Misadventures in machine learning for mathematics

I will describe some potential (but thus far largely unsuccessful) applications of ML techniques to proof discovery in geometric analysis.