AI Computer Beats Top Go Player
Artificial intelligence researchers say they've made a major breakthrough. A computer program was able to top a professional human player of the board game Go for the first time.
The details of the defeat were released on Wednesday in the science journal Nature. The program is called AlphaGo, and was developed by Google.
Computers have topped the best human players at most games seen as a measure of human intellect, including chess and backgammon. But with Go, humans had managed to maintain the upper hand until now.
Go is played with round black and white stones on a square grid. The point of the game is to occupy as much territory as possible.
With 361 intersections on the board, Go is extremely complicated compared to chess or the Japanese game called shogi. And it's the complexity that makes it such a great challenge for artificial intelligence researchers.
The program relies on what is called a "deep learning" approach. The software learns from the game and comes up with new collections of strategies.
"In Go, the computer needs to recognize situations and understand the meaning of patterns, which is difficult even for humans," said Takeshi Ito, an assistant professor of at the University of Electro-Communications. "That's why Go is a good example for developing (artificial intelligence)."
AlphaGo's next challenge will be to take on one of the world's top players in March. For the researchers though, they're looking even further down the road.
"We're excited about it, but it is just one on the ladder towards solving artificial intelligence," said Google DeepMind CEO Demis Hassabis.
Google researchers say they hope to eventually use artificial intelligence for real-world applications, including disaster prediction and medical sciences.