Infants are fascinated by other people, and they bring to this fascination a rich set of inferences and intuitions about the motivations that cause other people to act. An if an aim of AI is to build the flexible, commonsense thinker that human adults become, then AI might need to start like human adults do, from the same core abilities as infants. But a recent experiment that put 11-month old infants head-to-head with neural network models meant to mimic human behavior found that the babies still had the edge when it came to predicting the invisible goals and motivations behind the actions of animated shapes (representing people) on a screen.