From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!garrot.DMI.USherb.CA!uxa.ecn.bgu.edu!mp.cs.niu.edu!rickert Tue Jun  9 10:06:42 EDT 1992
Article 6061 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!garrot.DMI.USherb.CA!uxa.ecn.bgu.edu!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Newsgroups: comp.ai.philosophy
Subject: Re: Grounding: Virtual vs. Real
Message-ID: <1992Jun3.161959.32462@mp.cs.niu.edu>
Date: 3 Jun 92 16:19:59 GMT
References: <9614@kesson.ed.ac.uk> <1992Jun2.014828.2768@mp.cs.niu.edu> <9640@scott.ed.ac.uk>
Organization: Northern Illinois University
Lines: 134

In article <9640@scott.ed.ac.uk> sharder@cogsci.ed.ac.uk (Soren Harder) writes:
>rickert@mp.cs.niu.edu (Neil Rickert) writes:
>> Given that Harnad has broadened the idea of transducer to include all of
>>the internal components, transducers then become important.

>I don't think Harnad has changed his ideas, you just didn't pick them
>up right away because you think about the field in another way than he
>does. 

  You misread me.  I did not suggest that Harnad changed his ideas.  Only
that he changed the metaphor.  As for what I picked up, I am actually in
agreement with much of what Harnad says.  I'll agree that I think about
things differently from Harnad -- this problem needs to be looked at in
many different ways.  I really only differ with Harnad in two points -
(a) his insistence that analog is special and (b) his dogmatic (my
adjective) rejection of certain (i.e. digital) approaches.

>I believe that there might be a confusion of levels when you say that
>'computers are transducers'. They do not work the way they do
>(input-output-wise) _because_ they are made of transducers; actually
>you might say it is _in spite of_ their being made of transducers.

  Well now you might be getting confused too.  I did emphasize that
components of a computer are transducers in the sense that their input is
really analog.  That aspect is indeed trivial, and I mentioned it partly
as an example why the insistence on analog rather than digital is bogus.
But don't forget that a NAND gate is a transducer with two inputs and one
output, and you can build complete computers out of this type of
transduction.

  Having said this, I should be clear that I don't believe this is what
Harnad meant when he said that the human is a transducer.  I suspect that
he was pointing out that when you talk there is a transduction to sound,
that sound in turn reaches your ears for transduction back to interal
signals.  With certain emotions there is a transduction into hormonal
changes in the blood stream, which again causes other transductions which
feed back into the brain.  Or, in other words, the idea of the brain
as the intelligence organ is a myth -- there are many data flows which
depend on other parts of the body and involve transduction, and these data
paths are crucial to human intelligence.  And if this was Harnad's meaning,
I fully agree.  And indeed this will pose considerable difficulties to the
designers of AI, but I do not believe they are insuperable difficulties.

>> The point remains, however, that the major importance must lie within the
>>information and the way it is processed.  Whether the particular information
>>receptors happen to be analog or digital is completely irrelevant.  Whether
>>the processing (or transformation, or whatever) of the information is
>>purely digital or purely analog, or a combination, is also irrelevant.
>
>I believe you have a very strong argument there. (That is the second bit).
>The first bit I disagree with you. All systems have a limited
>set of forms they need their information in. If the information is not
>in that form there has to be a form of transduction. If you remove the
>transduction, e.g. cut the wires from keyboards to computers and

  Your example of cutting the wires to the keyboard is bogus.  That cuts off
the information entirely.  You need the information.  But the form in which
you receive the information is not cast in stone.  To use the keyboard
analogy you can have a keyboard which reports that the 'A' key is depressed.
Or you can have a keyboard which says that the key in column 3 row 4 is
currently depressed.  Computers have used either.  Once you have all the
information, in whatever form, you can in principle use computational
algorithms to internally transduce it to another form.

>Also for there to be intelligence, there has to be *some kind of*
>representation.

  Sure.  But what else is information if not a form of representation?

>Constructing a mind in the way you suggest is (or might be) possible
>under less favorable circumstances, but I think it would be a wrong
>way of doing it, *if* the processes in the brain are analog, *and* of

  You really have to get off this *analog* kick.  It is bogus.  Almost
any engineer will tell you that if you want to build a general purpose
analog computer today, the best way is to feed the input into a
A to D (analog to digital) converter, process it in a digital computer,
then feed the output to a D to A converter.

  But, more to the point, start looking at human intelligence, and ask
what is analog and what is digital.

  The blueprint is in the DNA.  That is surely digital.  So, in some sense
at least, everything analog about the body is achieved by a digital
synthesis of analog action.

  Now look at a complex feat of muscular coordination.  This is achieved
by individual muscle cells being sent individual signals to contract, and
the signals arrive in the correct sequence to achieve the coordination.
If we were to do this in a robot, we would describe the moving components
as digital to analog converters and the signalling pulses as digital
signals.

  Look at our vision.  The retina of the eye is composed of many receptors.
It looks as if the eye is build to break the image up into many pixels,
much like the way computer based images are managed (except that the
resolution of the eye is far higher).  It certainly has many aspects
of digitization of data.

  Don't think of human intelligence as analog.  The human body is a complex
mixture of analog and digital components.  The digital aspects are just not
so apparent, since they are not coordinated with a central timing clock, and
they therefore don't look much like the digital approach of computers.  But
biology has used digital approaches to solving many problems, particularly
where high precision is required.

>I meant what I said. (I don't usually do, though :-)). It might be a
>good idea to have some basic theories about how the human mind is
>constructed.

  Well, I'll say a little more.

  As I see it, the major requirement, and the most difficult to achieve,
is a good recognition system.  It must be able to recognize rapidly and
with good (but not perfect) accuracy, often based on only partial
information.  It requires an associated learning system, whereby patterns
which are frequently recurrent within the input can be learned, and can
then be recognized when they next appear.  The learning must be passive -
i.e. proportional to exposure - rather than of some goal-oriented form.

  Many people will disagee with the above assessment, and give "thinking"
and "consciousness" a higher priority than recognition.  I don't want to
deal with consciousness for the moment, since there is no general
agreement on what it is, except that it has some relationship to thinking.
As I see it, once you have the recognition in place, "thinking" is a
tractable problem, but without the recognition, thinking is very difficult,
perhaps impossible to fully achieve.

  Earlier in this posting I commented that the brain as the intelligence
organ is a myth, and the whole body is involved.  But I can clarify that.
It is likely that the recognition function is primarily a brain function.
Recognition involves the rest of the body only in the sense that recognition
usually triggers other actions (muscular actions etc).  Thinking, however,
does significantly involve other parts of the body.


