Re: Challenge to the behaviourists, #1

From: Bill Modlin (modlin1_at_metrocast.net)
Date: 09/19/04 

Date: Sat, 18 Sep 2004 23:05:35 -0400

"Glen M. Sizemore" <gmsizemore2@yahoo.com> wrote in message
news:20040917190942.325$Gy@news.newsreader.com...

> ... What is absurd is to
> explain "intelligence" by saying that the insides are, or could be,
> "intelligent." One does not explain higher-level observations by
saying that
> the observed, macroscopic properties are actually properties of the
level to
> which one is reducing the higher-level. That's like saying that
mercury
> atoms are silvery and slippery...It's stupid - just like you.

Since your gloss of what we are saying does not make sense, perhaps you
could consider the possibility that you are incorrectly interpreting our
words and look for a more sensible reading? Or are you more interested
in calling people stupid than discussing the issues?

In any case, I'll try to explain more clearly.

First, I think we all recognize that "intelligence" is a word with many
uses and meanings, so that any number of confusions can arise just
because people are using different senses of the word.

Even when we try to pin it down by giving a definition at the start of a
discussion, we may run into difficulty because it isn't the sort of
thing one can observe directly, like slipperiness or silveryness.
Intelligence, at least of the sort that I'm interested in talking about,
is a property of behavior-generating mechanisms that tends to produce
appropriate and effective context-dependent behaviors. We infer the
existence of intelligence from observations of the sorts of behavior
that we associate with the word, but none of those behaviors can be
taken as definitive. It matters not only what behavior is observed,
but why that behavior occurred, and in general for animate systems we do
not have access to the details of the process whereby a particular
behavior was generated to be certain whether it was an example of
intelligence at work, or something else... such as intensive training
for that specific behavior, or simply good luck.

Since intelligence is, when I use the word, a property of the internal
mechanisms which generate behavior rather than of the behaving organism,
Glen's snide remarks about slippery mercury atoms completely miss the
mark. To say that an organism "is intelligent" is a shorthand for
saying that its internal mechanisms for processing sensory data and
potentially generating behavior use appropriate procedures to do so...
that its methods are intelligent. It is then quite appropriate to
discuss those internal mechanisms and their principles of operation, and
to ascribe intelligence only when the right methods are used.

Of course, just what constitutes "the right methods" is yet to be
determined. That's the point of AI research, to discover what sort of
mechanisms are necessary to generate the wide range of behaviors from
which collectively we infer the existence of something loosely called
"intelligence".

The specific point that I was trying to make, which inspired the
slippery atoms jab, is that one can in principle program a computer to
emulate any specific pre-specified pattern of behaviors in response to
any similarly pre-specified set of detectable conditions... without
necessarily imbuing the program with any vestige of intelligence. So
we can have a computer (or a computer driven robot) which might do many
things that we would take as evidence of intelligence in a human, and
yet know that the system was still not at all intelligent. The classic
example is chess. We generally consider that playing a good game of
chess suggests intelligence in a person, yet few would wish to call Deep
Blue intelligent, since we know how it achieves grand-master level chess
playing and also know that the mechanisms it uses would be utterly
irrelevant to most other tasks.

I suppose it is possible that some here do define intelligence by the
production of specific behaviors, without reference to how or why those
behaviors were generated internally. If so, we have nothing to talk
about. I wish them well, but would not expect anything they say to be
relevant to my concerns.

Relevant Pages

  • Re: Challenge to the behaviourists, #1
    ... That's like saying that ... thing one can observe directly, ... mechanisms which generate behavior rather than of the behaving organism, ...
    (sci.cognitive)
  • Re: Challenge to the behaviourists, #1
    ... >Since your gloss of what we are saying does not make sense, ... >is a property of behavior-generating mechanisms that tends to produce ... What "internal mechanisms" are there ... I prefer to understand intelligence in terms of information theory, ...
    (sci.cognitive)
  • Re: Challenge to the behaviourists, #1
    ... That's like saying that ... response under stimulus control of behavior and its context. ... mechanisms which generate behavior rather than of the behaving organism, ... "specific pre-specified pattern of behaviors in response to any similarly ...
    (sci.cognitive)
  • Re: Landis and the Keystone Kops
    ... Scott said, "In my years at the UCLA lab, I've never seen anything ... selected by Landis to observe the retesting] demonstrates the lack of ... limiting our observation of the retesting is an example of one of the ... course he's saying what I've been saying all along. ...
    (rec.bicycles.racing)
  • Re: Vista beta vs. OSX
    ... >> tree falls in the forrest and nobody hear it, ... > so your position is that if we can't observe something directly, ... "Innovation is not about saying yes to everything. ...
    (comp.sys.mac.advocacy)
Quantcast