by Jessica Taylor 606 days ago | Ryan Carey likes this | link | parent Is there a reason why this won’t just converge to maximizing human approval?

 by Stuart Armstrong 606 days ago | link My original idea was to identify actions that increased human approval but were specifically labelled as not legitimate, and having the AI generalise those as a priority, but the idea needs work. reply
 by Jessica Taylor 605 days ago | link The thing I’m pointing at is that the generalization you should expect by default is “things are legitimate iff a human would label them as legitimate”, which is a kind of approval. It seems like you need some kind of special sauce if you want the generalization to be something other than that. reply
 by Stuart Armstrong 603 days ago | Patrick LaVictoire likes this | link How about something like this? I don’t expect this to work as stated, but it may suggest certain possibilities: There is a familiarity score $$F$$ which labels how close a situation is to one where humans have full and rapid understanding of what’s going on. In situations of high $$F$$, the human reward signals are take as accurate. There are examples of situations of medium $$F$$ where humans, after careful deliberation, conclude that the reward signals were wrong. The prior is that for low $$F$$, there will be reward signals that are wrong but which even careful human deliberation cannot discern. The job of the learning algorithm is to deduce what these are by extending the results from medium $$F$$. This should not converge merely onto human approval, since human approval is explicitly modelled to be false here. reply
 by Jessica Taylor 603 days ago | link This seems pretty similar to this proposal, does that seem right to you? I think my main objection is the same as the main objection to the proposal I linked to: there has to be a good prior over “what the correct judgments are” such that when this prior is updated on data, it correctly generalizes to cases where we can’t get feedback even in principle. It’s not even clear what “correct judgments” means (you can’t put a human in a box and have them think for 500 years). reply
 by Stuart Armstrong 602 days ago | link No exactly that. What I’m trying to get at is that we know some of the features that failure would have (eg edge cases of utility maximalisation, seductive-seeming or seductively-presented answer), so we should be able to use that knowledge somehow. reply
 by Paul Christiano 605 days ago | link You might be able to sample pairs $$(a, f(a))$$ without being able to actually evaluate $$a \mapsto f(a)$$. reply
 by Jessica Taylor 603 days ago | link I’m not sure why this would change things? If it samples a bunch of $$(a, f(a))$$ pairs until one of them has a high $$f(a)$$ value, then that’s going to have the same effect as sampling a bunch of $$a$$ values until one of them has a high $$f(a)$$ value. reply
 by Paul Christiano 603 days ago | Jessica Taylor likes this | link I meant that I may be able to sample pairs from some attack distribution without being able to harden my function against the attack distribution. Suppose that I have a program $$\widetilde{f} \in [0, 1]$$ which implements my desired reward function, except that it has a bunch of vulnerabilities $$\widetilde{a}_i$$ on which it mistakenly outputs 1 (when it really should output 0). Suppose further that I am able to sample vulnerabilities roughly as effectively as my AI. Then I can sample vulnerabilities $$\widetilde{a}$$ and provide the pairs $$(\widetilde{a}, -1)$$ to train my reward function, along with a bunch of pairs $$(a, \widetilde{f}(a))$$ for actions $$a$$ produced by the agent. This doesn’t quite work as stated but you could imagine learning $$f$$ despite having no access to it. (This is very similar to adversarial training / red teams). reply

### NEW DISCUSSION POSTS

[Note: This comment is three
 by Ryan Carey on A brief note on factoring out certain variables | 0 likes

There should be a chat icon
 by Alex Mennen on Meta: IAFF vs LessWrong | 0 likes

Apparently "You must be
 by Jessica Taylor on Meta: IAFF vs LessWrong | 1 like

There is a replacement for
 by Alex Mennen on Meta: IAFF vs LessWrong | 1 like

Regarding the physical
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

I think that we should expect
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

I think I understand your
 by Jessica Taylor on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

This seems like a hack. The
 by Jessica Taylor on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

After thinking some more,
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

Yes, I think that we're
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

My intuition is that it must
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

To first approximation, a
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

Actually, I *am* including
 by Vadim Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

Yeah, when I went back and
 by Alex Appel on Optimal and Causal Counterfactual Worlds | 0 likes

> Well, we could give up on
 by Jessica Taylor on The Learning-Theoretic AI Alignment Research Agend... | 0 likes