Damek

wait so how does this site work do you double post from twitter?

I agree, but how do you convince rationalist and more importantly policy people influenced by rationalists and their P(doom). IME conversation just devolves into (1) bet me next time I make a claim; (2) it’s better than nothing; (3) It doesn’t matter how bad initial guess is I just update my prior.

Important to couple cynicism with naïveté.

Interesting do you have a link to a write up/paper?

I guess we’re speaking kind of generally here. can you provide concrete examples?

putting aside theoretical guarantees does conformal empirically provide significantly better coverage than non conformal techniques? (Not sure whether this is what you’re saying.)

Is there an example application that really benefits from conformal prediction, I.e., where the results you get are significantly better than classical confidence intervals in some metric? I ask every person I meet if CP actually works in practice and I never get a satisfying answer.