One of the latest, really impressive research papers on Arxiv.org explores how DeepMind has created Player of Games — an AI system with a general-purpose algorithm that unifies previous approaches, combining guided search, self-play learning, and game-theoretic reasoning.
Player of Games is the first algorithm to accomplish strong empirical performance in large perfect information games like the Chinese board game Go and chess, and imperfect ones like poker.
Player of Games is a big step towards the creation of truly "general AI" game-playing systems. The central principle in the experiments was that the algorithm could perform better if given more computational resources. Martin Schmid, a leader of the research project, expects to considerably scale the approach used in the system in the foreseeable future.
Data Phoenix Newsletter
Join the newsletter to receive the latest updates in your inbox.