From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!mcsun!uknet!edcastle!aisb!aisb!philkime Tue Jan 28 12:17:44 EST 1992
Article 3146 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!mcsun!uknet!edcastle!aisb!aisb!philkime
>From: philkime@aisb.ed.ac.uk (Philip Kime)
Newsgroups: comp.ai.philosophy
Subject: Re: Penrose on Man vs. Machine
Message-ID: <1992Jan25.195815.21938@aisb.ed.ac.uk>
Date: 25 Jan 92 19:58:15 GMT
References: <1992Jan24.182929.8626@aisb.ed.ac.uk> <1992Jan25.170045.3587@csc.canterbury.ac.nz>
Sender: news@aisb.ed.ac.uk (Network News Administrator)
Reply-To: philkime@aifh.ed.ac.uk ()
Organization: Dept AI, Edinburgh University, Scotland
Lines: 52

In article <1992Jan25.170045.3587@csc.canterbury.ac.nz> chisnall@cosc.canterbury.ac.nz writes:

>From article <1992Jan24.182929.8626@aisb.ed.ac.uk>, by philkime@aisb.ed.ac.uk (Philip Kime):
>> The 'Dreyfus argument' tends to be accepted by degrees in the SF world.
>> I know many SF's who still adhere to SF but are, by differeing degrees,
>> wary of the Dreyfus problem. There are many who 'accept' at least some
>> of Dreyfus's points but who remain SF's. Whether this is consistent is
>> another matter and most certainly is down to whether the person in
>> question is a Strong AI enthusiast.....weak AIers will take what they
>> can get in terms of application etc. but Strong AIers are almost
>> duty-bound to give up SF in the face of a even a little acceptance of
>> Dreyfus's points.
>
>Would it be possible for you to summarise Dreyfus' argument for those of us
>who haven't heard of it?  References?

Sorry....you're right. Dreyfus's point comes across well in Dreyfus and
Dreyfus (1986 I think) paper called 'Making a Mind versus modelling the
Brain: AI back at a branch point'. This is in Boden's Philosophy of
Artificial Intelligence (OUP) (1981 I think...years are not my strong
point) or I can get the original reference if you can't get hold of this
book (it's very commonly in libraries). Basically, Drefus argues that SF
comes up against the problem of phenomenology... of trying to create of
theory of everyday experience grounded in 'atomic' aspects of this
experience. This has a very compositional flavour and the problem that
the great phenomenologists came up against (Husserl etc.) was that even
seemingly simple areas of experience required vast amounts of background
knowledge (i.e. other knowledge) in order to make any sense. It is like
trying to create a set of first-order predicate calculus expressions to
represent all of the knowledge needed to, say, make a sandwich. Think
about it...regardless of the adequacy of first-order calculus in
representing certain aspects of this scenario, the amount of expressions
you would have to write down in order to implement this knowledge in a
SF machine is vast. Try it. It's so vast a task in fact, Husserl,
Winograd (remember he had tried this with SHRDLU) and Dreyfus think that
it's either actually impossible or pragmatically intractable. Trying to
make all this knowledge explicit involves a compositional approach where
you define the atoms of experience and then define transformations to
more composite experience is a very ambitious program indeed. i.e. going
from coloured splotches on retinas to sandwich making is a very
frightening prospect. Dreyfus does have a point that this is possibly
overambitious. Whether it is impossible in principle or in practice is
very important to AI, at least the work is worthwhile if it is actually
possible. It's best to read Dreyfus as he conveys the anti-SF sense of
hopelessness well. He says that SF should take note of the work that's
already been done by Husserl et al as it is directly relevant to the
aims of SF....only fools would ignore it. These are very interesting
point...what do people think? Can we change the name of this thread or
have a separate one as it's name is getting misleading for new readers?
Husserl found it impossible and he attacked the problem for a great many
years exclusively....at the very least this should be evidential if not
a strong hint to SF.


