Three Dogmas of Reinforcement Learning