RT @kareem_carr
I think this is the right way to think about innovation in ML models. It’s basically a brute force search through many many options. Statisticians would never have done this. Too empirical for us.