Machine trains self to beat humans at world's hardest game
The ancient strategy game of Go may have met its ultimate match in Google’s AI.
The brain-taxing board game is a little like an Eastern version of chess, except many times more complex. It has millions of devotees in China, Korea and Japan. Many of them tuned in today to watch an artificial intelligence computer built by Google’s DeepMind beat the world champion, Lee Sedol, in the first of a five-game contest.
Duels like these don’t come often. That’s because the vast majority of human-versus-machine contests are rarely worth watching at all. Either the humans would so obviously win—try getting a machine to write jokes—or we’d so obviously lose. You wouldn’t run a race against a car, or try to out-hammer a steam drill.
But every once in awhile, a technological moment comes when the man-machine match-up gives us a fight worth watching. (Just ask John Henry.) DeepMind vs. Lee Sedol is one of those moments. But as the video above shows, the stakes of this week’s historic battle may not be what you think.
More Like This

A tragic bridge collapse in Miami echoes a similar event in Minnesota over a decade ago, one of the first signs of America’s growing infrastructure problem.

As gaming becomes the dominant form of entertainment this century, game developers increasingly track player behavior to tailor experiences that will keep people playing longer and spending more money.

We’ve teamed with PBS’ American Experience to take a look back at Freeman Dyson, who explored whether interplanetary space travel could be made possible by harnessing the power of a nuclear bomb.

Catastrophic accidents at power plants have heightened fears about the safety of nuclear energy, but environmentalists and others are giving it renewed attention as a way to fight global warming.