The strategy-stealing assumption

Categorising the ways that the strategy-stealing assumption can fail:

  • Humans don't just care about acquiring flexible long-term influence, because
    • 4. They also want to stay alive.
    • 5 and 6. They want to stay in touch with the rest of the world without going insane.
    • 11. and also they just have a lot of other preferences.
    • (maybe Wei Dai's point about logical time also goes here)
  • It is intrinsically easier to gather flexible influence in pursuit of some goals, because
    • 1. It's easier to build AIs to pursue goals that are easy to check.
    • 3. It's easier to build institutions to pursue goals that are easy to check.
    • 9. It's easier to coordinate around simpler goals.
    • plus 4 and 5 insofar as some values require continuously surviving humans to know what to eventually spend resources on, and some don't.
    • plus 6 insofar as humans are otherwise an important part of the strategic environment, such that it's beneficial to have values that are easy-to-argue.
  • Jessica Taylor's argument require that the relevant games are zero sum. Since this isn't true in the real world:
    • 7. A threat of destroying value (e.g. by threatening extinction) could be used as a bargaining tool, with unpredictable outcomes.
    • ~8. Some groups actively wants other groups to have less resources, in which case they can try to reduce the total amount of resources more or less actively.
    • ~8. Smaller groups have less incentive to contribute to public goods (such as not increasing the probability of extinction), but benefit equally from larger groups' contributions, which may lead them to getting a disproportionate fraction of resources by defecting in public-goods games.
Covid 2/11: As Expected

Ah, you were talking about this article. Me and Daniel were saying that "Kolmogorov Complexity" never shows up in the linked ssc article (thinking that Zvi accidentally wrote "Kolmogorov Complexity" when he meant "Kolmogorov Complicity").

Covid 2/11: As Expected

I can't find it either. Could you quote or screenshot?

Imitative Generalisation (AKA 'Learning the Prior')

Starting with amplification as a baseline; am I correct to infer that imitative generalisation only boosts capabilities, and doesn't give you any additional safety properties?

My understanding: After going through the process of finding z, you'll have a z that's probably too large for the human to fully utilise on their own, so you'll want to use amplification or debate to access it (as well as to generally help the human reason). If we didn't have z, we could train an amplification/debate system on D' anyway, while allowing the human and AIs to browse through D for any information that they need. I don't see how the existence of z makes amplification or debate any more aligned, but it seems plausible that it could improve competitiveness a lot. Is that the intention?

Bonus question: Is the intention only to boost efficiency, or do you think that IA will fundamentally allow amplification to solve more problems? (Ie., solve more problems with non-ridiculous amounts of compute – I'd be happy to count an exponential speedup as the latter.)

OpenAI: "Scaling Laws for Transfer", Hernandez et al.

It's worth noting that their language model still uses BPEs, and as far as I can tell the encoding is completely optimised for English text rather than code (see section 2). It seems like this should make coding unusually hard compared to the pretraining task; but maybe make pretraining more useful, as the model needs time to figure out how the encoding works.

Lessons I've Learned from Self-Teaching

I'm really surprised at how big your cards are! When I did anki regularly, I remember getting a big ugh-feeling from cards much smaller than yours, just because there were so many things that I had to consciously recapitulate. It was also fairly common that I missed some little detail and had to choose between starting the whole card over from scratch (which is a big time sink since the card takes so much time at every repeat) or accept that I might never remember that detail.

I'm super curious about your experience of e.g. encountering the function question. Do you try to generate both an example and a formalism, or just the formalism? Do you consciously recite a definition in words, or check some feeling of remembering what the definition is, or mumble something in your mind about how a function is a set of ordered pairs? Is the domain/range-definitions just there as a reminder when you read it, or do you aim to remember them every time? Do you reset or accept if you forget to mention a detail?

Prediction can be Outer Aligned at Optimum

Cool, seems reasonable. Here are some minor responses: (perhaps unwisely, given that we're in a semantics labyrinth)

Evan's footnote-definition doesn't rule out malign priors unless we assume that the real world isn't a simulation

Idk, if the real world is a simulation made by malign simulators, I wouldn't say that an AI accurately predicting the world is falling prey to malign priors. I would probably want my AI to accurately predict the world I'm in even if it's simulated. The simulators control everything that happens anyway, so if they want our AIs to behave in some particular way, they can always just make them do that no matter what we do.

you are changing the definition of outer alignment if you think it assumes we aren't in a simulation

Fwiw, I think this is true for a definition that always assumes that we're outside a simulation, but I think it's in line with previous definitions to say that the AI should think we're not in a simulation iff we're not in a simulation. That's just stipulating unrealistically competetent prediction. Another way to look at it is that in the limit of infinite in-distribution data, an AI may well never be able to tell whether we're in the real world or in a simulation that's identical to the real world; but they would be able to tell whether we're in a simulation with simulators who actually intervene, because it would see them intervening somewhere in its infinite dataset. And that's the type of simulators that we care about. So definitions of outer alignment that appeal to infinite data automatically assumes that AIs would be able to tell the difference between worlds that are functionally like the real world, and worlds with intervening simulators.

