Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!bloom-beacon.mit.edu!news!news.media.mit.edu.!push
From: push@media.mit.edu (Pushpinder Singh)
Subject: Re: Churchland takes on Dennett
In-Reply-To: rschoen@nyc.pipeline.com's message of 28 Jul 1995 00:43:38 -0400
Message-ID: <PUSH.95Jul28221046@media-lab.media.mit.edu>
Sender: news@media.mit.edu (USENET News System)
Organization: M.I.T. Media Laboratory
References: <3v1t2q$3c2@pipe6.nyc.pipeline.com> <806666094snz@chatham.demon.co.uk>
	<3v36qs$jkj@mp.cs.niu.edu> <3v74m4$off@pipe5.nyc.pipeline.com>
	<DCDtzM.Eqx@midway.uchicago.edu><806864301snz@longley.demon.co.uk>
	<3v9ptq$f2m@pipe5.nyc.pipeline.com>
Date: Sat, 29 Jul 1995 03:10:46 GMT
Lines: 37


In article <3v74m4$off@pipe5.nyc.pipeline.com> rschoen@nyc.pipeline.com
(Robert J. Schoen) writes:
|
| His point is well taken, especially in recent years as Connectionism and
| neuroscience have begun to rapidly approach one another. This is a critical
| point in Churchland's new book; human cognition appears to be pure
| connectionism and there is little evidence for the involvement of serial,
| rule-based processes.

I don't see how this matters.  No one in the world, not even Churchland,
has any real clue about the machinery in the brain at the level between a
few neurons and clumps of millions.  And that's the interesting level,
where the serial processes to which you refer would have to be happening.
So naturally we have little evidence of such processes, since that level is
entirely unexplored.

In a later post Robert writes:
|
| Churchland feels that Dennett's approach is cheating.  There is no need to
| posit any serial device to explain these phenomena, he argues; these serial
| processes can be explained simply with recurrent networks.

If all Churchland is saying is that recurrent networks can perform serial
computations, well that's no big deal at all.  They're perfectly capable of
being equivalent to serial machines.  Recurrent network aren't "serial
devices" nor are they "parallel devices" -- that's not the right way to
think about them.  What's important is the nature of the computation they
are performing, which can either be serial or distributed.

And while I'm not familiar with Dennett's theory, it's title "virtual
serial machine" sure does suggest he doesn't care about the underlying
machinery.  So what if the brain is basically built from recurrent nets
(still a pretty fuzzy idea); they would still be capable of embodying
Dennett's virtual machine.

-push
