[2407.10583] Three Dogmas of Reinforcement Learning