X-Message-Number: 25573
Date: Mon, 17 Jan 2005 11:30:21 -0500
From: Francois <>
Subject: One unsolved problem in Artificial intelligence

I think we will not succeed in creating true artificial intelligence until
we figure out how brains model reality. By that I mean how are objects and
events represented in such a way that they are recognizible in most
circumstances. Take a chair for example. I have seen hundreds of different
chairs in my lifetime, yet I have no trouble in identifying them as chairs,
nor do I have much trouble in deciding if a new object I see is a chair or
not. My recognition of a chair is not really influenced by the way it's
oriented. I will recognize a chair if I look at it from above, or below or
from any other direction. I will identify a chair regardless of where I see
it, be it in a living room, a jungle, a desert or even floating in the
ocean. I can even recognize a chair by touch if need be. This indicates that
our brains use incredibly flexible, yet remarkably reliable techniques to
model the outside world. And Peter Merel's story about butterflies returning
to the same blade of grass after their mating dances illustrates that
something as 'simple' as an insect's brain is capable of the same
recognition feats. We need something like that to make true general AI, and
unless some breakthrough has recently happened, we are still, in my opinon,
light years away from it.

The Devil fears those who learn more
than those who pray

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=25573