A China-developed artificial intelligence program defeated the world’s second-ranked Go player on Friday afternoon, nearly a year after the same player lost to AlphaGo, an AI project backed by Google, three games to zero.
Go is an ancient and famously complex board game played by two individuals who aim capture as much territory as possible by placing black or white stones on a 19-by-19 grid. There are more possible board configurations in Go than there are atoms in the known universe. In recent years, few players, if any, have been as dominant as Ke Jie of China, who has been playing Go professionally since the age of 10.
After launching on April 12, the Chinese AI, Golaxy, won 28 of 30 games against top-ranked Go players — including a victory over world No. 1 Park Junghwan of South Korea — prior to facing Ke. Funded by Beijing-based investment firm NCF Group, Golaxy is expected to rival or even surpass its predecessor, AlphaGo.
Golaxy can reach the same level of mastery as AlphaGo from playing fewer games, and it should reduce the amount of resources required to train human players, according to Golaxy chairman Jin Xing. “It can also play different kinds of games. For example, it can play a Go game where it doesn’t let up once it has established an advantage,” Jin said at a press conference. “This goes well beyond the AlphaGo framework.”
A screenshot from the livestreamed Go match between world No. 2 Ke Jie and China-developed AI program Golaxy in Fuzhou, Fujian province, April 27, 2018. From Weibo user @新浪科技
The match between Golaxy and Ke is intended not only to show off the lofty level of AI technology in China, but also to encourage people to think about how such sophisticated machine-learning processes could be applied to other fields in the future, said Lin Jianchao, head of the Chinese Go Association.
Last year’s face-off between Ke and AlphaGo was one of China’s most talked-about news events in 2017, attracting hundreds of millions of views on social media and considerable attention to the game, both in China and internationally. On video-streaming site Netflix, you can even find an English-language documentary about the triumphs of AlphaGo, which has since “retired” from competition.
In addition to Golaxy, there are a handful of other Chinese-developed Go AI programs from companies such as internet giants Alibaba and Tencent. Tencent’s program, dubbed “Fine Art,” won the 2017 Computer Go UEC Cup, a prestigious competition for Go programs held annually in Tokyo, and also defeated Ke in January of this year. On Monday, Fine Art was named the official training program for the Chinese national Go team.
Despite the existential sadness many feel when witnessing humanity’s best and brightest being outwitted by machines, programmers and players alike believe AI is good for Go. Programs like AlphaGo, Fine Art, and Golaxy “let people learn, understand, and enjoy the game on a deeper level,” according to Jin.
Beyond the fans AI programs attract to the game through hyped-up man-vs.-machine showdowns, they also benefit Go organizations and players of all skill levels. In March, for example, the Sina Go Institute began using AI programs to train players.
On Friday, the organizing committee of an amateur tournament explicitly banned computerized assistance after a player wearing earphones connected to a cellphone protruding from his breast pocket was suspected of cheating. Already, many Go tournaments have discontinued the practice of holding preliminary games online because of the risk of players at home using computer programs to gain an unfair advantage.
“2017 marks the first year of learning Go from artificial intelligence,” Li Zhe, a Go professional, wrote in a commentary in October 2017. He added that because of rapidly evolving AI technology, the entire game is likely to change as people adjust to playing styles they might never have seen before, or not have considered previously.
“Thanks to AI, Go and the culture surrounding it have a new tool,” Li said. “How to utilize AI to its full potential is now a critical question to competitive players.”
Contributions: David Paulk; editor: David Paulk.
(Header image: Ke Jie of China ponders his next move during the match against AI program Golaxy in Fuzhou, Fujian province, April 27, 2018. Zheng Shuai/IC)