Reinforcement learning:

A machine learning paradigm where some agent interacts with a constantly changing environment to learn the best actions it can take based on some reward signal. E.g. This is used as part of bigger systems to solve things like complex video/boardgames, self-driving etc.