From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!bronze!chalmers Fri Jan 31 10:27:11 EST 1992
Article 3285 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!bronze!chalmers
>From: chalmers@bronze.ucs.indiana.edu (David Chalmers)
Subject: Re: Two Materialisms (Was: Re: Strong AI and Panpsychism)
Message-ID: <1992Jan30.035035.27686@bronze.ucs.indiana.edu>
Keywords: ethics,functionalism,materialism
Organization: Indiana University
References: <1992Jan29.170943.4706@psych.toronto.edu> <1992Jan29.210141.26133@cs.yale.edu> <1992Jan30.023713.11333@cis.ohio-state.edu>
Date: Thu, 30 Jan 92 03:50:35 GMT
Lines: 116

In article <1992Jan30.023713.11333@cis.ohio-state.edu> chandra@boa.cis.ohio-state.edu (B Chandrasekaran) writes:

>I thought you were going to say, ".. in terms of
>mechanisms that we already understand, such as gravitation, quantum
>mechanics, electromagnetics, etc.," because those are the mechanisms that
>explain the behavior of *matter*, whose laws materialism presupposes
>are sufficient to explain minds.   In fact, "real" materialists -- such
>as the Churchalands, who are called, I think, eliminative materialists --
>disdain functionalism of the sort Drew McDermott subscribes to as 
>not materialism at all, since the causal stories he would tell for
>minds will never refer to the laws of physics.   For example, we never
>mention any law of physics in explaining Turing machines, feedback loops,
>winner-take-all networks, etc.   

I think this slightly mischaracterizes the Churchlands' position.  They
don't object to appealing to functional organization when it's
appropriate (and indeed appeal on occasion to feedback loops, matrix
multiplication, etc), but insist that this organization be directly
present in the brain.  One could call them "low-level functionalists",
as opposed to the high-level functionalism of symbolic AI.  Mostly
they are opposed to representational talk, especially talk couched
in the idiom of propositional attitudes (belief, desire, etc).

>It appears that there are three realms: the realm of matter, the realm
>of representations, and the realm of qualia/intentions/consciousness,
>not just two: matter and consciousness.

I like this distinction, although I think there might more naturally
be four realms to distinguish.

1. Physical realm: atoms, quantum physics, gravitation, electromagnetism, ...

2. Functional realm: feedback loops, networks, Turing machines, ...

3. Representational realm: knowledge, belief, desire,...

4. Consciousness realm: consciousness, qualia, pains, sensations, ...

On the various relations:

The functional realm is uncontroversially reducible to the physical realm.

Modern cognitive science holds that the representational realm is
reducible to the functional realm, and I agree, although some, e.g.
Searle, disagree, holding that it is only reducible to the physical
realm, and no doubt you could find some who don't accept even
that (e.g. people who hold that representation is dependent on
consciousness, and that consciousness isn't reducible to the physical;
Mikhail Zeleny may fit this description).

Consciousness is where things get tricky.  Some people think that
consciousness is reducible to representation (e.g. Dennett, Harman,
and possibly many cognitive scientists, though cognitive scientists
don't talk much about consciousness).  Many more think that
consciousness is reducible to function (Lewis, Armstrong, McDermott,
most functionalists, etc), but not necessarily to representation.
Still others (notably Searle) thinks that consciousness is reducible to
the physical, but not to the other realms.  Incidentally Searle also
differs from most others in holding that representation is itself
dependent on consciousness.

Penrose may in some strange sense hold that consciousness is reducible
to function, but believes that the right function can only be realized
through quantum effects (but he's so vague that one can't know).  A
Cartesian dualist like Eccles doesn't think that consciousness is
reducible to physics, but might just possibly think that it's reducible
to function that happens to be realized by ghostly spirits.

Then there are people like me who can't see how consciousness could
be reduced to physics (and consequently, to function or representation
either), though it would be nice if somehow it could be.

Finally there are people who don't think that consciousness really exists
as a phenomenon worth talking about (perhaps Minsky, and Dennett on
some days), so there's no point reducing it to anything.  Personally
I think that apart from the strong anti-reductionist position, this
is the only position that's at all tenable, although it does seem
to be prima facie implausible.

>In my paper on connectionism in AI Magazine a few years ago, I argued
>that part of the excitement surrounding connectionism in its early
>stages was the claim of skipping the second realm, i.e., it was more
>directly physical and not representational.   Most connectionists
>quickly abandoned this notion, since they came to see that it was
>every bit as representational as the symbolic systems, only different.
>Edelman seems to think he is offering a purely physicalist account,
>without any representational intervention, but I think there is
>representation there nevertheless.

I think that the enlightened connectionist position has always been
that the cognitive realm (which is more or less the representational
realm) is reducible to the functional realm -- you don't find many
connectionists talking directly in terms of physics.  It's just
that they want to reduce it to functional mechanisms quite different
from those used by traditional AI and cognitive science.

To be strictly accurate I should probably introduce the cognitive
realm as a fifth realm -- i.e. the realm concerned with the explanation
of behaviour -- and not beg any questions about the identification
of this with the representational realm.  Some cognitive scientists
want to skip the representational realm altogether, and go straight
from the cognitive to the functional (e.g. Brooks, Freeman, the
Churchlands on some days, and possibly some of the connectionists you
mention).  Most cognitive scientists, by contrast, take it for granted
that cognitive explanation will be representational explanation; or
more accurately, jointly representational and functional explanation.
A few people (Fodor, Pylyshyn, ...) might hold that one can perform the
reduction to the representational realm fully before a reduction to the
functional realm, but even these people will rely on nonrepresentational
functional explanations for some cognitive phenomena, e.g. perception.
But I'd better stop now.  The space of realms is getting cluttered.

-- 
Dave Chalmers                            (dave@cogsci.indiana.edu)      
Center for Research on Concepts and Cognition, Indiana University.
"It is not the least charm of a theory that it is refutable."


