Teaching a robot arm how to play tennis might be a possibility with reinforcement learning techniques and enough trial and error.