I’ve missed one or more facts that link those threads, can you tell me which part of critch’s approach stands out to you in this context and what made it do so? I agree that his is some of my favorite work, but it’s not yet obvious to me that it actually checks all the boxes, in particular whether humans deploying ai will have values that directly contract usage of his insights. it also still isn’t clear to me whether anything in his work helps with is/ought.
I was referring to his promotion of political approaches, which is what this post discussed, and which Eliezer has recently said is the best hope for avoiding doom, even if he’s still very pessimistic about it.
His alignment work is a different question, and I don’t feel particularly qualified to weigh in on it.
I’ve missed one or more facts that link those threads, can you tell me which part of critch’s approach stands out to you in this context and what made it do so? I agree that his is some of my favorite work, but it’s not yet obvious to me that it actually checks all the boxes, in particular whether humans deploying ai will have values that directly contract usage of his insights. it also still isn’t clear to me whether anything in his work helps with is/ought.
I was referring to his promotion of political approaches, which is what this post discussed, and which Eliezer has recently said is the best hope for avoiding doom, even if he’s still very pessimistic about it.
His alignment work is a different question, and I don’t feel particularly qualified to weigh in on it.