Xref: utzoo comp.ai:3222 talk.religion.misc:10619
Path: utzoo!attcan!uunet!lll-winken!ames!ncar!boulder!sunybcs!rutgers!mit-eddie!bu-cs!mirror!rayssd!raybed2!linus!mbunix!bwk
From: bwk@mbunix.mitre.org (Barry W. Kort)
Newsgroups: comp.ai,talk.religion.misc
Subject: Re: Elementary AI Philosophy
Summary: A Model Theory.
Keywords: Understanding and Comprehension, Reality and Modeling
Message-ID: <43872@linus.UUCP>
Date: 25 Jan 89 14:13:24 GMT
References: <18464@santra.UUCP> <1241@arctic.nprdc.arpa> <904@ubu.warwick.UUCP> <9423@ihlpb.ATT.COM> <43763@linus.UUCP> <7346@venera.isi.edu>
Sender: news@linus.UUCP
Reply-To: bwk@mbunix.mitre.org (Barry Kort)
Organization: Garden Golems, Inc., Norbert, WI
Lines: 59

In article <7346@venera.isi.edu> smoliar@vaxa.isi.edu.UUCP
(Stephen Smoliar) rejoins the discussion on computer models.
Stephen writes:

>	1.  First of all, there is the problem that you are going
>	to have to manage lots of models.  Thus, there are questions
>	of storage management.  There are also questions of retrieval:
>	How do you know what model to access and when to access it?

I delegate storage management to the operating system.  After all,
I use the corporate Library, which has a very nice system of storage
management (which I did not invent).  As to selection of the 
appropriate model, this is done by pattern-matching on the *structure*
of the knowledge base (without regard to the semantics).  Thus two
English sentences which have the same diagram are analogs, even
if the semantics are completely unrelated.

>	2.  Next, there is the issue of manipulating those models.
>	Metaphorically speaking, any model is going to have lots of
>	"knobs" on it.  Observing a model's behavior is a matter of
>	knowing how to twist which knobs when and what to look for.

Just as Carnegie-Mellon's Terragator goes out to play (so that
it can calibrate it's vision), we can let the computer play with
the model to discover interesting cause-and-effect linkages between
stimulus and response.  The play can be systematic (all possible
combinations in lexicographic order), or chaotic (totally random)
or something in between (heuristic rules with priorities and randomized
tie-breaking).

>	3.  This then brings up the issue of "thought experiments."
>	Like laboratory experiments, thought experiments must be planned.
>	Furthermore, those plans are designed in response to hypotheses
>	to be investigated.  Thus, before we can talk about thought
>	experiments, we have to talk about some agent which "thinks
>	them up."  Thus, while it may be true that you do not need
>	a homunculus to build your models (and I probably would be
>	willing to contest THAT point, too, given more time to think
>	about it), it would seem that your scenario ultimately depends
>	on having a homunculus to manipulate them.

Experiments are planned (or thought up) to resolve questions which
arise in theory construction or theory testing.  The starting
hypothesis is that a model or theory can be constructed at all.
The goal is to construct a compact, easy to use model with
explanatory power.  Unless someone is motivated to ask "Why does
the world work the way it does?", there is no motivation to
seek the answer.

>Is there any way to escape the homunculus?

A knowledge system reposes a symbolic replica of some chunk of the world.
The "understanding" occurs when the knowledge system captures (comprehends)
an accurate replica.  A model which imitates the structure and behavior
of real-world systems is a tool of survival.  The mariner's most valuable
instrument is his charts.  He who sojourns without a map is destined to
blindly walk into brick walls.

--Barry Kort