Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watmath!clyde!rutgers!seismo!ut-sally!berleant
From: berleant@ut-sally.UUCP
Newsgroups: comp.cog-eng
Subject: Re: The symbol grounding problem
Message-ID: <8309@ut-sally.UUCP>
Date: Mon, 22-Jun-87 05:30:35 EDT
Article-I.D.: ut-sally.8309
Posted: Mon Jun 22 05:30:35 1987
Date-Received: Tue, 23-Jun-87 03:44:10 EDT
References: <764@mind.UUCP> <768@mind.UUCP> <770@mind.UUCP>
Reply-To: berleant@ut-sally.UUCP (Dan Berleant)
Organization: U. Texas CS Dept., Austin, Texas
Lines: 47

In article <6670@diamond.BBN.COM> aweinste@Diamond.BBN.COM (Anders Weinstein) writes:
>Another well-known obstacle to moving from an objective to an intentional
>description is that the latter contains an essentially normative component,
>in that we must make some distinction between correct and erroneous
>classification. For example, we'd probably like to say that a frog has a
>fly-detector which is sometimes wrong, rather than a "moving-spot-against-a-
>fixed-background" detector which is infallible. Again, this distinction seems
>to depend on fuzzy considerations about the purpose or functional role of the
>concept in question.

An intriguing example! Maybe it's intrinsic to the fact that inference
takes place? The world is fuzzy enough that logical deduction is not
going to work infallibly, so every time there is inference, as in
categorizing=identifying=classifying a percept, errors in the results
are guaranteed some of the time. Thus errors may be expected to occur
anywhere in the path from percptual icon to concept symbol, rather than,
say, only at the point on this path where purposes or functions come
into play.

>Some of the things you say also suggest that you're attempting to resuscitate
>a form of classical empricist sensory atomism, where the "atomic" symbols
>refer to sensory categories acquired "by acquaintance" and the meaning of
>complex symbols is built up from the atoms "by description". This approach
>has an honorable history in philsophy; unfortunately, no one has ever been
>able to make it work.

Regardless of the form of the "atomic" symbols, complex concepts are
built from simpler ones. Conceptual combination allows us to go
from 'tomato' and 'juice' to 'tomato juice'. I assume there is no
argument that this new category may be acquired, sight unseen, 
by symbolic processing. Presumably there must be atomic=primitive
concepts, however, and where do these come from? It must be
by a process different from that usable to acquire the concept
'tomato juice'. What are the alternatives to acquisition "by 
acquaintance" (I'm not familiar with the term "by acquaintance")? Also,
what is meant by the contention that noone has been able to make it
work? There is research in AI on 'learning from examples', which is
suitable for 'primitive' concepts (ones described by properties rather
than other concepts). There is also research e.g. in the computational
linguistics field, on conceptual combination (tomato + juice = tomato
juice). I know of no system that does both, nor of any reason why we
should ask for one.


Dan Berleant
UUCP: {gatech,ucbvax,ihnp4,seismo...& etc.}!ut-sally!berleant
ARPA: ai.berleant@r20.utexas.edu