Evolving an expert checkers playing program without using human expertise

An evolutionary algorithm has taught itself how to play the game of checker s without using features that would normally require human expertise. Using only the raw positions of pieces on the board and the piece differential, the evolutionary program optimized artificial neural networks to evaluate a lternative positions in the game. Over the course of several hundred genera tions, the program taught itself to play at a level that is competitive wit h human experts (one level below human masters). This was verified by playi ng the best evolved neural network against 165 human players on an Internet gaming zone. The neural network's performance earned a rating that was bet ter than 99.61% of all registered players at the website. Control experimen ts between the best evolved neural network and a program that relies on mat erial advantage indicate the superiority of the neural network both at equa l levels of look ahead and CPU time. The results suggest that the principle s of Darwinian evolution may be usefully applied to solving problems that h ave not yet been solved by human expertise.