On Neural Network Solutions for the Ancient Game of GO

Abstract

GO is a difficult game for computers to master and the best commercial programs are still weaker than the average human player is. Artificial Intelligence (AI) techniques, especially neural networks, are currently strong candidates to replace the inadequate traditional game-playing techniques. In this paper, we explore a number of current techniques that are being used to implement GO game playing abilities in a computer program. Then a number of new possible approaches will be described that focus on neural networks as their primary tool. Finally, the paper concludes with a discussion of possible future research directions.