A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play | Synapse