Hungarian scientists have developed artificial intelligence agents that can play the classic arcade game Ms. Pac-Man, and improve their game-play strategy as they go. The researchers are trying to develop the agents’ reinforcement learning capacity.
Ms. Pac-Man has proven to be a good training ground because the movements of the ghosts that the little lady is always trying to avoid are random. A player can’t predict where the ghosts are going to go. Instead you have to constantly watch them, and adjust your strategy on-the-fly. The AI player makes these decisions based on a series of priorities, or rules–avoid the ghosts is number one. We can’t guarantee it will be as fun as the game, but you can read the paper here.