The AtlanticThe Atlantic

How Google's AlphaGo Beat a Go World Champion

By Christopher Moyer

28 Mar 2016 · 8 min read

On March 19, 2016, the strongest Go player in the world, Lee Sedol, sits down for a game against Google DeepMind’s artificial-intelligence program, AlphaGo. They’re at the Four Seasons Hotel in Seoul’s Gwanghwamun district, and it’s a big deal: Most major South Korean television networks are carrying the game. In China, 60 million people are tuning in. For the English-speaking world, the American Go Association and DeepMind are running an English-language livestream on YouTube, and 100,000 people are watching. A few hundred members of the press are in adjacent rooms, watching the game alongside expert commentators.

The game room itself is spare: a table, two black leather chairs, some cameras. Three officials presiding over the match sit in the back. Across from Lee sits Aja Huang, one of AlphaGo’s lead programmers; and beside him is a computer monitor that displays AlphaGo’s move choices. Huang’s job is to physically place AlphaGo’s pieces on the board. AlphaGo itself is not any one machine—it’s a piece of distributed software supported by a team of more than 100 scientists.

Sign in to informed

  • Curated articles from premium publishers, ad-free
  • Concise Daily Briefs with quick-read summaries
  • Read, listen, save for later, or enjoy offline
  • Enjoy personalized content