Machines Learn Better if We Teach Them the Basics

To improve reinforcement learning algorithms, computer scientists are training them as if they were human.