China wants a 'Go' at Google's DeepMind
Challenger emerges for AI's board-game crown
A group from China says it wants to build an AI to challenge the Google DeepMind AlphaGo system.
According to a Reuters and Shanghai Securities News report, the Chinese Association for Artificial Intelligence and the Chinese Go Association plan to challenge AlphaGo to a computer-versus-computer Go match.
The report did not give the name of China's Go-playing AI or how far along development of the system is, but reported a challenge to AlphaGo would be made by the end of the year. The news was announced as Google CEO Sundar Pichal visited a Go school in China as part of a tour of the country.
The AlphaGo program, based on Google's DeepMind computer learning system, achieved fame earlier this month when it routed the world's top Go player in a five-match series.
Mastering Go and competing at the level of top human professional players has been hailed as a significant accomplishment for artificial intelligence and computer learning researchers.
The immense number of potential moves involved in a Go game makes playing the game at a high level all but impossible with the traditional search-tree playing methods used for games such as chess.
Instead, DeepMind has to use its neural network structure to learn from previous games against both humans and itself to develop its own playing style and analyze each move. Lee Sedol, AlphaGo's human opponent in its last match, described the AI as making moves that "could not have been possible for human being to choose." ®