Posts
Comments
I would say that most of current published interpretability is very, very bad and sucks at its job.
I do have an overall belief that making interpretability that does not suck, is actually quite feasible, and that there's no particular reason to believe that it's going to be particularly difficult or take particularly long.
Many people have spent a lot of effort trying to make progress on interpretability. I wish you could find a way to express your opinion that is a little more respectful of their work.
It's possible that the notion of "emergence" arose as a reaction to a hard-core positivist view that there is no way for us to understand, say, biology, until we can deduce it from the behavior of individual quarks. So, possibly, the notion of "emergence" may have been invented just to say that it's actually ok to study biology even if you don't entirely understand how quarks combine into a mitochondrion. The fact that you don't have a perfect a model of how quarks lead to a mitochondrion does not forbid you to study what a mitochondrion does in a cell. You can say, there is this new thing that emerged, and I'm going to study this thing and its properties in their own right. I think that Comte's view that you ought to understand everything starting from the quarks (or, more plausibly, pure logic) and "moving up" is not too much of a strawman. I agree that "emergence" is now often used as a non-explanation though.