Best Buy's Anniversary Sale Samsung Could One-Up Apple Peloton Alternatives GMMK Pro Keyboard Review Natural Sleep Aids $59 Off Apple TV Equifax Error: Check Your Status Biggest Rent Increases
Want CNET to notify you of price drops and the latest stories?
No, thank you

Google's AI continues to crush humans at ancient game Go

In another win for the machines, AlphaGo has been secretly playing Go online and thrashing its human opponents.

Royalty-free image of wooden Go Game board with black and white tokens. Focus on centered stones.
Getty Images

AlphaGo continues its reign of board game domination.

The AI program developed by DeepMind, a Google-owned artificial intelligence company based in the UK, has been secretly playing and beating some of the world's best Go players online. DeepMind founder Demis Hassabis revealed in a tweet Wednesday that AlphaGo has been playing under the usernames "Magister" and "Master" on online Go sites Tygem and FoxGo.

Master, aka AlphaGo, has been crushing other players over the last few days and managed to beat Ke Jie, the world's reigning Go champion, twice, reported New Scientist. In total, AlphaGo won 50 out of the 51 online games it played, with one game being a draw due to an internet connection time out.

"We're excited by the results," said Hassabis, "and also by what we and the Go community can learn from some of the innovative and successful moves played by the new version of AlphaGo." DeepMind declined to comment further.

AlphaGo drew crowds last year when it defeated Lee Sedol, one of the world's top Go players. AlphaGo took four out of the five games in the contest, marking a noteworthy leap forward for computer smarts.

Go, which originated in China thousands of years ago, is played on a 19x19 grid with black and white stones. The board's size means the number of possible moves is greater than the number of atoms in the universe, according to Google, making it a more difficult programming challenge than chess.

AlphaGo uses sophisticated machine learning technologies like neural networks to evaluate board positions and determine which moves to make.

Update, January 6 at 8:45 a.m. PT: Adds DeepMind declining to comment.