Is reinforcement learning overhyped?