The automated discovery of algorithms for multiplication of matrices, using a deep reinforcement learning approach devised by DeepMind, is reported in Nature this week. This approach quickly re-discov