From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!uwm.edu!linac!att!att!fang!tarpit!cs.ucf.edu!news Mon Mar  9 18:34:29 EST 1992
Article 4191 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!rpi!uwm.edu!linac!att!att!fang!tarpit!cs.ucf.edu!news
>From: clarke@acme.ucf.edu (Thomas Clarke)
Subject: Re: Infinite Minds? (was re: Definition of understanding)
Message-ID: <1992Mar2.143901.7258@cs.ucf.edu>
Sender: news@cs.ucf.edu (News system)
Organization: University of Central Florida
References: <1992Mar1.192308.5252@neptune.inf.ethz.ch>
Date: Mon, 2 Mar 1992 14:39:01 GMT

In article <1992Mar1.192308.5252@neptune.inf.ethz.ch> santas@inf.ethz.ch  
(Philip Santas) writes:
| 
| In article <1992Feb29.014127.9300@husc3.harvard.edu>  
zeleny@zariski.harvard.edu (Mikhail Zeleny) writes:
| >Nonsense.  Why is m, the number of all possible sign-types, a finite
| >number?  Furthermore, if meaning is a function of the meaning of
| >constituent sign-tokens, which in turn is context-dependent, there is
| >yet another potentially infinite factor to be accounted for.
| 
| 1) The range of frequences the human ear can receive is limited, and discrete
|    (example: music tones). Proof: limited number of neurns and connections
|    among them; digital functioning.

The problem of AI is a corollary to the question of the fidelity of digital  
simulation.  I used to be a believe strongly that our current understanding of  
physics implies that anything in the world can be simluated on a computer,  
including the brain.  The argument is that the (partial) differential equations  
of physics are subject to theorems that show that if such equations satisfy  
nice conditions (Lipshitz continuity ...) then their solutions can be  
approximated aribitrarily closely by a digital computation.  Therefore,  
anything governed by physics seems to be digitally simulable. In particular,  
the brain can be simulated, and an AI could be implemented (in principle) by  
simulating the equations of the brain.

Since my predilection is to be skeptical of the current AI enterprise, I find  
the apparent implications quantum physics attractive.  According to QM, while  
the brain's wave functions are exactly simulable, the quantum observation  
process is not.  This seems to provide a loop-hole (somehat like Penrose)  
wherein the physical brain can be simulated, but not the mind.  Roughly  
speaking, the mind would be the result of particular observations of the  
brain's wave function.

Lately, chaos provides another, perhaps less mystical, way to produce a  
physical mind that is non-simulable.  Chaos shows that when the equations of  
phsyics are non-linear, their solutions cannot be digitally simulated to  
arbitrarily small accuracy over any significant time interval.  Thus, given  
complete knowledge of the (now classical) brain, simulation may be possible  
over a short time frame (milliseconds, seconds?) but over longer time frames  
nothing can be done. 

Both quantum and chaotic routes to limitation on the simulation of mind, leave  
the mind in the world.  It just happens that physics is such that the mind is  
the product of either necessarily hidden quantum processes, or is the product  
of a system too complex to simulate, or both.  Neither route precludes  
construction of an AI, but in neither case could we be said to know how it  
works, nor could the basis of construction be a digital computer.  The I would  
arise from a physical process that was beyond probing or detailed analysis.

There is a third possibility.  mind is not in the world.  But I'm not enough of  
a philosopher to discuss this intelligently.