And then, yeah, in practice I agree we won't be able to learn whether we're in a simulation or not, because we can't guarantee in-distribution data. So this is largely semantics. But I do think definitions like this end up being practically useful, because convincing the agent that it's not individually being simulated is already an inner alignment issue, for malign-prior-reasons, and this is very similar.

Prediction can be Outer Aligned at Optimum
Isn't that exactly the point of the universal prior is misaligned argument? The whole point of the argument is that this abstraction/specification (and related ones) is dangerous.


I guess your title made it sound like you were teaching us something new about prediction (as in, prediction can be outer aligned at optimum) when really you are just arguing that we should change the definition of outer-aligned-at-optimum, and your argument is that the current definition makes outer alignment too hard to achieve

I mean, it's true that I'm mostly just trying to clarify terminology. But I'm not necessarily trying to propose a new definition – I'm saying that the existing definition already implies that malign priors are an inner alignment problem, rather than than an issue with outer alignment. Evan's footnote requires the model to perform optimally on everything it actually encounters in the real world (rather than asking it to do as well as it can across the multiverse, given its training data); so that definition doesn't have a problem with malign priors. And as Richard notes here, common usage of "inner alignment" refers to any case where the model performs well on the training data but is misaligned during deployment, which definitely includes problems with malign priors. And per Rohin's comment on this post, apparently he already agrees that malign priors are an inner alignment problem.

Basically, the main point of the post is just that the 11 proposals post is wrong about mentioning malign priors as a problem with outer alignment. And then I attached 3 sections of musings that came up when trying to write that :)

Prediction can be Outer Aligned at Optimum

Things I believe about what sort of AI we want to build:

  • It would be kind of convenient if we had an AI that could help us do acausal trade. If assuming that it's not in a simulation would preclude an AI from doing acausal trade, that's a bit inconvenient. However, I don't think this matters for the discussion at hand, for reasons I describe in the final array of bullet points below.
  • Even if it did matter, I don't think that the ability to do acausal trade is a deal-breaker. If we had a corrigible, aligned, superintelligent AI that couldn't do acausal trade, we could ask it to scan our brains, then compete through any competitive period on Earth / in space, and eventually recreate us and give us enough time to figure out this acausal trade thing ourselves. Thus, for practical purposes, an AI that assumes it isn't in a simulation doesn't seem defective to me, even if that means it can't do acausal trade.

Things I believe about how to choose definitions:

  • When choosing how to define our terms, we should choose based on what abstractions are most useful for the task at hand. For the outer-alignment-at-optimum vs inner alignment distinction, we're trying to choose a definition of "optimal performance" such that we can separately:
    • Design an intent-aligned AI out of idealised training procedures that always yield "optimal performance" on some metric. If we successfully do this, we've solved outer alignment.
    • Figure out a training procedure that produces an AI that actually does very well on the chosen metric (sufficiently well to be aligned, even if it doesn't achieve absolute optimal performance). If we do this, we've solved inner alignment.

Things I believe about what these candidate definitions would imply:

  • For every AI-specification built with the abstraction "Given some finite training data D, the AI predicts the next data point X according to how common it is that X follows D across the multiverse", I think that AI is going to be misaligned (unless it's trained with data that we can't get our hands on, e.g. infinite in-distribution data), because of the standard universal-prior-is-misaligned-reasons. I think this holds true even if we're trying to predict humans like in IDA. Thus, this definition of "optimal performance" doesn't seem useful at all.
  • For AI-specification built with the abstraction "Given some finite training data D, the AI predicts the next data point X according to how common it is that X follows D on Earth if we aren't in a simulation", I think it probably is possible to build aligned AIs. Since it also doesn't seem impossible to train AIs to do something like this (ie we haven't just moved the impossibility to the inner alignment part of the problem), it seems like a pretty good definition of "optimal performance".
    • Surprisingly, I think it's even possible to build AIs that do assign some probability to being in a simulation out of this. E.g. we could train the AI via imitation learning to imitate me (Lukas). I assign a decent probability to being in a simulation, so a perfect Lukas-imitator would also assign a decent probability to being in a simulation. This is true even if the Lukas-imitator is just trying to imitate the real-world Lukas as opposed to the simulated Lukas, because real-world Lukas assigns some probability to being simulated, in his ignorance.
  • I'm also open to other definitions of "optimal performance". I just don't know any useful ones other than the ones I mention in the post.
Imitative Generalisation (AKA 'Learning the Prior')
We want to understand the future, based on our knowledge of the past. However, training a neural net on the past might not lead it to generalise well about the future. Instead, we can train a network to be a guide to reasoning about the future, by evaluating its outputs based on how well humans with access to it can reason about the future

I don't think this is right. I've put my proposed modifications in cursive:

We want to understand the future, based on our knowledge of the past. However, training a neural net on the past might not lead it to generalise well about the future. Instead, we can train a network to be a guide to reasoning about the future, by evaluating its outputs based on how well humans with access to it can reason about the past [we don't have ground-truth for the future, so we can't test how well humans can reason about it] and how well humans think it would generalise to the future. Then, we train a separate network to predict what humans with access to the previous network would predict about the future.

(It might be a good idea to share some parameters between the second and first network.)

Load More