Eight years ago, Go grandmaster Lee Sedol from South Korea lost to AlphaGo, an artificial intelligence system developed by Google.
The game was a best of five series. Lee, the world’s best human Go player at the time, won one game and AlphaGo won four. It marked the first time that a pro human Go player was defeated by an AI system.
Google Korea sat down with Lee Sedol to discuss what was a defining moment in his career, and during the interview Lee confessed that had he known what AlphaGo was capable of, he may have “gone in a different direction” in his adult life.
Go: Lee vs. AlphaGo
Go is a strategy chess game. The aim is to capture and occupy as much space on the board as possible, squeezing the opponent out. Go originated in China over 2,500 years ago, and is now widely played in China, Japan and South Korea.
AlphaGo was developed in 2014 by DeepMind, Google’s AI unit, which has created a slew of the company’s AI products including, most recently, Gemini.
Prior to playing against AlphaGo, Lee had won major international Go titles from 2002 to 2011, and was regarded as the best Go player in the world in 2007 and 2008.
Google reached out to Lee in 2016, and both Lee and AlphaGo began a series of matches in March of that year. It would redefine how the game was played and studied by millions across East Asia.
“I thought I would definitely win,” Lee said. “At the time, I thought it was just an experiment or something Google was doing. So, I said yes without much hesitation … but I knew it was not the case after it came out all over the news.”
Lee described how bizarre it felt playing against a computer for the first time, due to the lack of human interaction.
“Should I call it a momentum or force?” he asked. “Possessing and controlling that upper hand is a very important part of Go. It’s like when you play tennis and you hit the tennis ball against the wall.”
AlphaGo won the first three rounds of Go, and Lee won the fourth. AlphaGo won the fifth and final round.
Google’s AI had Lee shaken
Lee announced his retirement from pro Go in November 2019, citing the increasing dominance of AI. “AI is an entity that cannot be defeated,” he told South Korean media at the time.
Lee confessed that had he known about AI’s potential earlier, he may have chosen a different career path.
“Before, when I was asked if I would play Go again, I would say that I would definitely learn Go again and become a professional. But after AlphaGo came out, my thoughts have changed a lot. Maybe I would go in a different direction, like having an AI-related career,” he said.
How AI changed Go
Lee stated in the interview that since the advent of AlphaGo, keeping records of Go games for learning purposes has not been the same.
“Now, [old Go games] only hold historical value. In fact, they are no longer used to study Go because AI’s records are much better in terms of content. After looking at the AI’s record, one might learn by thinking, ‘This is how it should be placed here, and this is how it should be placed there’.”
On social media fans of Go are lamenting how the arrival of AI has changed a player’s relationship with the game.
“A person’s play style is an expression of themselves – like seeing an artist’s style in a painting. It’s personal and meaningful. I can’t help but feel a sense of sadness and loss,” wrote Jingna Zhang, a photographer and Go enthusiast, on X.
So, what is Lee Sedol doing these days? He said he plans to design his own board game. Lee said he’s experimented with Gemini, Google’s AI assistant. Gemini found itself in hot water in February for its supposedly woke AI-generated images.
“I think it would be a fun experience creating new board games with Gemini,” Lee said.
Images: Google Korea via YouTube
