Reinforcement learning is an approach that helps a machine learn by rewarding desirable actions and penalizing undesirable ones. If the artificial intelligence does not require any human inputs to learn, it progresses by trial and error. When its decision or action brings it closer to the agreed goal, it is given positive feedback. This is how it can remember which actions allow it to optimally perform the task.

It is similar to how you would teach a dog new tricks. You give it a treat when it does a command correctly so that it associates the reward with the correct response to a command.

Other interesting terms…