A human has at last defeated an artificial intelligence (AI).
From the beginning, it was an insurmountable match. It was a heat between a human player and a computer program, AlphaGo, Google DeepMind’s artificial intelligence program, and its 1,202 standard processing units (CPUs).
The formidable AlphaGo had calculated every possible move that could be made by its opponent. The human go player, however, had only been able to get a few hints as to the computer program’s skill level from its previous games.
Lee Sedol, one of the world's top go grandmasters, finally beat AlphaGo on March 13, when he won the fourth showdown of five after 180 moves. With the win, Lee has prevented a whitewash, after losing the first three games.
In the middle of the game, Lee made some decisive moves, occupying territory in adjacent corners and then moving on both his right and left sides. Meanwhile, AlphaGo surrounded more territory, mainly in the top and middle.
Make-or-break moves were made in the middle of the board, where Lee, playing white, started to reduce black's territory. At that moment, AlphaGo went off on a tangent, making a series of weaker moves on the right. Then, a big move came: move No. 78. Lee played his 78th stone in an unexpected location in the middle of the board, turning the tide.
Things started to turn bad for the computer program, even though it can calculate as many as 100,000 possible moves per second. After 30 more moves, AlphaGo failed to turn the tables.
As Lee kept reducing the formerly large amount of black territory, AlphaGo finally resigned, showing on screen, “The result 'W+ Resign' was added to the game of information.”
The human go player's win against a seemingly impregnable opponent instantly made headlines across international news networks. The Chinese CCTV broadcaster reported that, “Lee Sedol has at last grabbed his first win. This win has restored face to humanity.”
The Associated Press said that, “Today’s victory is something priceless that cannot be replaced by anything else. AI cannot be perfect yet.” AFP, too, reported that, “After a series of humiliating defeats, the Korean go grandmaster finally gave us the joy of victory that humanity has outdid machinery.”
The Guardian, a British daily, said, “For the final match of the series, scheduled for Tuesday, Lee has offered to play as black, saying it would make any victory more meaningful.” It then added, “DeepMind founder Demis Hassabis said the loss was a valuable learning tool and would help identify weaknesses in the program that need to be fixed.”
Finally, Michael Redmond, one of the match’s commentators in English at DeepMind's YouTube channel, and himself a professional go player, pointed to what he saw as a clever strategy and as the game’s turning point: move No. 78.
By Wi Tack-whan, Sohn JiAe
Korea.net Staff Writers
Photo: Baduk TV, Yonhap News
whan23@korea.kr
From the beginning, it was an insurmountable match. It was a heat between a human player and a computer program, AlphaGo, Google DeepMind’s artificial intelligence program, and its 1,202 standard processing units (CPUs).
The formidable AlphaGo had calculated every possible move that could be made by its opponent. The human go player, however, had only been able to get a few hints as to the computer program’s skill level from its previous games.
AlphaGo resigns on March 13 in the fourth of five matches against go grandmaster Lee Sedol.
Lee Sedol, one of the world's top go grandmasters, finally beat AlphaGo on March 13, when he won the fourth showdown of five after 180 moves. With the win, Lee has prevented a whitewash, after losing the first three games.
Go grandmaster Lee Sedol clinches his first victory against AlphaGo on March 13 in Seoul. In a post-game press conference, Lee says, “I couldn’t be happier because I won after the previous losses. I think that I've been able to win at least one game thanks to your support and encouragement.”
In the middle of the game, Lee made some decisive moves, occupying territory in adjacent corners and then moving on both his right and left sides. Meanwhile, AlphaGo surrounded more territory, mainly in the top and middle.
Make-or-break moves were made in the middle of the board, where Lee, playing white, started to reduce black's territory. At that moment, AlphaGo went off on a tangent, making a series of weaker moves on the right. Then, a big move came: move No. 78. Lee played his 78th stone in an unexpected location in the middle of the board, turning the tide.
Things started to turn bad for the computer program, even though it can calculate as many as 100,000 possible moves per second. After 30 more moves, AlphaGo failed to turn the tables.
As Lee kept reducing the formerly large amount of black territory, AlphaGo finally resigned, showing on screen, “The result 'W+ Resign' was added to the game of information.”
The human go player's win against a seemingly impregnable opponent instantly made headlines across international news networks. The Chinese CCTV broadcaster reported that, “Lee Sedol has at last grabbed his first win. This win has restored face to humanity.”
The Associated Press said that, “Today’s victory is something priceless that cannot be replaced by anything else. AI cannot be perfect yet.” AFP, too, reported that, “After a series of humiliating defeats, the Korean go grandmaster finally gave us the joy of victory that humanity has outdid machinery.”
The Guardian, a British daily, said, “For the final match of the series, scheduled for Tuesday, Lee has offered to play as black, saying it would make any victory more meaningful.” It then added, “DeepMind founder Demis Hassabis said the loss was a valuable learning tool and would help identify weaknesses in the program that need to be fixed.”
Finally, Michael Redmond, one of the match’s commentators in English at DeepMind's YouTube channel, and himself a professional go player, pointed to what he saw as a clever strategy and as the game’s turning point: move No. 78.
By Wi Tack-whan, Sohn JiAe
Korea.net Staff Writers
Photo: Baduk TV, Yonhap News
whan23@korea.kr
Most popular
- Slew of festivals, events scheduled in downtown Seoul in May
- Korea.net welcomes 2025 K-influencers, Honorary Reporters
- 2025 Honorary Reporter class pledges to spread 'real Korea' worldwide
- US urged to exempt tariffs on Korea in first '2+2' trade talks
- Korean culture festival in Cuba marks 1st year of bilateral ties
