Comment by ctoth

Reinforcement learning from self-play/AlphaWhatever? Nah must just be datasets. :)