From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!utgpu!cs.utexas.edu!sun-barr!olivea!uunet!mcsun!uknet!edcastle!edcogsci!sharder Sun May 31 19:04:53 EDT 1992
Article 5986 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!utgpu!cs.utexas.edu!sun-barr!olivea!uunet!mcsun!uknet!edcastle!edcogsci!sharder
>From: sharder@cogsci.ed.ac.uk (Soren Harder)
Newsgroups: comp.ai.philosophy
Subject: Re: Grounding: Virtual vs. Real
Message-ID: <9597@scott.ed.ac.uk>
Date: 30 May 92 12:12:21 GMT
References: <1992May27.183408.4868@spss.com> <1992May27.193153.19128@mp.cs.niu.edu> <9571@scott.ed.ac.uk> <1992May29.152559.226@mp.cs.niu.edu>
Organization: Centre for Cognitive Science, Edinburgh, UK
Lines: 196

rickert@mp.cs.niu.edu (Neil Rickert) writes:

>In article <9571@scott.ed.ac.uk> sharder@cogsci.ed.ac.uk (Soren Harder) writes:
>>rickert@mp.cs.niu.edu (Neil Rickert) writes:
>>
>>>  Picture a TTT system receiving virtual reality input.  As part of that
>>                                                             ^^^^
>>>virtual reality it is listening to music from a compact disk, and
>>>is connected to a computer via a modem.
>>...
>>>  What I am proposing is that we replace the two modems by a null modem
>>>cable, and replace the CD player/stereo system/loadspeaker/transducer
>>>with a digital CD-ROM reader.  Apparently we are to suppose that the
>>>thinking and intelligence suddenly disappears.
>>
>>First of all as you just said yourself: this is _part_ of the system.

>  Did you really want me to expand my article to several hundred thousand
>pages, by listing all kinds of devices to deal with different types of
>input?

If you had read my mail, including the conclusions/context that you
deleted when responding. (Sorry for these allegations, but it seems to
be the tone) this particular part you've chosen can (_perhaps_) work
with (almost, [see below]) direct digital input. (A better example
would be a text). But there still has to be _symbol_ grounding; the AI
needs some previous experience with the world.

>  Let me restore the context.

>  If I understood him correctly, Harnad said that, under the hypothesis that
>a TTT is possible, you could have a computer create a virtual reality as input
>to the TTT, and that TTT would act intelligently.  But once you remove the
>transducers, and replace them with a direct digital connection from the
>virtual reality generator to the core computer of the TTT, the intelligence
>disappears.

The problem is that the transducers doesn't need to translate the data
back into the form they had in the digital representation of the
virtual world. I can illustrate this with a metaphor: imagine a
symbolic representation, an assembly manual for something in Chinese,
and a 'putting in the world' a Chinese doing what the manual tells him
to do. A transducer (a clever English speaking man) would then be able
to bring it back into a symbolic form, an English text. The product of
this process is different from the input. 

>  I nowhere claimed that intelligence was possible only with listening to
>music.  I nowhere claimed that it was possible to scale up primitive

What you have shown is that a sufficiently intelligent system (how
ever it became intelligent) (we imagine) can enjoy music, *if it is
received in the right format*. But that is trivial. If it is not
received in the right format, like for humans the score of the piece,
or a print out of the information on a CD. (Or a acustic
representation of the information, other than the normal). What Harnad
claims is that part of that intelligence is producing that representation.

>systems, but nor did I claim it was impossible.  I was discussing only
>the claim that there is intelligence with transducers and that the
>intelligence disappears - even though behaviour remains identical - once

Why do you think the behaviour remains identical? (How could it?)

>the transducers are replaced by a direct digital input.

>>Secondly (I believe Harnad has said this himself) the transducer is
>>not just digitizing its input, but is (/is part of) a complex analogue
>>system. The intelligence is (at least partly) in the transducer.

>  Yes, that is my interpretation of Harnad's statement.  And that is
>exactly what I am questioning.

Was that a comment to the last sentence or to the whole paragraph?

>>                                        Take a look at how the brain
>>processes sound; not by constructing spectogram (~ digital
>>representation),

>  Excuse me, but are you under the impression that a CD contains a
>spectrogram of the music?  If so, you are badly misinformed.

I didn't say that. (And didn't mean it either).

>  Incidently, the inner ear contains a bunch of structures which resonate
>at different frequencies, and would seem ideal for developing a
>spectrogram.  Are you proposing that these are mere adornments and have
>nothing to do with how a human hears, and that human hearing does not
>use spectral analysis?

