Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!panix!zip.eecs.umich.edu!newsxfer.itd.umich.edu!gatech!howland.reston.ans.net!news.sprintlink.net!EU.net!ieunet!hdl.ie!tony
From: tony@hdl.ie (Tony Veale)
Subject: Re: Quit Slagging Penrose
Message-ID: <CyC5xs.Mr4@hdl.ie>
Sender: news@hdl.ie (News system)
Nntp-Posting-Host: saavik
Organization: Hitachi Dublin Laboratory
X-Newsreader: TIN [version 1.2 PL1]
References: <1994Oct27.115326.13522@cs.tcd.ie>
Date: Thu, 27 Oct 1994 14:37:03 GMT
Lines: 103

Frank Kelly (kellyfj@cs.tcd.ie) wrote:
: Hello,
: 	I'd like to agree with the person who proposed that certain
: people should stop criticising Penrose' new book. I've read the
: "Emperor's New Mind" and although it is not a great book (it doesn't
: propose testable solutions to problems in AI) it is better than
: a lot of works out there (I won't name names, except on request :-))

: Lately I have read a few articles (e.g. WCNN'93 & 94) which have
: research results which SEEM to indicate that maybe Penrose was indeed
: right.

: However I would edge more towards Gerald Edelman's thinking (that
: Quantum Mechanics etc. are not really necessary), nevertheless, with Penrose'
: track record in the Physics field I would not immediately down any of
: suggestions before there was *real* research done to verify/refute his
: hypotheses. 


: If Penrose' new book is half as good as the "Emperor's New Mind" then
: I'm going to buy it. He may be wrong but I doubt if he would say things
: unless he had quite a bit of evidence to back himself up, unlike some of
: the "critics" in this group.

: Comments invited.

: --Frank

Gee, am I glad you addressed all of my criticisms directly, rather than
just issuing a blanket statement of disgust ;^)


How about this evidence then? Doug Lenat of the CYC project (which is
variously lauded and criticized, both for good reasons) makes a good
point when he says AI folks have a tendency to suffer from "Physics
Envy". I know that he refers to the desire on out parts to create grand
unifying visions, and he's not attacking the stuff of physics directly,
yet I think his phrase is certainly applicable here. This physics hero
worhsip is risible - I accept Penrose has down damned good work in the
physics field but this does not qualify him to preach computer science
to computer scientists. Why don't you start thinking for yourself, instead
of putting your faith in someone's opinions on the basis of reputation, and
a sexy new paradigm.


I am not saying that physicists don't make excellent computer
scientists - we have them to thank, to a large degree, for the current
revival of connectionist science. I'm with Hopfield on this one, and I
await proper demonstration of i) the necessity for QM in intelligence
(no populist appeal-to-his-sense-of-human-specialness bullshit) and
ii) the workings of QM in intelligence in such a way that invalidates
symbolic/connectionist models.

In my opinion (here I go again) Penrose needs a nice close shave with
occam's razor. Why don't you tell me why conventional arguments against
AI (symbolic and connectionist) don't work, and why we should be looking
for a new paradigm? Don't you believe that the complexity exhibited by
the sheer scale and connectivity of the brain is staggering, and this
in itself is a very good reason to believe that an algorithmic description
of intelligence (not necessarily human) is feasible using current CS 
beliefs?

Criticism makes people tougher and their theories more robust (because
they have to become answerable to the scientific community). It is true
that people have in the past laughed at, and branded as charlatans, such
great minds as Newton (action at a distance - Leibnitz had a belly laugh
at that one), Kepler (what do you mean, elliptical orbits?) and Copernicus
(been sipping the mass wine again) were eventually proved correct. This
doesn't mean we should take an idea seriously just because it flies
in the face of conventional wisdom, quite the opposite. Let Penrose fight
for his beliefs and make us listen if he wants us to (not that this matters;
those of us in the AI community are not qualified to carry out the 
research program Penrose advocates, even if we wanted to). Penrose is 
going to have to fight for respect if he want's to be accepted in another
field (Edelman, for instance, gets quite a bit of opposition from 
traditional AIers, and it only serves to make his arguments that
much stronger). Penrose is the Cold Fusion of AI at the moment, and because
he's sneered at doesn't make him right.

So if you want to be Penrose's champion, then, and wear his ribbon at
the joust (how chivalrous of you!) why don't you explain why his theories
have merit? Clank! There goes the gauntlet hitting the floor. Why don't
you pick it up? All it takes is for you to give us one good example
of a non-computational function performed by humans, and demonstrate
why exactly it is non-computable (no Platonic-reality mysticism please).

Demons begone ...

Tony


__________________________________________________________________________
Tony Veale                                             Tel: +353-1-6798911
Hitachi Dublin Laboratory, O'Reilly Institute,         Fax: +353-1-6798926
Trinity College, Dublin 2, Ireland.                    e-mail: tony@hdl.ie

"Courageous, untroubled, mocking and violent - that is what Wisdom wants us
to be. Wisdom is a woman, and loves only a warrior."

						Fred Nietzsche,
						Zarathustra, book I
__________________________________________________________________________

