DeepMind's David Silver and Richard Sutton's position paper on the possible upcoming future of AI heavily based on reinforcement learning.