From newshub.ccs.yorku.ca!ists!torn!utcsri!rpi!zaphod.mps.ohio-state.edu!mips!apple!apple!mips.mitek.com!spssig.spss.com!markrose Tue Jun 23 13:21:22 EDT 1992
Article 6327 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!torn!utcsri!rpi!zaphod.mps.ohio-state.edu!mips!apple!apple!mips.mitek.com!spssig.spss.com!markrose
>From: markrose@spss.com (Mark Rosenfelder)
Newsgroups: comp.ai.philosophy
Subject: Re: The Turing Test is not a Trick
Message-ID: <1992Jun19.225601.6595@spss.com>
Date: 19 Jun 92 22:56:01 GMT
References: <491@tdat.teradata.COM> <1992Jun18.164543.42825@spss.com> <502@tdat.teradata.COM>
Organization: SPSS Inc.
Lines: 57
Nntp-Posting-Host: spssrs7.spss.com

In article <502@tdat.teradata.COM> swf@tdat.teradata.com (Stanley Friesen) writes:
>How do we judge the presence of 'complex sensory processing'?  What measure
>do we use to determine 'complexity'?

This could be approached on a sense by sense basis; for instance, on vision
we could look at the resolution, sensitivity to amount of light, to
differences in frequency, to motion, etc.  Or perhaps we could see 
how big a swath of brain is used for sensory processing.

>So, is there any reason other than evolutionary accessibility that requires
>complex sensory processing be *integral* to intelligence?

I prefer to look at it from the other direction: Given that large portions
of the brain are devoted to sensory perception, and the same organ is used
both for perception and for more self-evidently "intelligent" capacities
such as calculation, language use, or creativity, on what grounds can we
*separate* intelligence from perception?

For instance, I strongly suspect that human semantics is inextricably
integrated with sensorimotor processing.  The lexical entry for "dog", so
to speak, contains a pointer to pictures of dogs, experiences with dogs, 
etc., and these would be uninterpretable without the sensorimotor portion
of the brain.

>|I'd be quite surprised if alien communication were significantly less
>|sophisticated and multi-channelled.  
>
>I can agree with this principle, but I feel that specifying intonation
>per se is too anthropocentric.  How about changing it to something
>more general like 'uses multi-mode communication'.

OK.

>|Just to give one example, watching an alien successfully repair its broken 
>|space scooter would give you good prima facie evidence for its intelligence,
>|even if it never uttered a word (besides a photic obscenity or two).
>
>Hmm, maybe, maybe not.  I am not sure that such behavior is not possible
>to a much less 'intelligent' being than a human.  In fact it might even be
>possible to train a chimpanzee to do simple mechanical repairs.

I doubt it-- we're talking animals whose top-level technological achievement
is the grass-stalk termite grabber-- but we won't know till we've tried it.  

>This really looks like a job for net-brain-storming.  Any ideas anyone?
>How should we specify this to make the test actually usable?

Ask the robot, or the alien, to repair the flat tire on your bicycle.
To be precise, the testee must have no prior knowledge of how to do
this task.  It has to learn how to do it from the tester, and (after
a few practice tries) successfully do it on its own.

This could be called (thinking of the Tour de France) the Touring Test.
(Sorry.)

I think it would be possible with some thought to come up with plenty 
of good tests of nonverbal intelligence.


