Former Go champion crushed by DeepMind retires after declaring AI invincible

The South Korean Go champion Lee Se-dol has retired from skilled play, telling Yonhap news agency that his choice was motivated by the ascendancy of AI.

“With the debut of AI in Go video games, I’ve realized that I’m not on the prime even when I change into the primary by frantic efforts,” Lee instructed Yonhap. “Even when I change into the primary, there’s an entity that can not be defeated.”

For years, Go was thought of past the attain of even essentially the most refined pc applications. The traditional board recreation is famously advanced, with extra potential configurations for items than atoms within the observable universe.

This popularity took a knock in 2016 when the Google-owned synthetic intelligence firm DeepMind shocked the world by defeating Se-dol four matches to one with its AlphaGo AI system. The video games had a world impression, alerting the world to a brand new breed of machine studying applications that promised to be smarter and extra inventive than AI of previous.

Lee, who was the world’s primary ranked Go participant within the late 2000s, initially predicted that he would beat AlphaGo in a “landslide” and was shocked by his losses, going as far as to apologize to the South Korean public. “I failed,” he mentioned after the tournament. “I really feel sorry that the match is over and it ended like this. I wished it to finish effectively.”

Regardless of the end result, Go consultants agreed that the match produced excellent play. AlphaGo shocked the world with its so-called “move 37,” which human consultants initially thought was a mistake, however which proved decisive in recreation two. Lee made his personal impression together with his “hand of God” play (transfer 78), which flummoxed the AI program and allowed Lee to win a single recreation. He stays the one human to ever defeat AlphaGo in match settings. (Throughout training AlphaGo misplaced two time-capped video games to Go participant Fan Hui.)

For the reason that match, although, DeepMind has solely improved its AI Go programs. In 2017, it created AlphaGo Zero, a model of this system which surpassed even AlphaGo.

Whereas the unique AI realized to play Go by finding out a dataset of greater than 100,000 human video games, AlphaGo Zero developed its expertise by merely enjoying itself, again and again. After three days of self-play utilizing massively highly effective pc programs that allow it play video games at superhuman speeds, AlphaGo Zero was capable of defeat its predecessor 100 video games to nil. DeepMind mentioned on the time that AlphaGo Zero was probably the strongest Go participant in historical past.

In an announcement given to The Verge, DeepMind’s CEO Demis Hassabis mentioned Lee had demonstrated “true warrior spirit” in his video games with AlphaGo. Stated Hassabis: “On behalf of the entire AlphaGo crew at DeepMind, I’d prefer to congratulate Lee Se-dol for his legendary decade on the prime of the sport, and want him the easiest for the long run … I do know Lee will probably be remembered as one of many biggest Go gamers of his era”

In response to Yonhap, Lee isn’t utterly giving up on enjoying AI, although. He plans to commemorate his retirement in December by enjoying a match towards a South Korean AI program known as HanDol, which has already crushed the nation’s prime 5 gamers. Lee will probably be given a two-stone benefit.

“Even with a two-stone benefit, I really feel like I’ll lose the primary recreation to HanDol,” Lee instructed Yonhap. “As of late, I don’t observe Go information. I wished to play comfortably towards HanDol as I’ve already retired, although I’ll do my finest.”

Replace: Remark from DeepMind CEO Demis Hassabis has been added to the story in addition to clarification about gamers who’ve crushed AlphaGo in numerous settings.

Source link