Early game playing AI
Building AI's to play games started in the 50's with researchers building programs that played checkers and chess. These two games have a few properties in common:
They are zero-sum games. Any reward that one player receives is a corresponding loss to the other player and vice versa. When one player wins, the other loses. There is no possibility of cooperation. For example, consider a game such as the prisoner's dilemma; here, the two players can agree to cooperate and both receive a smaller reward.
They are both games of perfect information. The entire state of the game is always known to both the players unlike a game such as poker, where the exact cards that your opponents are holding is unknown. This fact reduces the complexity that the AI must handle. It also means that a decision about what the best move can be made is based on just the current state. In poker, the hypothetical optimal decision about how to play would require information that is not just on your...
