Go master Lee defeats AI in retirement match
Lee secured the victory after 92 stones thanks to an early blunder by HanDol, a program developed by Korea’s NHN Entertainment.
Lee announced his retirement from Go, or baduk in Korean, last month, saying he felt he could no longer beat AI programs.
Lee famously defeated AlphaGo of Google’s DeepMind in 2016 in one match of a five-match meeting and remains the only human Go player to have defeated the AI system.
Lee had an advantage of two stones at the start of Wednesday’s match, taking into account the inherent edge for the AI.
But after taking the opening match, the next match will be on a level playing field.
Lee pocketed 200 million won ($172,000) from the match - 150 million won for his appearance and 50 million won for the win.
And as was the case during the series against AlphaGo, an amateur player was behind the computer to place stones for HanDol.
When Lee faced AlphaGo, Aja Huang, a DeepMind programmer and an amateur 6-dan Go player, made the moves on the board after watching AlphaGo’s moves on the monitor.
While the job may sound simple, it required an understanding of Go rules, and Lee later said he was appreciative of Huang’s manners during the match.
Lee also said he was so impressed with Huang that he felt the programmer was like AlphaGo itself.
For HanDol, Korean amateur 5-dan player Lee Hwa-seop will play the role of Huang.
Lee, who handles operations of NHN’s online baduk game, has gone deep into a couple of international amateur tournaments this year.
Lee Se-dol is calling it quits after an illustrious career that featured 18 international titles and 32 domestic titles.