S. Korean Go champion expresses frustration after match with AI program

By Lim Chang-won Posted : January 24, 2019, 11:34 Updated : January 24, 2019, 11:34


SEOUL -- South Korea's reigning champion in the ancient Chinese game of Go expressed frustration after he lost to a home-made artificial intelligence program which has scored a clean victory in a series of matches with top players.

"I think it will be a little difficult for my generation of professional players to win, even if they work hard," Shin Jin-seo, 18, declared after his match Wednesday with HanDol, an AI program developed by NHN Entertainment, the country's largest game publisher. He became the fifth professional Go player to lose.

Inspired by Google's AI program AlphaGo which stunned the world in March 2016 with an overwhelming 4-1 victory against South Korean Go master Lee Sedol, NHN has developed its own program.

"There was a big difference in judgment," Shin said. "In matches with humans, the opponent can make a mistake at any time. It's fun to find such a mistake, but artificial intelligence basically makes no mistakes, and I felt a lot of frustration there."

In 2017, Google's AI team DeepMind unveiled a more powerful program called "AlphaGo Zero" that can achieve superhuman performance in the most challenging domains with no human input, vowing to create general-purpose AI that can learn on its own and, eventually, be used as a tool to solve pressing problems from climate change to disease diagnosis.

HanDol is better than the 2016 version of AlphaGo, but it lags behind AlphaGo Zero, NHN said, adding it would develop Go and other board games based on AI and use it for other purposes as DeepMind did.
기사 이미지 확대 보기