Home>Latest

AlphaGo 1, human 0 in first of 3 games

By MA SI in Wuzhen, Zhejiang (China Daily) Updated : 2017-05-24

柯洁.jpg

Ke Jie, the world's top human Go player, loses a game on Tuesday to the artificial intelligence program AlphaGo. XU YU / XINHUA

Google's artificial intelligence program AlphaGo is one step closer to omnipotence in the ancient game of Go on May 23 after it prevailed in the first of three matches against Ke Jie, a 19-year-old Chinese prodigy who is the world's No 1 human player.

The victory shows the big leaps AI has made in the 3,000-year-old board game, which at one time was thought to be beyond the reach of computer algorithms because of its complexity, experts said.

"AlphaGo is completely different from what it was one year ago. In the past, it had shortcomings. But now it is progressing so fast that I have not found any of its weaknesses yet," Ke said at a news conference in Wuzhen, Zhejiang province, after losing a close game by half a point.

But Ke, the holder of multiple world titles, wasn't ready to give up before games on Thursday and Saturday. "I am not that satisfied with my performance. I could have done better. For the next two games, I will spare no efforts to make it," he said.

The match is the latest showdown between elite human Go players and AlphaGo, which defeated South Korean Go master Lee Se-dol 4-1 in March 2016.

Demis Hassabis, CEO of Google DeepMind, said the company has upgraded AlphaGo into a more capable version since then.

It now uses 10 times less computational power than its predecessor, and can work from a single PC connected to Google's cloud server, the company said.

Go, which originated in ancient China, is a game in which two players take turns placing black and white stones on a 19-by-19 grid of squares to try to control the most territory.

In the first game, Ke played the black stones, with Alpha-Go taking the white. During the competition, Ke looked as if the pressure were on, given his strong expressions and fidgeting as he pondered his next moves against AlphaGo.

"Ke tried to imitate Alpha-Go's playing style at the beginning, and did make several good moves, but still failed to outcompete AlphaGo," said Wang Runan, chairman of the Chinese Weiqi Association in an interview with news portal QQ. Weiqi is Pinyin for Go.

There is still a chance for Ke to win the best two out of three, Wang said, if he gains inspiration from AlphaGo and comes up with new moves and ideas.

Xiang Yang, an AI expert at the China Center for Information Industry Development, said the competition highlights once again that AI is developing at an unexpected pace.

"After showcasing its technological prowess, more efforts are needed to leverage AI for commercial applications. The game is a tipping point, but using it to solve problems will be the next big challenge," Xiang said.