ABSTRACT
Reinforcement learning was critical to AlphaGo's victories over the Go Masters, but the esoteric game, no matter how intriguing, is that well known in Western countries. Machines employing reinforcement learning were successful in game-playing against humans because the machine learned just like human beings learn, from experience with rewards and punishments, and any gamer knows that the more games you play, the better you will get, but unlike humans, a machine's skill can be tirelessly honed through millions of games against only expert humans, but other machines, and once establishing its superiority, it can play against earlier versions of itself, ultimately going far beyond the skill of the best humans through machine self-strengthening. In reinforcement learning, an agent in a given state can perform an action chosen from a set of all actions that the agent can take with respect to the domain in which it finds itself, changing its action as necessary to adapt, confront changes in that environment.
