An interesting example of reinforcement learning