Google AI AlphaGo Wins Its First Go Match Against Go Champion Lee Sedol

Google’s AI (Artificial Intelligence) program AlphaGo has won its first match against South Korean professional Go player Lee Sedol, the first time a computer has ever beaten a professional in the 3,000-year-old Chinese game.

As MSN reports, South Korean professional Go player and world champion Lee Sedol believed going into this week’s tournament that he would have the upper hand. Two weeks ago, he predicted that he would sweep the best of five tournament 5-0. However, AlphaGo got the best of Sedol from the beginning and wound up winning the historic first match.

“I was very surprised because I did not think that I would lose the game. A mistake I made at the very beginning lasted until the very last.”

In a grueling four-hour match, which you can watch in the video below, the Google AI punished the South Korean champion for every slip-up, eventually forcing the world champion to resign.

“I had a lot of fun playing Go and I’m looking forward to the future games.”

Sedol’s loss shocked the Go community, especially in South Korea, where the game is immensely popular. Observer Kim Sung-ryong, also a Go expert, praised AlphaGo’s ability to recover from mistakes by taking emotion out of the equation.

“It did not play like a human at all.”

Now facing a 1-0 disadvantage, Sedol may yet find Google’s AI insurmountable. AlphaGo only needs two more victories out of the next four games to win the tournament and take away the $1 million prize from Sedol (Google will donate the prize to charity if AlphaGo wins).

Go is an almost deceptively simple game. Players compete on a 19 x 19 grid, taking alternating turns placing black and white tiles (called “stones”) at intersections on the board (called “points”). The goal is to control more territory than your opponent at the end of the game.

Google AI Go Match
A game of Go in progress. [Image courtesy of Goban1 via Wikimedia Commons by License]

Go has long been considered one of the last hills to conquer in the world of artificial intelligence (AI). AI got the upper hand in chess after decades of trying in 1997, when IBM’s Deep Blue AI bested world champion Garry Kasparov.

But building a computer that could beat a champion human at Go has proved elusive.

The problem is that Go, unlike chess, is basically infinite. Despite its simple setup, simple rules, and simple objective, Go has orders of magnitude more possible outcomes than chess. A chess game has 10120 possible outcomes. A game of Go, by comparison, has 10761 possible outcomes.

Also, Go has an intuitive element to it; players rely on “intuition and feelings” to plan out their next move, rather than a series of possibilities with a finite number of outcomes.

To learn the game, according to Engadget, programmers gave AlphaGo some 30 million moves, gleaned from professional Go players, and then set the AI to playing against itself. Using a process called “reinforcement learning,” Deepmind then developed its own strategies through trial and error.

Games have long been the proving ground for artificial intelligence. AI is essentially accomplished by teaching computers to “think” – that is, consider all possible outcomes in any given situation and respond with the best possible response. And the best means of achieving that has been to teach the computers to play games. And for decades, that game has been chess.

But now that Go has been conquered, AI has taken a step further into someday becoming a reality. AlphaGo’s programmers believe that since a computer has conquered intuition, AI is a step closer to being useful in fields such as medicine and law enforcement, among other areas.

[Photo by Kim Hee-Chul-Pool/Getty Images]