Reinforcement learning trains computers through a feedback system of rewards and penalties.

Imagine teaching a dog to sit. When they do it, they get a treat. When they don't, they get nothing. This is exactly how reinforcement learning works. A computer program, called an agent, tries different moves in an environment. If a move gets it closer to its goal, it receives a virtual reward. If it makes a mistake, it gets a penalty. Over time, the computer figures out which actions bring the biggest rewards.