Machine Reinforcement Learning
Reinforcement learning is a way for artificial intelligence to learn without being programmed. It involves having machines carry out tasks over and over, and when the desired result occurs, favoring that, until the machine understands how to carry out the task. It’s no different than the familiar test of mice and cheese. You put the mouse in a box with two levers. One gives them cheese, the other shocks them. After experimenting, the mouse will learn to press only the button that gives them cheese. Reinforcement learning does the same with computers, at an infinitely increased speed, and replacing getting cheese with things like winning the board game Go or successfully merging self-driving cars.
Return To Article
