I don’t buy the empirical claim about when recognition is easier than generation. As an example, I think that you can recognize robust formulations much more easily than you can generate them in math, computer science, and physics. In general I think “recognition is not trivial” is different from “recognition is as hard as generation.”
I don’t buy the empirical claim about when recognition is easier than generation. As an example, I think that you can recognize robust formulations much more easily than you can generate them in math, computer science, and physics. In general I think “recognition is not trivial” is different from “recognition is as hard as generation.”