Image: Counting the Stones of Life by Lee Se-dol. Credit: Woongjin Knowledge House.
Years before ChatGPT, the 2016 match between Lee Se-dol 이세돌 and AlphaGo had ablaze with interest in artificial intelligence as Lee, considered one of the greatest go (called baduk 바둑 in Korean) players in the history of the game, squared off an AI engine developed by Google DeepMind.
Though AI had long been good enough to defeat the best humans at chess, go was considered too complex a game for AI in 2016. But in a stunning upset, AlphaGo defeated Lee in four out of their five games. Lee retired three years later; his one victory over AlphaGo has become totemic among fans as a symbol of humans over machines. In a memoir published this week, Counting the Stones of Life 인생의 수읽기, Lee looked back at the match of the century.
Lee recalls that when he reviewed AlphaGo’s games five months prior to his match, the AI engine was no better than an amateur. Just ten turns into their first match, however, Lee could sense that AlphaGo had become a strong player. He felt that he could win Game 3 with an aggressive opening, which he expected the AI to struggle with, but was proven wrong.
Lee ultimately prevailed in Game 4 by making a move - one he would not have made against a human - designed to induce a calculation error for the computer. Lee played it straight in Game 5, feeling he had cheated in Game 4, and lost again.
Lee acknowledges that AlphaGo made new, innovative moves that conventions would have prevented human players from thinking of. But overall, he finds the impact of AI on go a negative one. “Baduk is a game where each player seeks the best path forward in an abstract field. But after AlphaGo, it became a game with a predetermined answer. What is the role of a pro in a game with a predetermined answer? A pro gamer? But not even e-sports pros just mimic what the computer tells them to do. Today, baduk is mostly about how closely you study AI.”
