Google Just Made a Giant Leap in Artificial Intelligence
A human has lost a game of Go to a computer for the first time in 2.5K years
By Michael Harthorne,  Newser Staff
Posted Jan 27, 2016 4:47 PM CST

(Newser) – It seems Skynet is one step closer … if all the Terminator wanted to do is play a 2,500-year-old game. On Wednesday, Google DeepMind announced the creation of an artificial intelligence that successfully beat a professional Go player five games in a row, Wired reports. According to a study published in Nature, it's "a feat previously thought to be at least a decade away." Go was once believed to be impossible for a computer to master due to how many moves are possible, Quartz reports. According to Wired, there's an average of 35 potential moves per turn in chess; with Go, it's 250. “There’s more configurations of the board than there are atoms in the universe,” DeepMind's CEO says in a video. Even supercomputers aren't fast enough to analyze every potential outcome with that many moves, according to Quartz.

So DeepMind created an AI—AlphaGo—that thinks more like a human. It's composed of two neural networks—"computer systems that are modeled after the human brain," Quartz reports. One is responsible for analyzing its position on the board, the other for choosing a move that "feels right." According to Wired, AlphaGo learned 30 million moves by observing the world's best Go players, then played itself millions of times to learn new moves that could beat the best. Last October, AlphaGo took on the European Go champion and crushed him, Quartz reports. In March, it will face off against the world champion. One expert tells Wired the ramifications of AlphaGo are huge because it can be applied to "anything that you can conceive of as a game," including "war or business or [financial] trading."