All of Hyperion's Comments + Replies

There's also the possibility that a CCP AGI can only happen through being trained on Western data to some extent (i.e., the English language internet) because otherwise they can't scale data enough. This implies that it would probably be a "Marxism with Chinese characteristics [with American characteristics]" AI since it seems like that just raises the "alignment to CCP values" technical challenge difficulty a lot.

If so much effort is being focused into AI research capability, I'd actually expect modally Agent-3 to be better than typical OpenBrain employee but completely incapable of replacing almost all employees in other fields. "capabilities are spiky" is a clear current fact about frontier AI, but your scenario seems to underestimate it.

2Max Harms
This is a good point, and I think meshes with my point about lack of consensus about how powerful AIs are. "Sure, they're good at math and coding. But those are computer things, not real-world abilities."

I suppose I mean influence over politics, policy, or governance (this is very high level since these are all distinct and separable), rather than actually being political necessarily. I do think there are some common skills, but actually being a politician weighs so many other factors more heavily that the strategic skill is not selected on very strongly at all. Being a politician's advisor, on the other hand...

Yes, it's a special case, but importantly one that is not evaluated by Brier score or Manifold bucks.

I guess that's the main element I didn't mention: many people on this forum would suggest judging via predictive skill/forecasting success. I think this is an ok heuristic, but of course the long time horizons involved in many strategic questions makes it hard to judge (and Tetlock has documented the problems with forecasting over long time horizons where these questions matter most).

Mostly, the people I think of as having strong strategic skill are closely linked to some political influence (which implicitly requires this skill to effect change) such as a... (read more)

2Neel Nanda
Political influence seems a very different skill to me? Lots of very influential politicians have been very incompetent in other real world ways This is just a special case (and an unusually important one) of a good forecasting record, right?

Nice post! As someone who spends a lot of time in AI policy on strategic thought and talking to people who I think are amongst the best strategic thinkers on AI, I appreciated this piece and think you generally describe the skills pretty well.

However, you say "research" skill by default does not lead to strategic skill, which is very true, but this varies drastically depending on the type of research! Mechanistic interpretability, in fact, appears to me to be an example of a field which is so in the weeds empirical with good feedback loops, that it makes i... (read more)

5Neel Nanda
Interesting. Thanks for the list. That seemed like a pretty reasonable breakdown to me. I think mechanistic interpretability does train some of them in particular, two, three and maybe six. But I agree that things involve thinking about society and politics and power and economics etc as a whole do seem clearly more relevant. One major concern I have is that it's hard to judge skill in domains with worse feedback loops because there is not feedback on who is correct. I'm curious how confident you are in your assessment of who has good takes or is good in these fields, and how you determine this?

This is very impressive work, well done! Improving compute/training literacy of the community is very valuable IMO, since I have often thought that not knowing much of this leads to poorer conclusions.

Note that the MLPerf benchmark for GPT-3 is not on the full C4 dataset, it's on 0.4% of the C4 dataset.

See: https://twitter.com/abhi_venigalla/status/1673813863186452480?s=20

This is an intuition only based on speaking with researchers working on LLMs, but I think that OAI thinks that a model can simultaneously be good enough at next token prediction to assist with research but also be very very far away from being a powerful enough optimizer to realise that it is being optimized for a goal or that deception is an optimal strategy, since the latter two capabilities require much more optimization power. And that the default state of cutting edge LLMs for the next few years is to have GPT-3 levels of deception (essentially none) and graduate student levels of research assistant ability.

I don't think it's odd at all - even a terrible chess bot can outplay almost all humans. Because most humans haven't studied chess. MATH is a dataset of problems from high school competitions, which are well known to require a very limited set of math knowledge and be solveable by applying simple algorithms. 

I know chain of thought prompting well - it's not a way to lift a fundamental constraint, it just is a more efficient targeting of the weights which represent what you want in the model.

