Thanks for the comment. My probabilities sum to 165%, which would translate to me saying that I would expect, on average, the next paper to do 1.65 things from the list, which to me doesn't seem too crazy. I think that this DOES match my expectations. 

But I also think you make a good point. If the next paper comes out, does only one thing, and does it really well, I commit to not complaining too much. 

Some relevant papers to anyone spelunking around this post years later:

https://arxiv.org/abs/2403.05030
 https://arxiv.org/abs/2407.15549
 

Thanks, I think that these points are helpful and basically fair. Here is one thought, but I don't have any disagreements. 

Olah et al. 100% do a good job of noting what remains to be accomplished and that there is a lot more to do. But when people in the public or government get the misconception that mechanistic interpretability has been (or definitely will be) solved, we have to ask where this misconception came from. And I expect that claims like "Sparse autoencoders produce interpretable features for large models" contribute to this. 

Thanks for the comment. I think the experiments you mention are good (why I think the paper met 3), but I don't think that its competitiveness has been demonstrated (why I think the paper did not meet 6 or 10). I think there are two problems. 

First, is that it's under a streetlight. Ideally, there would be an experiment that began with a predetermined set of edits (e.g., one from Meng et al., 2022) and then used SAEs to perform them. 

Second, there's no baseline that SAE edits are compared to. There are lots of techniques from the editing, finetuning, steering, rep-E, data curation, etc. literatures that people use to make specific changes to models' behaviors. Ideally, we'd want SAEs to be competitive with them. Unfortunately,  good comparisons would be hard because using SAEs for editing models is a pretty unique method with lots of compute required upfront. This would make it non-straightforward to compare the difficulty of making different changes with different methods, but it does not obviate the need for baselines. 

Thanks for the useful post. There are a lot of things to like about this. Here are a few questions/comments.

First, I appreciate the transparency around this point. 

we advocate for what we see as the ‘minimal viable policy’ for creating a good AI ecosystem, and we will be open to feedback.

Second, I have a small question. Why the use of the word "testing" instead of "auditing"? In my experience, most of the conversations lately about this kind of thing revolve around the term "auditing." 

Third, I wanted to note that this post does not talk about access levels for testers and ask why this is the case. My thoughts here relate to this recent work. I would have been excited to see (1) a commitment to providing auditors with methodological details, documentation, and data or (2) a commitment to working on APIs and SREs that can allow for auditors to be securely given grey- and white-box access. 

Thanks. I agree that the points apply to individual researchers. But I don't think that it applies in a comparably worrisome way because individual researchers do not have comparable intelligence, money, and power compared to the labs. This is me stressing the "when put under great optimization pressure" of Goodhart's Law. Subtle misalignments are much less dangerous when there is a weak optimization force behind the proxy than when there is a strong one. 

See also this much older and closely related post by Thomas Woodside: Is EA an advanced, planning, strategically-aware power-seeking misaligned mesa-optimizer?

Thanks for the reply. I think this resolves the thread and puts us on the same page :)

I have been thinking a lot lately about evals and what differences black- vs. white-box access makes for evaluators. I was confused about the appendix on black-box evals. My best attempt to make sense of it hinges on two instances of the authors intending something different than what was written. 

First, the appendix hinges somewhat on this point. 

For now, we think the easiest way to build control techniques that avoid making incorrect assumptions about scheming models is to make conservative assumptions about their internals, by rejecting any strategy which can be evaded by a red-team-constructed model.

I agree with this, but saying that "control evals should be black-box" is a non sequitur. This is making the point that bad reverse-engineering of the model is bad and seems to assume that good reverse-engineering is hard/unlikely implicitly.

Second, even if good reverse-engineering is hard/unlikely, reverse-engineering is not the definition of a white-box eval. There are a lot of things that you can do with white-box access for evals other than interpretability-based things: gradient-based attacks, gradient-guided attacks, latent-space attacks, and finetuning -- and all of these things actually seem plausibly useful for control evals.  Plus I don't see why interp should be off limits for evaluators who know the limitations of the tools they are using. 

Ultimately, white-box attacks offer strictly more options for any type of eval compared to black-box ones. For any property that an AI system might have, if there are two teams of competent people trying to verify if the system has that property, all else equal, I'd bet on the success of the team that has more access. 

So if I understand things correctly, and if the intended point of this appendix is that "Interpretability-based evals for control seem unreliable and possibly could lead to pitfalls," I would recommend just saying that more directly. 

Thanks!

I intuit that what you mentioned as a feature might also be a bug. I think that practical forgetting/unlearning that might make us safer would probably involve subjects of expertise like biotech. And if so, then we would want benchmarks that measure a method's ability to forget/unlearn just the things key to that domain and nothing else. For example, if a method succeeds in unlearning biotech but makes the target LM also unlearn math and physics, then we should be concerned about that, and we probably want benchmarks to help us quantify that.

I could imagine an unlearning benchmark, for example, with $n$ textbooks and $n$ ap tests. Then for each of $k$ different knowledge-recovery strategies, one could construct the $n\times n$ grid of how well the model performs on each target test for each unlearning textbook.