Match 1 - Google DeepMind Challenge Match: Lee Sedol vs AlphaGo - YouTube
Today a computer defeated the world champion in one of the most complex games created. Commentators will no doubt weigh in on the technique and meaning of the victory, and there remain several games left in the tournament, but before they do it is worth considering why this challenge is different from all others.
Go is a game with a branching factor far in excess of what computers can handle through brute-force methods. The number of possible configurations of the board exceed the number of atoms in the universe, meaning it is not possible to build a machine that can retain or explore each outcome. Human players (being entities in that same universe) suffer from the same limitation, and so cannot play Go analytically. Instead, they must play “intuitionally,” using learned patterns to prune the search tree and relying on “gut” to point their minds towards optimal solutions. “Intuition” and “gut” are the kinds of things we consider to be the unique province of human intelligence. Today a new contender lays claim to such capacities, and it is far from human.
It may be tempting to hail our new Go-playing overlords, but I believe such resignation betrays a conceit over human minds. It is not necessary to ascribe unfathomable and terrifying powers to a machine taking its first steps into the realm of cognition. It may rather be more accurate to downgrade our notion of what constitutes an intelligent mind.
If all of humanity submits questions to your search engine, the most surprising observation (to me) is how rather predictable humans are. All the thoughts of all the humans in all situations is a spectrum so wide we consider it to be effectively infinite and abandon all hope of ever corralling it into a formal procedure. Yet each human can have only a finite number of thoughts (however defined), and there are a finite number of humans, and thus all the thoughts of all of humanity is also a finite set whose size relative to modern compute capacity may not even be terribly large. Without grappling with every possible outcome, a finite number of humans with finite brains navigate unfathomably many complex situations with the very same “intuition” and “gut” we seem unwilling to ascribe to machines. Yet how complex can those notions truly be, being as they are the product of individual minds that are small relative to humanity?
The days when computers routinely defeat humans at Go are less interesting than the point in history when we realize that it was never a fair fight to begin with.