It really isn't hard. No new paradigms are required. The proo

... (read more)
3porby
I think you may underestimate the difficulty of the MATH dataset. It's not IMO-level, obviously, but from the original paper: Clearly this is not a rigorous evaluation of human ability, but the dataset is far from trivial. Even if it's not winning IMO golds yet, this level of capability is not something I would have expected to see managed by an AI that provably cannot multiply in one step (if you had asked me in 2015). {Edit: to further support that this level of performance on MATH was not obvious, this comes from the original paper: Further, I'd again point to the hypermind prediction market for a very glaring case of people thinking 50% in MATH was going to take more time than it actually did. I have a hard time accepting that this level of performance was actually expected without the benefit of hindsight.} It was not targeted at time complexity, but it unavoidably involves it and provides some evidence for its contribution. I disagree that I've offered no evidence- the arguments from complexity are solid, there is empirical research confirming the effect, and CoT points in a compelling direction.  I can understand if you find this part of the argument a bit less compelling. I'm deliberately avoiding details until I'm more confident that it's safe to talk about. (To be clear, I don't actually think I've got the Secret Keys to Dooming Humanity or something; I'm just trying to be sufficiently paranoid.) I would recommend making concrete predictions on the 1-10 year timescale about performance on these datasets (and on more difficult datasets).

I mean, to me all this indicates is that our conception of "difficult reasoning problems" is wrong and incorrectly linked to our conception of "intelligence". Like, it shouldn't be surprising that the LM can solve problems in text which are notoriously based around applying a short step by step algorithm, when it has many examples in the training set.

To me, this says that "just slightly improving our AI architectures to be less dumb" is incredibly hard, because the models that we would have previously expected to be able to solve trivial arithmetic problems if they could do other "harder" problems are unable to do that.

3porby
I'm not clear on why it wouldn't be surprising. The MATH dataset is not easy stuff for most humans. Yes, it's clear that the algorithm used in the cases where the language models succeeds must fit in constant time and so must be (in a computational sense) simple, but it's still outperforming a good chunk of humans. I can't ignore how odd that is. Perhaps human reasoning is uniquely limited in tasks similar to the MATH dataset, AI consuming it isn't that interesting, and there are no implications for other types of human reasoning, but that's a high complexity pill to swallow. I'd need to see some evidence to favor a hypothesis like that. 1. It was easily predictable beforehand that a transformer wouldn't do well at arithmetic (and all non-constant time algorithms), since transformers provably can't express it in one shot. Every bit of capability they have above what you'd expect from 'provably incapable of arithmetic' is what's worth at least a little bit of a brow-raise. 2. Moving to non-constant time architectures provably lifts a fundamental constraint, and is empirically shown to increase capability. (Chain of thought prompting does not entirely remove the limiter on the per-iteration expressible algorithms, but makes it more likely that each step is expressible. It's a half-step toward a more general architecture, and it works.) 3. It really isn't hard. No new paradigms are required. The proof of concepts are already implemented and work. It's more of a question of when one of the big companies decides it's worth poking with scale.

Mostly Discord servers in my experience: EleutherAI is a big well known one but there are others with high concentrations of top ML researchers.

I happened to be reading this post today, as Science has just published a story on a fabrication scandal regarding an influential paper on amyloid-β: https://www.science.org/content/article/potential-fabrication-research-images-threatens-key-theory-alzheimers-disease

I was wondering if this scandal changes the picture you described at all?

3AβMale
Not a ton. I'd also recommend this article, including the discussion in the comments by researchers in the field. A crucial distinction I'd emphasize which is almost always lost in popular discussions is that between the toxic amyloid oligomer hypothesis, that aggregates of amyloid beta are the main direct cause of neurodegeneration; and the ATN hypothesis I described in this thread, that amyloid pathology causes tau pathology and tau pathology causes neurodegeneration. The former is mainly what this research concerns and has been largely discredited in my opinion since approximately 2012; the latter has a mountain of evidence in favor as I've described, and that hasn't really changed now that it's turned out that one line of evidence for an importantly different hypothesis was fabricated.