Google's AlphaGo isn't taking over the world, yet – CNET
A supercomputer programmed by Google just beat a human at a really complicated game. But don’t worry about the computer winning at anything else.
On Saturday in South Korea, a Google artificial-intelligence program dubbed AlphaGo beat world champion Lee Sedol in Go, an ancient and complex board game in which strategy and tactics collide with intuition and cunning.
“When I look back on the three matches, even if I were to go back and redo the first match, I think I would not be able to win because I misjudged AlphaGo,” Lee said at a postgame press conference on Saturday, following his third straight defeat.
Against the odds, Lee came back and beat Google’s computer the following day, but it wasn’t enough to tip the scales of the contest.
The Google DeepMind Challenge, which has taken place at the Four Seasons Hotel in Seoul, has drawn tens of thousands of online spectators who have followed the matches live on YouTube. Streams of the software versus wetware competition have received more than 3 million views since the contest started on Wednesday. Though Lee has officially lost the best-of-five contest, he will play one more match to establish a final score in the face-off, which ends Tuesday.
The public interest wasn’t piqued solely by the popular game, which is played widely in Japan, China and Korea. For many, Google’s success raises questions about how the relationship between man and machine will evolve. After all, if Google’s software can win a game that hinges on little more than feel, won’t it someday be able to do something less complex, like your job?
Not anytime soon, say computer scientists. Despite their fast advances, robots remain too woefully single-minded to give humanity a giant pink slip.
Oren Etzioni, the CEO of the Allen Institute for Artificial Intelligence, says teaching computers to read and then answer questions about that content is still a work in progress. For example, when Etzioni had computers read eighth-grade science texts, they could answer only about 60 percent of the questions on a test.
“Understanding a single sentence can be a lot more complicated than playing Go,” Etzioni says. Computers, he says, have yet to demonstrate that they “can solve fuzzier problems where things are more nuanced.”
Go’s complexity is what drew Google’s artificial intelligence team to it. Facebook is also working on a program that can play the game.
Go, which originated in China thousands of years ago, is played on a 19×19 grid with black and white stones. The board’s size means the number of possible moves is greater than the number of atoms in the universe, according to Google, making it a more difficult programming challenge than chess.
Artificial intelligence expert David Levy says Go’s complexity makes Google’s win a bigger victory than IBM’s 1997 triumph over world chess champion Gary Kasparov.
“Singularity is much closer than most people previously thought,” said Levy, referring to the mooted moment in computer science when computers outpace humans. Levy has twice won the Loebner Prize, an award for creating the most humanlike computers.
Nolan Bushnell, founder of Atari, was also impressed by AlphaGo’s feat.
“Go is the most important game in my life,” Bushnell said. “It’s the only game that truly balances the left and right sides of the brain. The fact that it has now yielded to computer technology is massively important.”
Still, Google’s Go victory doesn’t mean the end of humanity as we know it is nigh.
“For all its difficulty, Go is still an artificial problem with very simple rules,” says Pedro Domingos, a computer science professor at the University of Washington. “Building a home robot, for example, is something of a different order of magnitude altogether — the robot needs common sense, physical dexterity, etc, which are all still sorely lacking in AI.”
There you have it. Your job is safe for now.
Updated at 11:48 a.m. PT on March 13: adds results of fourth contest.
Read this article: