RE: IQ testing for half-baked AI's

From: Ben Goertzel (
Date: Sat Jul 28 2001 - 15:06:51 MDT

> At 2:00 PM -0400 7/28/01, Ben Goertzel wrote:
> >What would be valuable, I think, is if we could agree upon a
> series of "IQ
> >tests" for baby AI's, that we could use to objectively assess which ones
> >were more promising and generally should have more attention
> paid to them.
> >I also think it may be valuable to each AI developer to see how others'
> >systems perform on the same tasks that his system is working on.
> As I understand it, intelligence is something a process has or
> doesn't. What you want us to test for is smartness. While knowing
> how smart a process is, that doesn't really clarify if it is an
> intelligence or not.

Terms like "intelligence" and "mind" represent folk-psychology concepts,
whose meanings are ambiguous and multifaceted. It's not reasonable to
expect any reasonably compact mathematical definition to fully capture the
sense of such a term, nor any reasonable executable experimental protocol to
measure a quantity association with such a term.

Further, it is it clear that these folk concepts, precisely as commonly
considered, are precisely the concepts one needs for scientific analysis.

Our aim with formalizations of intelligence, mind or related concepts, or
tests desired to measure these things, should be to create scientific
concepts which capture significant aspects of the associated folk concepts
with which they share names -- and which we believe are useful for science
and engineering, perhaps more so than the corresponding folk concepts in
their full ambiguity.

So, to me, quibbling over what's intelligence and what's smartness isn't all
that interesting. We need to devise practical tests to allow us to detect
whether our AI systems are on the right track or not, and objectively
compare different Ai approaches -- whatever words you want to use to
describe this...


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT