home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!destroyer!sol.ctr.columbia.edu!usc!news.aero.org!marken
- From: marken@aero.org (Richard Marken)
- Newsgroups: bit.listserv.csg-l
- Subject: Re: language
- Date: 5 Nov 1992 22:11:44 GMT
- Organization: The Aerospace Corporation, El Segundo, CA
- Lines: 62
- Message-ID: <1dc670INNr27@news.aero.org>
- References: <92Nov5.124237pst.29192@hmmm.parc.xerox.com>
- NNTP-Posting-Host: aerospace.aero.org
-
- [From Rick Marken (921105.1400)]
-
- In article <92Nov5.124237pst.29192@hmmm.parc.xerox.com> Penni Sibun <sibun@PARC.XEROX.COM> writes:
- >(penni sibun 921105.1300)
- >
- > [From Rick Marken (921104.1400)]
- >
- > penni sibun (921103.1600) on the relation between her model and pct model
- >
- > >i think the major difficulty in mapping bet. yr model and mine is the
- > >``intended meaning'' part.
- >
- > I agree -- intentions (reference signals) ARE the ONE BIG difference
- > between PCT and ALL other models of living systems.
- >
- >nonsense. intentions are one of the many banes of traditional ai.
-
- Well, maybe you are right. They do talk about intentions a lot.
- My impression is that in most cases the intention is modelled as a
- command to produce action rather than as a specification for perception.
- But I'm sure that my impression is wrong (from your point of view) --
- perhaps you could describe such an ai model.
-
-
- >re: below. i can't make much sense of yr cryptic equations. i don't
- >think you answered my question. i didn't ask whether you thought
- >there were hidden loops in my model. i asked how you could make a pct
- >model that accounted for the same phenomena. obviously, i believe
- >that the phenomena my model captures are important. i'd like to see
- >how you think pct could account for the same process. could you
- >please try again?
-
- Ok. This is a very fair question. Perhaps you could describe the
- phenomenon that your model "captures". Maybe give some quatitative
- data on the phenomenon. From your discussion with Avery it sounds like
- your model (Salex, right?) produces sentences that describe scenes.
- Some of the descriptions are based on inferences derived from the
- scene but not explicitly "stated" -- these are verbally described
- scenes, right. I think you already have a discreet PCT model -- which
- is fine. The references are each of the "descriptions" that should be
- produced. The perceptions are logical variables -- either they are
- true of false. The references for each variable is set at true. Then
- you go through you inference engine and parser and produce states
- of those variables. A higher order system is looking for all the
- reference perceptions to be set to true -- then you are done. So you
- probably already have a control model -- controlling logical variables.
-
- But I think it would be good to see the data on the phenomenon before
- judging the value of the model -- control model or not.
-
- Best
-
- Rick
-
- **************************************************************
-
- Richard S. Marken USMail: 10459 Holman Ave
- The Aerospace Corporation Los Angeles, CA 90024
- E-mail: marken@aero.org
- (310) 336-6214 (day)
- (310) 474-0313 (evening)
-
-