“David Silver strongly believes that true superintelligence must be achieved through reinforcement learning from a machine's own experience, not by training on human-generated data which he considers inferior because it contains mistakes.”