At this point in the progression of machine-learning AI, we’re accustomed to specially trained agents that can utterly dominate everything from Atari games to complex board games like Go. But what if ...