Facebook create Go playing AI

Mark Zuckerburg, founder of the popular social networking site Facebook, today announced his new AI’s ability to compete against human players at the ancient Chinese game ‘Go’

The game, thought to be more than 2,500 years old, is played by two players, one with black stones, one with white. The players take turns to place the stones on a grid, the goal being to dominate the board by surrounding the opponent’s pieces with their own. Once played a stone cannot be moved unless it is surrounded by the opponent’s pieces and thus captured.

‘Go’ as a game has approximately 10^700 possible scenarios of play making it truly monumental computational task for AI to be trained in and learn. In contrast the popular game chess, which IBM’s ‘Deep Blue’ computer was trained in and beat the at the time world chess champion Garry Kasparov in 1997, only has approximately 10^60 possible scenarios.

Mark Zuckerburg wrote on his Facebook page that ‘Go is one of the last games where the best human players can still beat the best artificial intelligence players’.

The AI still has some ways to go, admitting that there are some major flaws in the design of the AI where if the AI starts loosing, it panics and plays almost random seeming moves.

Despite this, it appears that we may be that one step closer to loosing supremacy of another game to AI.

Author Description

Martyn Price

Main editor at PCI-News aswell as IT technician, network engineer, systems administrator, website designer, manager and Data recovery specialist at PCI Xpress ltd

No comments yet.

Join the Conversation