give little weight to.[1][2]
Separately, I think you’re wrong about UK AISI not putting much credence on extinction scenarios? I’ve seen job adverts from AISI that talk about loss of control risk (i.e., AI takeover), and I know people working at AISI who—last I spoke to them—put ≫10% on extinction.
- ^
Why do I give little weight to Metaculus’s views on AI? Primarily because of the incentives to make very shallow forecasts on a ton of questions (e.g. probably <20% of Metaculus AI forecasters have done the equivalent work of reading the Carlsmith report), and secondarily that forecasts aren’t aggregated from a select group of high performers but instead from anyone who wants to make an account and predict on that question.
- ^
Why do I give little weight to AI expert surveys such as When Will AI Exceed Human Performance? Evidence from AI Experts? I think most AI experts have incoherent and poor views on this because they don’t think of it as their job to spend time thinking and forecasting about what will happen with very powerful AI, and many don’t have great judgment.
There’s an old (2006) Bostrom paper on ~this topic, as well as Yudkowsky’s ‘Anthropic Trilemma’ (2009) and Wei Dai’s ‘Moral Status of Independent Identical Copies’ (2009). Perhaps you’re remembering one of them?
(Bostrom disagrees with the second paragraph you cite, as far as I can tell. He writes: ‘If a brain is duplicated so that there are two brains in identical states, are there then two numerically distinct phenomenal experiences or only one? There are two, I argue.’)
Yes, I prefer my 6 bullets to the repugnant conclusion, although of course population ethics is tricky stuff and I expect that idealised-Willi[1] would scoff at current-Will’s naive thinking.
Partly, what’s going on is I’m skeptical there’s such a thing as positive welfare, which makes the quiz hard to answer.[2] Beyond that, I’m partial to lexicality, and I believe person-moments are the relevant unit of analysis, not lives. If all those lives in Z-world have ups and downs, then that implies a huge number of suffering person-moments—which I find horrifying.
Subscript i because there’s no unique way to idealise. (Which raises its own problems. But that’s beside the point here.)
Here’s my answer sheet, fwiw.