South Korean Grandmaster Lee Sedol has won the fourth game of Go in the five-match duel against the Google-developed supercomputer. The human Go champion said he managed to find weaknesses in the software.
The AlphaGo computer program suffered its first defeat in the match-up with Lee, resigning some five hours into the Sunday game.
Lee is among world's best players of the ancient Asian board game of Go, where the players fight for territory by placing black or white stones on a 19-by-19 grid.
The human grandmaster, however, lost three consecutive matches to Google's AphaGo software after agreeing to take the machine on for the best-of-five series.
This is Lee's first victory since the duel started last week in Seoul.
"I couldn't be happier today...this victory is priceless," the 33-year old Lee told the audience. "I wouldn't trade it for the world."
Millions of Go fans around the world have been following the match, along with numerous artificial intelligence scientists.
"I can't say I wasn't hurt by the past three defeats...but I still enjoyed every moment of playing so it really didn't damage me greatly," Lee added.
AI 'pressured' into mistakes
The South Korean winner of 18 international championships said that he finally managed to discover two weaknesses in the Google-developed software.
According to Lee, AlphaGo has more difficulty when playing with black stones, which give the player right to start the game and award additional points to the white player. Also, the machine responded badly when Lee made an unexpected move, he said.
Demis Hassabis, the head of the AlphaGo developer Google DeepMind, described Lee as an "incredible player" and the Sunday loss as an important learning experience for the engineers.
"It was doing well...but then, because of Lee's fantastic play, it was pressured into some mistakes," he said of his program.
"Actually we are very happy because this is why we came here, to test AlphaGo and its limit and find out what its weaknesses were," he added.
Deeper than Deep Blue
The match has drawn parallels to the famous duel between the chess grandmaster Gary Kasparov and the Deep Blue program in 1997.
However, the number of combinations in Go is far bigger than in chess, barring the engineers from relying solely on brute-force calculations. To hone its skills, AlphaGo studies old matches and uses simulated games, mimicking human intuition.
Darko Janjevic (AP, AFP, Reuters)