LAST OCTOBER, a computing system developed by DeepMind, Google’s artificial intelligence company, beat the European champion player at the game of Go, an ancient Chinese boardgame. In March, the machine -called AlphaGo -will play against Lee Sedol, the current world champion of the game. The matches will take place beginning March 9, and will be broadcasted live on YouTube.
So what’s the big deal? Computers have been beating humans at boardgames like chess for many years (in 1997, for example, IBM’s Deep Blue machine defeated Garry Kasparov, the then world chess champion), sure, but Go is different. The number of possible positions in the game outnumber the number of atoms in the universe, so computers cannot play their old trick of analyzing every possible combination and choosing the best one. Intuition made humans better Go players than machines, whose computing abilities were not enough to win. Until now.
DeepMind’s approach uses deep neural networks, making it possible for the machine to learn, get better, and develop something similar to our intuition, something AI experts thought wouldn’t happen so soon -they expected at least another decade to pass before we saw it. Will AlphaGo manage to defeat Lee Sedol in March? The player says he is confident he will win. Everyone else is starting to doubt where to put their money.