Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: What is AI?

Author: Dan Homan

Date: 06:25:54 12/17/99

Go up one level in this thread


On December 16, 1999 at 21:17:46, Dann Corbit wrote:

>A similar thread brings up an interesting question, "What is AI?"
>
>An old test was supposed to be that if we are playing a remote opponent we can't
>tell if it is a human or a machine.  I think that can probably be achieved now
>(especially if we throw in a bit of randomness).
>
>Then what tends to happen is that we say, "That's not really artificial
>intelligence.  After all, it's just a machine, so it _can't_ be."  We simply
>move the target and we are safe from the encroachment of the machine into "our"
>domain.

I don't think it is a matter of limiting the domain.  I think it is a matter
of starting with poor definitions.  We keep trying to define intelligence
in terms of what humans do and don't do.  Functional definitions are very
difficult - particularly when limited to a single task or a restricted set of
tasks.  Just because we use intelligence to do a particular thing
doesn't mean that intelligence is required to do that thing.

In my view, intelligence isn't limited to humans - animals clearly have
intelligence.  Other primates, whales, and dolphins are obvious examples,
but I have even seen adaptive learning and  problem solving skills in my cats
and dogs.  Do these animals approach our level of ability?  No, but I think it
is a matter of degree and not a matter of kind.

IMHO, intelligence is a kind of generalized leaning and problem solving ability.
Higher levels of intelligence impart the ability to abstract from direct
or even indirect experience which makes learning and problem solving
extremely efficent.  A key hallmark of intelligence seems to be the
ability to adapt to new situations based on past experience (rather
than pre-programmed instinct).

Computers are a generalized problem solving tool.  Clearly they could
act as an intelligence if properly programmed.  Do computers act as
an intelligence?  We need to define intelligence properly and then
we will know.  With the definition I proposed in another thread,
some computer chess programs (like KnightCap) would display a
limited form of intelligence in their ability to generalize from old
experience to new experience in a useful way.

The fact that an algorithm can generalize from past experience (even
if generalize is used here in a rather limited sense) is extremely
interesting.  It tells us (that at least in principle) there is nothing
special about our ability to generalize from our past experience.

 - Dan


>
>If (for instance) I was playing some opponent using Winboard and I only knew it
>was one of:
>"Kasparov"
>"Deep Blue"
>
>I would have no way of guessing which was which, since either one would pound my
>stuffings out effortlessly.
>
>So the question stands, "What is AI?" and along with it, the related question,
>"Are chess programs intelligent?"



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.