Craig S. Kaplan on Nostr: nprofile1q…estct Isn't part of the problem also that ML researchers don't really ...

nprofile1qyt8wumn8ghj7un9d3shjtnyd968gmewwp6kytcqyrqhdf27mzjlgfqd6c25lqwlq9mfnzagd82ghm6ht3r7x0yjql2ty7estct (nprofile…stct) Isn't part of the problem also that ML researchers don't really understand why these techniques work as well as they do? As a result, there aren't really good empirical evaluation methods that go beyond "poke it with a stick and see what happens".