This site may earn chapter commissions from the links on this page. Terms of use.

go

Google acquired the British artificial intelligence startup DeepMind but over 2 years agone, but at the time information technology wasn't clear what the secretive company was working on. Most of DeepMind'due south work has been under the radar, but Google has now announced DeepMind's research has led to a pregnant AI milestone. A new program called Alpha Get has been developed that can shell a professional human histrion at the game of Become, something no computer has managed to do before.

Nosotros're all familiar with chess-playing computers — Deep Blue famously beat Garry Kasparov 20 years agone. Go, which was created more than than ii,500 years ago in Mainland china, is considered a more significant challenge for AI because the overwhelming complexity makes it an "intuitive" game. The goal in Go is to identify your pieces on the board to surround and capture the opponent'south pieces until you control more than one-half of the board. It'due south a game of pattern recognition and skill; there'southward no luck involved, making it a perfect problem to examination artificial intelligence.

The complication of Go comes from the huge number of board configurations. In chess, there are only 32 full pieces and 64 squares on the board. Additionally, each piece can but move in certain means. It'southward possible for a computer to brute force all the potential lath configurations and program many moves in advance. Get is played with identical pieces that are placed on a 19 by 19 grid (361 potential locations). The number of board configurations is huge — more than the number of atoms in the universe, making it impossible for a computer to just creature force the search infinite. Y'all need a computer that can larn to play the game like a homo, and that's what Alpha Go did.

Most serious Go players tin't explicate exactly why certain moves are the right ones — thus, the intuitive aspect. Most programmers felt until recently that Get was then complex it would take decades for a computer to best a human. So, Alpha Get defeated European Become champion Fan Hui five games to zero in a recent match. This coming March, Alpha Become will take on Lee Sedol, one of the best players in the earth.

Google isn't the only AI company that has been interested in cracking Get, and now that it has, many of the same techniques could be practical to other issues. DeepMind researchers adult general AI methods, and so they're not locked into merely playing Get; that would not make for a very useful AI. There are ii basic learning networks inside Alpha Become — 1 network learns to predict probable upcoming moves and the other predicts the consequence of different arrangements of game pieces. It doesn't try to simulate an entire game with all the uncountable board configurations, but instead only thinks a few moves ahead like a human player would.

With a different data set, these algorithms could tackle big bug like medical diagnosis and climate modeling. For at present, DeepMind is focusing on the friction match with Lee Sedol. Alpha Go tin run through millions of games per day to improve its agreement of the game. That might help it win more games, but playing games is simply the beginning.