Hacker News Reader

Well, if I understood correctly, the RL DeepMind implementation is basically making a RL algorithm work with a supervised model.

This has been done since the 90s. The Deepmind paper is about a few more tricks.