DeepMind chief reflects on AlphaGo, AI future

“It feels like yesterday that we played that match, even though it has been 10 years. At the same time, considering how much has changed, it also feels like a century has passed.”

Demis Hassabis made the remarks on April 29 in Seoul, his first visit to South Korea in a decade. He spoke at the “Google for Korea 2026” event at the Westin Josun Seoul, marking 10 years since the landmark 2016 match between DeepMind’s AlphaGo and Lee Sedol. Hassabis joined Lee and writer Cho Seung-yeon for a discussion titled “AlphaGo 10 Years On: A Vision for AI for All.”

Hassabis described the match as a turning point that signaled the rise of modern AI. He said DeepMind began its research in 2010, when few were focused on the field, driven by the belief that AI would become a powerful tool for scientific discovery. He predicted major breakthroughs in science over the next decade and said AI could help usher in a new era of human prosperity.

Since AlphaGo, DeepMind has developed AlphaFold, an AI system that predicts protein structures and has generated about 200 million such predictions. For that work, Hassabis received the 2024 Nobel Prize in Chemistry. He said it is possible that many diseases could be addressed within the next 10 to 20 years.

Lee, who is widely known for securing one victory against AlphaGo in their five-game match, cautioned that working with AI requires balance. He said there is a risk that people could cede control of their thinking to machines and stressed the need for careful use.

Hassabis later played a brief 10-minute exhibition match against Shin Jin-seo, the world’s top-ranked Go player, at an event hosted by the Korea Baduk Association. Shin said he was struck by how closely Hassabis’ play resembled AI patterns. While many players now study the game using AI, he said there are still moves that can unsettle it.

In a subsequent discussion, Hassabis addressed artificial general intelligence, emphasizing the need for safeguards. He said it is essential to ensure AI operates within clear guardrails and is not misused by malicious actors.


최지원 기자 jwchoi@donga.com