'AlphaGo': Film Review

Courtesy of Greg Kohs/Reel As Dirt
An involving sports-rivalry doc with an AI twist.
9/29/2017

Documentarian Greg Kohs follows the team of programmers trying to build an algorithm for the world's most complicated game.

One more milestone in humanity's path to a robo-dominated future is passed in Greg Kohs' AlphaGo, the story of the AI program that mastered the most complicated board game ever invented. Taking pains to say the achievement it chronicles is still a long way from Skynet, the doc does persuasively convey the shock with which men and women suddenly realize they have been overtaken years before they expected to be. Involving and poignant if sometimes less informative than it might be, the doc should have appeal on cable after its short theatrical run.

Go is the millennia-old Chinese game played with white and black stones on a 19-by-19 line grid. Amazingly, it is said that there are more possible configurations of that board than there are atoms in the observable universe; all the computers in the world could work for a million years and still not evaluate every possible move in a game.

So, while the things some experts say in the doc's intro scenes — playing Go is "like putting your hand on the third rail of the universe," one claims — are needlessly overblown, writing a computer program to play this game makes writing the Garry Kasparov-beating Deep Blue chess program look like coding Pong.

The Google-owned DeepMind company, "an Apollo kind of program" for artificial intelligence, decided to take on this challenge by using neural networks. Rather than having Go players teach a computer how to play, they set things up so the program could learn for itself. How this works is, unsurprisingly, a challenge for the film to explain. But we quickly see the results, as a former European champion player, Fan Hui, is recruited for a five-game tournament against the AlphaGo program and loses them all.

Response to this news in Asia was skeptical, claiming Fan was not a sufficiently impressive player. So the DeepMind team arranged a 2016 match in Seoul against one of the world's best players, Lee Sedol. This is the conflict occupying the bulk of the film, and those who don't know the result will do well to remain ignorant until viewing. But this doc wouldn't exist if, as everyone expected would happen, Sedol wiped the floor with the computer.

Setting aside the difficulty of explaining advanced computing to a lay audience, Kohs doesn't do as well as he might have at chronicling the games themselves. He ignores some details attentive viewers might wonder about (like the handicap, or "komi," that figures into a game's score), and as the games unfold he almost never visualizes how a given move affects play. Non-Go players must rely on the expressions — sometimes astonished — of TV commentators doing play-by-play if we want to understand how the contest is going.

Another perspective comes from watching the face of our human champion, a soft-spoken and introverted young man; and listening to the chatter back in the DeepMind suite, where the programming team raptly monitors all the calculations that lurk behind the algorithm's every choice.

But even watching all three of these parties doesn't unlock the mysteries of the "wrong" moves AlphaGo makes that turn out right in the end. By Game Four, Lee has adapted, making head-scratching decisions of his own that sometimes work out. In the end, observers wonder if AlphaGo's odd variety of intuition might not kill Go as an intellectual pursuit but shift its course, forcing the game's scholars to consider it from new angles. So maybe it isn't time to welcome our computer overlords, and won't be for a while — maybe they'll teach us to be better thinkers before turning us into their slaves.

Production company: Moxie Pictures
Director-director of photography: Greg Kohs
Producers: Gary Krieg, Josh Rosen, Kevin Proudfoot
Executive producers: Robert Fernandez, Dan Levinson
Editor: Cindy Lee
Composer: Volker Bertelmann

In English and Korean

90 minutes

comments powered by Disqus