Reinforcement Learning: Connections, Surprises, and Challenge