A robust alignment scheme would likely be trivial to transform into an AGI recipe.

Perhaps if you did have the full solution, but it feels like that there are some things of a solution that you could figure out, such that that part of the solution doesn't tell you as much about the other parts of the solution.

And it also feels like there could be a book such that if you read it you would gain a lot of knowledge about how to align AIs without knowing that much more about how to build one. E.g. a theoretical solution to the stop button problem seems like it would not tell you that much about how to build an AGI compared to figuring out how to properly learn a world model of Minecraft. And knowing how to build a world model of minecraft probably helps a lot with solving the stop button problem, but it doesn't just trivially yield a solution.

If you had a system with “ENTITY 92852384 implies ENTITY 8593483" it would be a lot of progress, as currently in neural networks we don't even understand the interal structures.

I want to have an algorithm that creates a world model. The world is large. A world model is uninterpretable by default through it's sheer size, even if you had interpretable but low level abels. By default we don't get any interpretable labels. I think there are ways to have generic dataprocessing procedures that don't talk about the human mind at all, that would yield more interpretable world model. Similar to how you could probably specify some very general property about python programs, such that that program becomes easier to understand by humans. E.g. a formalism of what it means that the control flow is straightforward: Don't use goto in C.

But even if you wouldn't have this, understanding the system still allows you to understand what the structure of the knowledge would be. It seems plausible that one could simply by understanding the system very well, make it such that the learned datastrucutres need to take particular shapes, such that these shapes correspond some relevant alignment properties.

In any case, it seems that this is a problem that any possible way to build an intelligence runs into? So I don't think it is a case against the project. When building an AI with NN you might not even think about that the interal representations might be wierd and alien (even for an LLM trained on human text)^[1], but the same problem persists.

1. ^{^}

   I haven't looked into this, or thought about at all, though that's what I expect.

You Need a Research Log

I definitely very often run into the problem that I forget why something was good to do in the first place. What are the important bits? Often I get sidetracked, and then the thing that I am doing seems not so got, so I stop and do something completely different. But then later on I realize that actually the original reason that led me down the path was good and that it would have been better to only backtrack a bit to the important piece. But often I just don't remember the important piece in the moment.

E.g. I think that having some kind of linking structure in your world model, that links objects in the model to the real world is important such that you can travel backward on the links to identify where exactly in your world model the error is. Then I go off and construct some formalism for a bit, but before I got to the point of adding the links I forgot that that was the original motivation, and so I just analyzed the model for a couple of hours before realizing that I still haven't added the linking structure. So it even happens during the same research session for me if I am not careful. And if you want to continue the next day, or a week later, having organized your thoughts in a way that isn't so painful to go through that you won't do it is extremely helpful.

I recognized a couple of things as important so far for being able to do it correctly:

1. Make it fun to make the notes. If you can't make this information processing activity fun you basically can't do it.

• My brain somehow seems to like doing it much more when I put all the notes on a website.
• Also taking lots of ADHD medication helps.

2. Make the notes high quality enough such that they are readable, instead of a wall of garbage text.

• Writing thoughts mainly on a whiteboard, and analog journals (including reflection) seems to help a lot (in general actually).

3. Integrate note-taking tightly into your research workflow.

• Don't rely on postprocessing, i.e. having a separate step of producing research notes. At least I didn't manage to get this to work at all so far. As much as possible make the content you produce in the first place as good as possible (analog tools help a lot with this). That means writing up notes and reflections as you are working, not at some time later (which never actually comes).

The interpretability safety case section seems to make the rather strong claim that the activation of known sabotage-related features is necessary for sabotage to occur (emphasis mine):

[C1.1] The model will be unable to accomplish organizational sabotage (T1) or to undermine oversight (T3) without triggering our feature-based monitoring (M1.1) and a detailed analysis of the session.

Afaict, the way that they establish this link (that a feature activation is necessary for a behavior to occur) is through ablation studies. Basically, if deactivating a feature prevents a model organism from committing a dangerous behavior, then they conclude that the feature causes the behavior. And although it does seem like this sort of test will generate Bayesian evidence about which features contribute to sabotage-capacity, I’m confused why the author thinks this would generate strong Bayesian evidence, much less identify the necessity conditions for sabotage with ~certainty?

Like, it seems plausible to me that some other unknown combination of features might create sabotage-like behavior, or that some relevant features were missed, or that there were other, ill-described-as-features mechanisms which can cause sabotage, etc. And my guess is that the evidence this sort of testing does generate will become weaker over time—i.e., once our safety precautions start to really matter—since more intelligent systems will presumably be able to cause sabotage in more numerous and creative ways.

Some of these potential setbacks are mentioned, and I get that this is just a preliminary sketch. But still, I’m confused where the confidence is coming from—why does the author expect this process to yield high confidence of safety?

Norvid on Twitter made the apt point that we will need to see the actual private data before we can really judge. Not unusual for lucky people to backrationalize their luck as a sure win.

you don't think it's "snobbish or discriminatory" to pretend it's something more because you count yourself among its users?

Fair point. I had already provided special justification for it but I agree with your reasoning, so I'll leave it out. Thanks for the example.

I think number of neurons in neocortex (or even more prefrontal cortex - but unfortunately i didn't quickly find how big the orca prefrontal cortex is - though I'd guess it to still be significantly bigger than for humans) is a much much better proxy for intelligence of species than brain size (or encephalization quotient). (E.g. see the wikipedia list linked in my question here [LW · GW].)

(Also see here [LW(p) · GW(p)]. There are more examples, e.g. a Blue and yellow macaw has 1.9 billion, whereas brown bears have only 250million.)

EDIT: Tbc I do think that larger bodies require more neurons in touch-sense and motor parts of the neocortex, so there is some effect of how larger animals need a bit larger brains to be similarly smart, but I don't think this effect is very strong.

But yeah there are other considerations too, which is why I am only at 50% that orcas could do science significantly better than humans if they tried.

On one hand, I feel a bit skeptical that some dude outperformed approximately every other pollster and analyst by having a correct inside-view belief about how existing pollster were messing up, especially given that he won't share the surveys.  On the other hand, this sort of result is straightforwardly predicted by Inadequate Equilibria, where an entire industry had the affordance to be arbitrarily deficient in what most people would think was their primary value-add, because they had no incentive to accuracy (skin in the game), and as soon as someone with an edge could make outsized returns on it (via real-money prediction markets), they outperformed all the experts.

On net I think I'm still <50% that he had a correct belief about the size of Trump's advantage that was justified by the evidence he had available to him, but even being directionally-correct would have been sufficient to get outsized returns a lot of the time, so at that point I'm quibbling with his bet sizing rather than the direction of the bet.

Fwiw I didn't find the post hostile. 

I'm assuming "natural abstraction" is also a scalar property. Reading this paragraph, I refactored the concept in my mind to "some abstractions tend to be cheaper to abstract than others. agents will converge to using cheaper abstractions. Many cheapness properties generalize reasonably well across agents/observation-systems/environments, but, all of those could in theory come apart."

And the Strong NAH would be "cheap-to-abstract-ness will be very punctuated, or something" (i.e. you might expect less of a smooth gradient of cheapnesses across abstractions)