X-Message-Number: 14964
From: "Pat Clancy" <>
Date: Sun, 19 Nov 2000 19:35:42 -0800
Subject: Re: Simulating People and Animals

Henri Kluytmans wrote:

> On the contrary, simple primitive animal activities (functions) can 
> already be simulated.
> 
> I can remember simulated organisms with simulated 
> neural nets that where genetically evolved using certain 
> survival rules. These virtual organisms developped movements 
> and also behaviors (competition strategies) you also see in 
> real biological organisms.
> 

Programmed simulations of animal behaviour are always poor imitations that 

require a carefully set-up artificial environment to do something very limited.
They lack the flexibility and adaptability that even simple organisms display. 
You can certainly create a program that does something you might call 
"competitive behaviour" within its virtual environment, but to me that's a far 
cry from saying you've duplicated some aspect of the biological mind, or that 
you could keep on making such programs more complex until you'd achieved 
something comparable to a mind.

> 
> To avoid any miscommunication : Do you consider 3D cellular automata 
> (CA) based neural nets (containing millions or billions of artificial 
> neurons) which grow and evolve inside cellular automata machines (CAMs) 
> (in based FPGA field programmable gate arrays, hardware) a Turing 
> machine ?
> 
> When I used the term "Turing Machines" I was always reffering 
> to step by step symbolic computing systems with only a low 
> number of processors. I was not refferring to massively 
> parallel systems, like hardware implemented neural nets, 
> although in principle they also should be considered Turing 
> machines.
> 

There seems to be this idea that even if you couldn't implement a mind on a 
single processor machine, you'd still be saved by having a huge number of 
processors, that that would make it doable. This is not so. Now I'm not 
saying you don't believe a mind could be implemented on a single processor 
computer; if you believe that then we simply disagree in our basic 
assumptions and that's that. However, the number of processors and how 
they're connected makes no difference at all to the question "can a mind be 
implemented on a Turing machine?". Any Turing machine can run any 
program that can be run by any other Turing machine. The only qualification 

to this is that it refers to "universal" Turing machines, meaning those with an
infinite tape. The tape corresponds to the storage in an actual computer. So 
to restate it, any Turing machine _with sufficient storage_ can run any 
program that can be run by any other Turing machine. Of course, a single 
processor machine may run a given program much slower than a massively 
parallel machine (depending on the program). But let's say sometime in the 
far future there's a program called "Consciousness 1.0" that runs on 
Windows 8000 and normally utilizes 10^12 processors; that same algorithm 
could in theory be run on the computer you're using now, if you had enough 
storage (memory + disk space), it just would be a lot slower. (Of course I'm 
making some assumptions about the computer you're using now!)

Pat Clancy

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=14964