An algorithm that learns through rewards may show how our brain does too

By optimizing reinforcement-learning algorithms, DeepMind uncovered new details about how dopamine helps the brain learn.