Professional Go player retires after admitting that Google’s AI cannot be beaten
DeepMind remains victorious.
A professional Go player has retired in response to losing four out of five matches against an AI developed by Google’s DeepMind Technologies.
Lee Se-dol, a 36-year-old hailing from South Korea, has retired after a 24-year career in which he won 18 international competitions. The Go professional became famous in March 2016 after becoming the only human to ever beat Google’s AlphaGo AI, which was developed to test the capabilities of artificial intelligence.
Despite managing to successfully bamboozle the AlphaGo program into making a poor move and eventually admitting defeat, Se-dol claimed in an interview with the Yonhap News Agency that he only won because of a bug known to occur in a similar Chinese Go computer program called Fine Art: “It loses in a strange way. It’s due to a bug.”
After losing to the AlphaGo AI one last time, Se-dol felt that he simply couldn’t claim mastery over the professional Go sphere if there was one opponent that he couldn’t best.
“Even if I become the number one, there is an entity that cannot be defeated,” he said.
Se-dol’s decision to retire was said to also have been driven by the community’s critical reaction to his losses, which made him feel a sense of defeat when going up against such a monumental challenge like the AlphaGo AI.
Go, which is also known as baduk in Korea, is a game for two people wherein players take turns to place black and white stones on 19-by-19 checkerboard grid. The abstract strategy game dates back to more than 2,500 years ago and is considered one of the oldest board games still being played today.
Though he may be retiring, Se-dol will be celebrating the end of his career by challenging yet another AI opponent, HanDol, which was developed in 2018 by the South Korean NHN Entertainment Corporation and has a track record of defeating five of South Korea’s top Go players.
AlphaGo doesn’t just prove a formidable opponent when it comes to playing Go, as its root artificial intelligence program, AlphaZero, also happens to be pretty good at chess and shogi - a Japanese form of chess. We’ll have to wait and see whether Google produces any more AI board game champions.