We humans all got a little nervous back in the late 90s. The IBM-engineered artificial intelligence, Deep Blue, (in)famously beat the world champion and Grandmaster, Garry Kasparov, at that most intellectual of human passtimes, chess. Kasparov sensed human creativity in Deep Blue’s moves, though. He accused IBM of foul play, demanding a rematch. The company refused him, promptly dismantling Deep Blue, and humanity’s fears that the first step toward machine dominance would be a knight-to-queen-checkmate were put to rest. For a while, anyway.
20 years have passed and–to state the obvious–machines have grown exponentially smarter. It wouldn’t be surprising if Kasparov’s phone could defeat him, now. Of course, engineers are never ones to leave well enough alone. The sharp minds at the Google think tank, DeepMind, have trained the latest generation of A.I. in Go, a game infinitely more complex than chess. “Infinitely” isn’t hyperbole, either. As their rep explains in the video below, Go has 180 more options available for any given move, so the number of possible board configurations are more numerous than all the atoms in the universe. If that is the case, you’d need an extra-intelligent A.I. to play it at an expert level, no?
Enter: AlphaGo, a game-playing machine that’s like a T-1000 to Deep Blue’s classic Terminator.
Since this video was recorded, AlphaGo has, in fact, followed-up its victory over European champion Fan Hui by beating Lee Sedol, a top-ranked pro Go player. The man had a shot to single-handedly save mankind from Skynet work camps, but he lost three matches in a five-game set against the machine and forfeited his last. So… go humans? You can watch all four hours of his first losing match against AlphaGo here.
Oh, and for those unfamiliar with Go: the game is not only more complex than chess, it’s also significantly older. It started in China, some 2500 years ago, and has since spread throughout Asia, being known as “the circling game” in some languages. One player is given a bowl of a black “stones,” the other a bowl of white ones, and they alternately cover the intersecting “points” of a 19×19 grid in a continual effort to encircle and capture each other’s pieces. There’s no checkmate equivalent, however. A game concludes, and pieces are counted, whenever the players no longer wish to make additional moves. Sorta like when us lowly humans eventually give up in the face of our synthetic overlords.
Any Go players here? Figure you could take on AlphaGo? Does the gaming prowess of A.I. make you hopeful for tomorrow, or fearful of the future? Sound off in the talkback.
Featured Image Credit: DeepMind