I'm not a physicist, and know very little about turbulence, so the
following description might be wrong. If it is it is still valid as a
metaphor. 

In the ear when the sound has hit the eardrum and has been transformad
into mechanic movement transported through the malleus and the
stirr-up bone, it reaches the cochlea which is filled with a liquid,
and puts that liquid into motion. The motion of this liquid is (I
believe, at present) not 100% computable. Of course it might very
(very, very) well be computable on a large computer in the not so far
future, but this is only one of a lot of transducers, and it would be
a lot easier to make the transduction/computation analogly (and AI is
hard enough without taking the hard way).

To answer your question. Yes, the ear contains these structures, but
it contains other structures as well, which work in the way I have
described. It also seems to contain structures that are especially
useful to speech understanding. (Children are born with an ability to
distinguish speech sound from other sound, and group speech sounds
like adults, independent of acoustic features, i.e. two sounds that
'sound alike' (are tokens of the same phoneme), but are acoustically
different are grouped together, but two sounds that 'sound different'
from eachother, but are acoustically close are not grouped together).

>>                 but by a complex network of feature detectors:
>>rising-tone detectors, falling-tone detectors, harmony-detectors
>>(???).  The human transduction in the ear produces data with features
>>that is vastly different from the salient features of the
>>representation on the CD.

>  Perhaps you can explain what you meant by that comment.  Are you
>perhaps claiming that there is information (rising tone, falling tone, etc)
>which is not present in the data on the CD?  If this is your claim, would
>you please explain how this missing information is magically reinserted
>into the music when the CD player is connected to an audio system, and how
>this information magically disappears when music is digitized?  Or are you
>perhaps claiming that it although all the information is on the CD, it
>is theoretically impossible to extract it with digital means?

I mean that they are vastly different from the salient features in the
representation on the CD. I haven't said anything about missing
information or magic. What I am saying is that there is no
_representation_ of these features on the CD. All information on the
CD is monadic [right word?] and static, gives information about all the
instruments in one jumble at one exact point of time (or rather
averaged over a small time window). When you play it it becomes
dynamic it becomes a rising tone (as part of a jumble), and when you
hear it, you are able to hear the violins as a seperate entity (not
part of a jumble) (this is a rather complex process), and you are able
to hear 'a rising tone' (*not* a C# going to a Eb) (this is a very
primitive process, not deduced but transduced).

>>>  Of course this is preposterous.  Where was the thinking and the
>>>intelligence?  Was it in the modem?  Was it in the stereo?  Was it in
>>>the loadspeaker?
>>
>>Can *you* answer those questions yourself? I'll help you: you need
>>three NO's. If you can give us the fourth answer (and we deem it to be
>>sufficiently specific) you can use adjectives like 'preposterous', not before.

>  Well thank you for answering my rhetorical questions.  I am glad to see
>you agree with me that there is no intelligence in the transducers, and
>that you are thereby (perhaps unintentionally) supporting my claim that
>the transducer argument is bogus.

I might have misunderstood what you meant with a 'modem'. If you mean
a 'transducer' then I retract one no. But I'm looking forward to the
answer of your quiz: Where was the thinking and the intelligence, if
not in the system. 

Some one else has given a reference to Dennetts 'Consciousness
Explained'. I have the idea that you are wandering along close to the
'dualism' trap. You are cutting away parts of the system saying 'This
just engineering. No intelligence' believing that the intelligence is
somewhere in what is left, because you don't understand it. The
intelligence is in the system, and if you cut the system in half the
intelligence disappears.

>>>  This reminds me of the nonsense that went on about classical music
>>>recordings in the early days of CDs.  Many music buffs claimed that
>>
>>Should we keep this off comp.ai.philosophy :-)

>  Once again you excell.  By taking this statement completely out of context
>you have made it appear to be unrelated to the issues.

I'm sorry if you feel I have mistreated your message. I felt that that
part of your posting was an acceptable example, but it seemed a bit to
me as you had included it to slag off even more people than
participate in this debate. Perhaps you should watch your language a
bit more. Words like 'excell', 'nonsense', 'bogus' and 'totally
misinformed' seems to some people (in the context you use it in) as
slightly impolite to use about other peoples beliefs. 


Soren

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Soren Harder, (MSc student)
Centre for Cognitive Science, 2 Buccleuch Place, Edinburgh
E-mail: sharder@cogsci.ed.ac.uk
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>-- 
>=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
>  Neil W. Rickert, Computer Science               <rickert@cs.niu.edu>
>  Northern Illinois Univ.
>  DeKalb, IL 60115                                   +1-815-753-6940


