From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!uwm.edu!ogicse!plains!vender Mon May 25 14:07:26 EDT 1992
Article 5871 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!uwm.edu!ogicse!plains!vender
>From: vender@plains.NoDak.edu (Brad Vender)
Newsgroups: comp.ai.philosophy
Subject: Re: Grounding: Virtual vs. Real
Summary: Possible solution
Keywords: transduction, analog
Message-ID: <18039@plains.NoDak.edu>
Date: 24 May 92 02:27:55 GMT
Article-I.D.: plains.18039
References: <1992May20.034459.8223@Princeton.EDU> <6906@pkmab.se>
Organization: North Dakota State University, Fargo, ND
Lines: 44

I have a possible solution to the grounding problem:
  semantics are grounded because they refer to stimili sets.
  The origin of these stimili can be the system itself (human or
  AI) or the outside world processed through the senses.

'hunger' has meaning because it refers to the stimili we associate
  with the condition (weakness, desire for food, etc.)

'dog' has meaning because it refers to all the sensory data
  and information we possess about various forms of dogs.
  The fact that we ignore certain superficial characteristics
  in various 'dog' specimens when forming this link is
  important in formation of the linkage.

Or more importantly, when I consider one of my friends, the
  symbol dereferences to all the events I've experienced with
  my friends, and also the characteristics associated with
  the friends themselves (weight, eye color, attractiveness,
  etc.).

The symbols can be formed by composition of other symbols.
  I suppose imagination is the capability to do this.  After all,
  if we construct the reference for 'unicorn' from 'horse' and
  'possesses a single horn' we get an image which has no direct
  connection to reality, but which still has meaning.

This supposition (that semantics is the mapping of symbols to
  real or false sensory data) allows us to ignore the condition
  of whether the world we perceive is "real" or not.

In other words, stop arguing about whether or not transducers
  can actually connect a real world to the symbols.  Semantics
  allows us to review past sensory data by linking it with
  various symbols.  The fact that in order to form common
  linkages (to make sure that everyone is using the same
  data for 'apple' for instance) we have to have common
  sensory data (both seen some common examples of 'apple')
  is merely a supporting case.

All of this is still new to me, but I hope someone out there
  understands it and will respond.  Is the idea good or not?
  Let me know.

--Brad


