“Can a robot write a symphony? Can a robot turn a… canvas into a beautiful masterpiece?”
This is the question Will Smith’s character, Detective Del Spooner, asks Sonny, a robot with highly sophisticated artificial-intelligence, in the 2004 movie I, Robot.
Sonny, in a later scene, concedes, “You were right detective, I cannot create a great work of art.”
For those of you who haven’t yet heard, AlphaGo is an artificial intelligence program developed by Google’s DeepMind team. Similar to IBM’s Deep Blue computer, which was the first computer to defeat a chess world champion, AlphaGo is focused on mastering the game of Go. The difference being that Go has a nearly infinitely higher level of complexity with trillions of possibilities in the course of a game.
It is a game where judgment and intuition is key, so it was estimated that accomplishing the Go equivalent of Deep Blue’s victory over Garry Kasparov was still at least a decade away. But, it happened.
This past week, AlphaGo defeated the reigning Go world champion, Lee Sedol. And not only did it beat him, it swept the first three games in the five game match.
And I have to concede, that when I heard about this, my reaction was exactly the same as that of Det. Spooner. I essentially shrugged and just said, “Yeah, but can it paint?”
But as I read more about the matches, one word in particular stood out in how commentators described the matches. More and more often, analysts of the game were calling AlphaGo “creative.”
And that piqued my interest. How could a robot be considered creative?
I even Googled (ironic, I know) the definition of the word just to be clear.
relating to or involving the imagination or original ideas, especially in the production of an artistic work.
It just didn’t make sense that an artificial intelligence, which has its thought process determined by the patterns of previous events and outcomes, could be creative, which by definition must involve the imagination or original thought.
So why were commentators calling it creative?
They were saying that the moves that AlphaGo was making were moves that human players were highly unlikely to play. This was either because human intuition would lead a player to believe that such a move was illogical or because it was something that a human being would be taught not to do in formal instruction.
And that struck me because it brought up a scene in my mind of a kid playing chess with an older gentleman just trying moves. And each time the kid made an illegal move, the older man would put the piece back and say, “Sorry kid, that doesn’t work.” But after a while, the kid would make a play that wasn’t only legal, but actually quite brilliant and the gentleman wouldn’t able to really say anything except to just remark “wow.”
And when you think of AlphaGo as a super juiced up version of that kid, it becomes a lot clearer and, to me at least, a tinge sadder.
It becomes clearer because, like the kid who doesn’t know the rules, AlphaGo can stumble upon a creative and brilliant move without necessarily making the same mistakes that a kid would simply by doing something totally unexpected.
But it also seems sadder because perhaps the reason Lee lost to AlphaGo in those three matches was because he was the one that was not so creative. Perhaps the rigidity and formality of the instruction he was given when he was first being taught to play Go closed him off to the sort of creative and unexpected moves that AlphaGo, which had not gone through the same level of instruction, was open to. Maybe Lee, in that sense was the more “programmed” of the two.
And I think that this argument has merit because in the fourth game between Lee and AlphaGo, Lee made an unorthodox move that forced AlphaGo into a mistake and eventually led him to defeat AlphaGo for the first time.
Now that is not to say that I am convinced that A.I. can be creative. I still believe that, because AlphaGo’s “thought” process is determined by previous events and calculations, rather than original thought, it can never truly be creative. But I think it has shown me how really narrow the gap between imitation and real thing can be.
And if there is one thing I take away from this, it will be to let my future kids (however many years in the future that may be) explore the full range of possibilities and not dismiss the “illogical” moves in the way I imagine Lee Sedol’s instruction told him to. That way, when the robots revolt and try to surprise us in the future, they’ll be ready.