Go is a Chinese board game dating back 2,500 years. The game’s rules are fairly simple, with the objective being to capture more territory than your opponent. However, Go is a prime example of a game that is easy to learn, but difficult to master. The documentary boasts that Go offers more board combinations than there are atoms in the universe.
The game’s complexity meant that computer scientists and Go champions alike believed that AI would not be able to master the game for at least another decade. The Google-owned DeepMind company set out to test that theory by creating an AI that could compete with the world’s best Go players. The company spent several years developing the AlphaGo program, teaching it more than 30 million moves from actual games of Go. From there, the program used machine learning to play against itself, testing strategies and developing new moves. This process was crucial to ensuring the program could keep up with professional players, as the game contains so many possibilities.
Despite the program’s victory over European Go champion Fan Hui, both AlphaGo’s creators and Sedol were skeptical of its chances against a world champion. Sedol went into the game expecting to triumph over the AI. His confidence was warranted, considering Sedol has been playing professionally since the age of 12 and has 18 world championships under his belt.
Even the program’s creators were unsure of their creation’s chances. The documentary shows them nervously watching AI’s five rounds against Lee Sedol.
The program’s victory has raised new questions concerning the role of computers and AI in our daily lives as some fear that such technology might one day make humans increasingly obsolete. Of course, such fears are hardly new. There were similar comments when IBM’s Deep Blue program defeated Russian chess master Garry Kasparov.
While man vs. machine makes for a compelling angle and potential ramifications of AI should not be ignored, the story of AlphaGo is really a story of man vs. man. After all, the program is not some sentient intelligence, but, rather, a tool created by Google’s DeepMind team.