The game of chess is the most widely-studied domain in the history of artificial intelligence. The strongest programs are based on a combination of sophisticated search techniques, domain-specific adaptations, and handcrafted evaluation functions that have been refined by human experts over several decades.
DeepMind’s AlphaGo Zero program, by contrast, recently achieved superhuman performance in the game of Go, by tabula rasa reinforcement learning from games of self-play.
Now, in a new paper, researchers from the Google-owned company have shown that the algorithm could be generalized that can achieve, blank slate, superhuman performance across many challenging domains. The research authors, which include Demis Hassabis have been rapidly advancing their system since their historic win at the game of Go only last year.
Starting from random play, and given no domain knowledge except the game rules, AlphaZero achieved within 24 hours a superhuman level of play in the games of chess and shogi (Japanese chess) along with Go, and convincingly defeated a world-champion program in each case.
The game of chess represented the pinnacle of AI research over several decades. Garry Kasparov’s defeat by IBM’s Deep Blue system remains a landmark in the field of study. Today, state-of-the-art programs are based on powerful engines that search many millions of positions, and use handcrafted domain expertise and sophisticated domain adaptations.
By contrast, DeepMind’s AlphaZero is a generic reinforcement learning algorithm – originally devised for the game of Go – that achieved superior results within a few hours, searching a thousand times fewer positions, given no domain knowledge except the rules of chess.
Interestingly as one user posted on Reddit, AlphaZero may be pointing to a decided mathematical disadvantage to play black in chess. In the game when AlphaZero versed another program, Stockfish, when AlphaZero was white, it won 25 and drew 25 games. When it was black, it won only three and drew 47 games.
Shogi is arguably a significantly harder game, in terms of computational complexity, than chess. It is played on a larger board, and any captured opponent piece changes sides and may subsequently be dropped anywhere on the board.
That the same algorithm was applied without modification to the more challenging game of shogi, again outperforming the state of the art within a few hours, demonstrates the system to be a much more adaptable, general form of artificial intelligence.
While AlphaZero remains a game playing domain-constrained AI, this work does appear to be a big step towards full AGI.