Friday, March 11, 2016

Why, all our ranks are broke.

First they came for Tic-Tac-Toe. Then they came for checkers. And they even took down chess. But exhausting the search space for those games was inevitable. More than a googol but apparently not more than a Google.

Go was always the game the computers were never meant to win. Too many moves to brute force or create massive move databases for. But it looks like Google's DeepMind is going to beat one of very best human players.

The initial skeptic in me found it hard to see DeepMind as anything but a kind of savant in the same vein as Deep Blue. At least with Watson's victory over Ken Jennings, the AI applications were more generalizable. But DeepMind, if the writing on the tin is correct, uses neural networks to learn; it is able to play video games by looking at the screen.

So it does seem to be a real breakthrough in the field of AI. But with so many of the means that become available to us, the ends we seek disappoint.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.