Advertisement

SKIP ADVERTISEMENT

Google’s AlphaGo Defeats Chinese Go Master in Win for A.I.

Ke Jie, the world’s top Go player, reacting during his match on Tuesday against AlphaGo, artificial intelligence software developed by a Google affiliate.Credit...China Stringer Network, via Reuters

HONG KONG — It isn’t looking good for humanity.

The world’s best player of what might be humankind’s most complicated board game was defeated on Tuesday by a Google computer program. Adding insult to potentially deep existential injury, he was defeated at Go — a game that claims centuries of play by humans — in China, where the game was invented.

The human contender, a 19-year-old Chinese national named Ke Jie, and the computer are only a third of the way through their three-game match this week. And the contest does little to prove that software can mollify an angry co-worker, write a decent poem, raise a well-adjusted child or perform any number of distinctly human tasks.

But the victory by software called AlphaGo showed yet another way that computers could be developed to perform better than humans in highly complex tasks, and it offered a glimpse of the promise of new technologies that mimic the way the brain functions. AlphaGo’s success comes at a time when researchers are exploring the potential of artificial intelligence to do everything from drive cars to draft legal documents — a trend that has some serious thinkers pondering what to do when computers routinely replace humans in the workplace.

Last year, it was still quite humanlike when it played,” Mr. Ke said after the game. “But this year, it became like a god of Go.”

Perhaps just as notably, the victory took place in China, a rising power in the field of artificial intelligence that is increasingly seen as a rival to the United States. Chinese officials perhaps unwittingly demonstrated their conflicted feelings at the victory by software backed by a company from the United States, as they cut off live streams of the contest within the mainland even as the official news media promoted the promise of artificial intelligence.

AlphaGo — which was developed by DeepMind, the artificial intelligence arm of Google’s parent, Alphabet Incorporated — has already pushed assumptions about just how creative a computer program can be. Since last year, when it defeated a highly ranked South Korean player at Go, it changed the way the top masters played the game. Players have praised the technology’s ability to make unorthodox moves and challenge assumptions core to a game that draws on thousands of years of tradition.

In the first game, Mr. Ke made several moves that commentators said were reminiscent of AlphaGo’s own style. Wearing a blue tie and thick-framed black glasses, the boyish Mr. Ke kept things close in the early going. By AlphaGo’s own assessment, it did not have a big statistical advantage until after the 50th move, according to a DeepMind co-founder, Demis Hassabis.

Mr. Ke, who smiled and shook his head as AlphaGo finished out the game, said afterward that his was a “bitter smile.” After he finishes this week’s match, he said, he would focus more on playing against human opponents, noting that the gap between humans and computers was becoming too great. He would treat the software more as a teacher, he said, to get inspiration and new ideas about moves.

“AlphaGo is improving too fast,” he said in a news conference after the game. “AlphaGo is like a different player this year compared to last year.”

Go, in which two players vie for control of a board using black and white pieces called stones, is considered complex because of the sheer number of possible moves. Even supercomputers cannot simply calculate all possible moves, presenting a big challenge for AlphaGo’s creators.

AlphaGo instead relies on new techniques that help it learn from experience playing a large number of games. This time, Mr. Hassabis said, a new approach allowed AlphaGo to learn more by playing games against itself. In the future, computer scientists hope to use similar techniques to do many things, including improving fundamental scientific research and diagnosing illnesses.

AlphaGo’s victory represents a marketing success for Google and Alphabet. The Mountain View, Calif., software company pulled out of mainland China seven years ago rather than submit to the country’s censorship requirements. But it has continued to express interest in the vast market, which has the world’s largest population of internet users.

Notably, the Go match took place in the city of Wuzhen, where the Chinese internet authorities hold an annual conference on cyberspace regulation.

China has been drawn to AlphaGo since its victory last year over a South Korean Go master, Lee Se-dol. Officials responsible for technology development in China responded quickly by creating new artificial-intelligence programs and making more funding available to researchers, according to Chinese computer science professors, marking a sort of Sputnik moment for the country.

“AlphaGo truly had a big impact” in China, said Wang Shengjin, a professor at the department of electronic engineering at Tsinghua University in Beijing. “Before, we would be discussing how to apply the technology, but it was hard to be clear exactly how to do it, so AlphaGo gave us a vivid example of that.”

Still, China showed some skittishness at game time. Despite huge interest, many Chinese became consternated when it became apparent there was no obvious live video of the event online. A site that follows Chinese censorship orders, China Digital Times, posted a translated notice from the government calling for all websites to block the broadcast.

“Anything that demonstrates that something special about China has turned out to be just another artificial intelligence problem that Google is better solving than any other company is additionally problematic,” said Clay Shirky, a professor at New York University Shanghai, “because it threatens the specialness of the culture.”

The Cyberspace Administration of China, the country’s main digital censor, did not respond to a request for comment.

While AlphaGo had already racked up an impressive record against humans over the past year, the match against Mr. Ke offered a final showdown. Like AlphaGo, Mr. Ke had beaten South Korea’s Mr. Lee in several recent major competitions.

After Mr. Lee’s loss, Mr. Ke had said publicly on Chinese social media that the program “can’t beat me.”

Mr. Ke’s tone changed earlier this year, after he lost three online speed games to the program. At that point, he said on Chinese social media that computers seemed to be showing that some of what humanity thought about the game was incorrect.

Mr. Ke will have two more chances to get the better of AlphaGo with games on Thursday and Saturday. Most experts do not give him much of a chance. But last year, his rival Mr. Lee surprised, winning one game against AlphaGo out of five after a brilliant and unconventional move stumped the software.

Carolyn Zhang in Shanghai contributed research.

A version of this article appears in print on  , Section B, Page 3 of the New York edition with the headline: In Win for A.I., Google Program Humbles Master of a Mind-Boggling Game. Order Reprints | Today’s Paper | Subscribe

Advertisement

SKIP ADVERTISEMENT