RE: Computers that improve themselves

From: Ben Goertzel (ben@webmind.com)
Date: Mon Apr 23 2001 - 10:00:14 MDT


> In other words, I think using evolutionary techniques to discover new
> possibilities in technology is a great thing, but the new possibilities
> should be thoroughly understood and then incorporated into a
> human-designed
> final product. Using a raw evolution-derived piece of technology without
> completely understanding how it works is a recipe for disaster IMO.

My take on this is a little different.

In the WM AI Engine, one has a designed system which has components that are
evolved internally. The evolved subsystems operate in ways that are NOT
thoroughly understood, in each particular case. But the role that they play
in the overall system are thoroughly understood. I think this is the only
way to do it. Whether you explicitly call part of your AI dynamic
"evolutionary" or not doesn't matter, but there's no way to do cognition
without some kind of combinatory, stochastic mechanism for generating new
candidate procedures and concepts, and then selecting them based on their
inferred or experienced effectiveness at achieving certain goals. This is
evolution, a key part of cognition.

ben



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT