Newsgroups: comp.ai.philosophy,comp.ai,comp.robotics,comp.cog-eng,sci.cognitive,sci.psychology
From: ohgs@chatham.demon.co.uk (Oliver Sparrow)
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!gatech!swrinde!pipex!demon!news.demon.co.uk!chatham.demon.co.uk!ohgs
Subject: Re: Grounding Representations: ("Grounding" is the wrong word)
Distribution: world
References: <3nhlk5$i7o@percy.cs.bham.ac.uk> <798969311snz@chatham.demon.co.uk> <sjbdLEC00YUvESv1oT@andrew.cmu.edu> <D7pIGq.Knp@gpu.utcc.utoronto.ca> <3nr51r$8ok@news.panix.com> <3o072m$bes@jhunix1.hcf.jhu.edu> <799256142snz@longley.demon.co.uk>
Organization: Royal Institute of International Affairs
Reply-To: ohgs@chatham.demon.co.uk
X-Newsreader: Demon Internet Simple News v1.27
Lines: 45
X-Posting-Host: chatham.demon.co.uk
Date: Mon, 1 May 1995 07:26:08 +0000
Message-ID: <799313168snz@chatham.demon.co.uk>
Sender: usenet@demon.co.uk
Xref: glinda.oz.cs.cmu.edu comp.ai.philosophy:27406 comp.ai:29430 comp.robotics:20293 comp.cog-eng:3118 sci.cognitive:7466 sci.psychology:40950

Re: Aaron's megapiece.

It is, I think, generally the case that when something is thought about,
the thinking agent deploys a set of tools which have been formed through
experience, either directly or by proxy, encoded either in the 'software'
of the thinker or in structures determined by its architectural blueprint
which predispose it to perceive and think in certain ways. "Grounders"
believe that this link is close, immediate and necessary for the continuance
of thought. Those who advance whatever is the antithesis to grounding - 
Spokites? - tend to suggest that logical structures can be built upon logical 
structures, such that new things can be created from pure mentation; and that 
if this can be done once, then it falsifies any general objection to AI. (There 
is, of course, at least two non-sequiturs in that but I shall not pursue them.)

The point which I tried to explore in my previous entry was that this is
not a helpful area in which to look for absolutes. In general, mentation
is the outcome of iteration, either through personal history or the cultural
transmission of norms, attitudes and ideas. UNdoubtedly, once a concept is
fixed, then we are able to build abstract structures from it. The Eiffel Tower 
will, indeed, remain in both my mind and in reality without a continuous 
connection between the two! The key issue is, I think, less *whether* a 
perceiving agent (organism or widget) needs to be connected to the things that 
it is to perceive and more *how much* it needs to be connected, and when. 

In the absence of equivalent of cultural transmission amongst machine-based 
knowledge (and given the absolute role of intermediation of humans until quite 
recently) this does seem to be a basic issue. A knowledge engine seems to 
require (1) the propensity to acquire knowledge from data, something which is
gene-mediated in animals and (2) a proper learning environment, be it burrow
or nursery. Management of the process of knowledge acquisition seems to be
very important to both mechanical networks and living organisms. 

One can, perhaps, say that a system which is *never* grounded - born aware in a 
sea of random grey flecks, perhaps - will be most unlikely to generate the 
framework in which to pose the deep issues of number theory on which we might 
imagine it to specialise. A system born of the knowledge of its creators has a 
structure, however, and is in this sense, grounded. A system which is caught in 
a rich set of feedback between its knowledge formation processes, an observable 
environment, its ability to act upon that environment and within a framework of 
cultural transmission is, however, that much more likely to be a rich and 
complex, evolving system. 

----------------------------------
  Oliver Sparrow
  ohgs@chatham.demon.co.uk
