From newshub.ccs.yorku.ca!torn!utcsri!rpi!think.com!ames!olivea!pagesat!spssig.spss.com!markrose Mon Nov  9 09:37:05 EST 1992
Article 7543 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!torn!utcsri!rpi!think.com!ames!olivea!pagesat!spssig.spss.com!markrose
>From: markrose@spss.com (Mark Rosenfelder)
Newsgroups: comp.ai.philosophy
Subject: Re: grounding and the entity/environment boundary
Message-ID: <1992Nov9.001936.23167@spss.com>
Date: 9 Nov 92 00:19:36 GMT
References: <1992Nov3.234343.16571@spss.com> <qk0TTB2w165w@CODEWKS.nacjack.gen.nz>
Sender: news@spss.com (Net News Admin)
Organization: SPSS Inc.
Lines: 37

In article <qk0TTB2w165w@CODEWKS.nacjack.gen.nz> system@CODEWKS.nacjack.gen.nz 
(Wayne McDougall) writes:
>markrose@spss.com (Mark Rosenfelder) writes:
>> It's this generalized grounding that isn't diminished when you close your
>> eyes, or sleep, or take a year's sabbatical.  And I'd maintain that this
>> grounding takes years to acquire, and lays the basis for whatever quick 
>> adaptations to changing circumstances we are capable of.
>
>I'd hoped that was exactly the distinction I was making. I think I'm 
>suggesting that it is possible to maintain grounding with low-bandwith 
>communications for those areas in which we (or an AI system) has a 
>generalized grounding. 

I'm not sure I agree; but the question is fortunately empirical.  Someone
needs to stay for a decade or two in a sensory deprivation tank, connected
only to the Internet.  Any volunteers?

>However, this is why closing your eyes for 10 
>years is a problem - your generalized grounding wears out 

Why do you say that?  People who go blind don't forget what colors are
in that time, do they?  And we generally remember quite a bit of our
childhood experiences-- what our houses looked like, what our relatives
were like, etc.  

>(ain't transistors wonderful!). 

No more wonderful than neurons.  I very much suspect that brains forget
things only when they want to.

>How's this for a twist on the Turing Test? A Artificially Intelligent 
>system is something that is worthwhile to talk to.

Cute, but no improvement on the usual TT.  For one thing, it's even more
subjective than usual-- as Marvin Minsky might say, "worthwhile" is a
two-place not a one-place predicate: X is worthwhile *to Y*.  For another,
I know humans who'd fail this test...


