From newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!caen!malgudi.oar.net!sunc.osc.edu!cis.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!rickert Mon Jun 15 16:04:40 EDT 1992
Article 6207 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!caen!malgudi.oar.net!sunc.osc.edu!cis.ohio-state.edu!pacific.mps.ohio-state.edu!linac!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Subject: Re: Physical versus Computaional (was Re: Transducers)
Message-ID: <1992Jun11.153432.4670@mp.cs.niu.edu>
Organization: Northern Illinois University
References: <4138.708217481@mp.cs.niu.edu> <1992Jun11.132823.7139@cs.ucf.edu>
Date: Thu, 11 Jun 1992 15:34:32 GMT
Lines: 41

In article <1992Jun11.132823.7139@cs.ucf.edu> clarke@acme.ucf.edu (Thomas Clarke) writes:
  (in reply to my earlier article <4138.708217481@mp.cs.niu.edu>)  

>You are using "physical" in a much different way than I would.
>The phrase "physical growth" seems to imply what I would 
>call something like "neuromorphological" (?)
>
>My use of physical versus computational in the context of the brain
>would be to distinguish between the analog of neural net implementation 
>of x*y versus a digital implemenation of x*y.

  Excuse me, but neural net implementations usually run on digital
computers.  Sure they are physical - physical electrons move around.  But
that is a trivial meaning of physical.  How can such a neural net
implementation not be computational?  Analog computation is still
computation.  Slide rules were analog devices commonly used for
computation only a few decades ago.

  Compare this to thought.  There are things that you can think about that
will make you so sad tears will show up in your eyes.  Those tears are
physical.  They indicate the occurrence of physical actions in the body
beyond the transfer of neural signals.  The question of how central these
physical actions are to thought remains controversial.

>Idealized neurons with continuous sigmoidal functions can compute 
>x*y by "biasing" the neurons into a region where the activation is
>well-approximated by the square of the input, a=s^2.

  They still compute.  It is still a computation.  The fact that the
multiplication x*y didn't use the multiplier hardware in the CPU in
no way alters the fact that it did a multiplication.

>A computational implementation would involve converting x and y into
>some sort of coded representation (e.g. binary) and using the 
>computational resources (e.g. neurons) to implement the necessary
>combinatorial logic to compute the coded representation of x*y.
>Many neurons, larger delay times, but digital accuracy.

  You are right.  I am using "physical" in a different way from your
usage.  Your usage of the terms just does not make sense to me.



