3 ms·Deep Reinforcement Learning Doesn't Work (yet)45 points by stochastician 9y agocozos 9y agoI laughed so hard at the Local Optima example videomav3r1ck 9y agoThis article deserves way more attention. The Atari DQN paper was a really buzzy thing when it came out. It’s been over four years now and we haven’t seen DQN and similar Deep RL techniques generalize well.