Human Shin Jinseo Defeats Strongest AI KataGo... 2-Point Handicap, Wins by 4.5 Points
Translated from Korean, summarized and contextualized by DistantNews.
At a glance
- Go grandmaster Shin Jinseo defeated the top AI program KataGo in a 2-point handicap match, marking a significant human victory.
- Shin's win is the first time a human player has beaten KataGo with a 2-point handicap, showcasing human resilience against advanced AI.
- The match highlighted the strategic depth of human play, with Shin Jinseo emphasizing the 'story and charm' of human Go over pure computational power.
South Korean Go prodigy Shin Jinseo, ranked world No. 1, has achieved a landmark victory against KataGo, the current strongest Go artificial intelligence program. In a highly anticipated match played with a 2-point handicap, Shin emerged victorious after 290 moves, winning by 4.5 points. This marks the first time a human player has defeated KataGo under such conditions, a significant moment in the ongoing human-AI rivalry in the ancient game.
The match, held in Seoul, saw Shin Jinseo demonstrate remarkable composure and strategic prowess. In the first game, he was caught off guard by an early move from KataGo, resulting in a swift defeat. However, in the second game, Shin played cautiously, maintaining a high win rate throughout the mid-game. Facing an AI with computational power far exceeding supercomputers, even a single mistake can be fatal. Shin had previously stated that a win rate dropping below 98% would be dangerous.
I was flustered by KataGo's moves, which were more ordinary than in the first game, but I was fortunate that a large-scale joseki unfolded in the bottom right corner.
Despite the AI's overwhelming processing capabilities, the game's outcome hinged on the complex central territory. Shin Jinseo, presented with countless possibilities, faced a critical decision: engage in combat or seek compromise. Against the conventional wisdom that humans should avoid direct confrontation with AI, Shin chose a bold strategy, cutting off an opponent's stone near the center of the board. This decision, a testament to human free will, caused his win rate to fluctuate, dropping below 97% at times, and narrowing the point difference to just 4.
There were countless moments when I wanted to counter KataGo's moves and attack, but I held back and endured. However, I decided I couldn't anymore and came out to the center to cut. I'm not an AI, but I thought I could fight without losing too much.
Ultimately, Shin Jinseo's strategic acumen and ability to execute under pressure secured the win. He utilized his remaining time effectively, outmaneuvering the AI in the crucial final stages. Post-match, Shin expressed relief at navigating the opening smoothly with a standard joseki and acknowledged his desire to fight KataGo's moves, but ultimately chose to concede less damage before engaging in the central combat. He believes that while human skill cannot compare to AI, human Go possesses a unique 'story and charm,' and winning through combat held significant meaning.
The match rules stipulated a 5-hour thinking time and one 30-second period of overtime for Shin, while KataGo had to play each move within 20 seconds. The AI was given a 0.5-point komi (handicap) for the black stones. Shin's match fee was 50 million won per game, with an additional 50 million won for a win. A victory in both games would also earn him a Genesis G90 car. Lee Dan-bi, a professional Go player, made KataGo's moves on the board.
Human skill is incomparable to artificial intelligence. However, human Go has story and charm. I want to emphasize that I won against artificial intelligence through a small battle.
Originally published by Hankyoreh in Korean. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.