From newshub.ccs.yorku.ca!torn!cs.utexas.edu!zaphod.mps.ohio-state.edu!saimiri.primate.wisc.edu!usenet.coe.montana.edu!news.u.washington.edu!ogicse!reed!news Mon Nov  9 09:36:57 EST 1992
Article 7532 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!torn!cs.utexas.edu!zaphod.mps.ohio-state.edu!saimiri.primate.wisc.edu!usenet.coe.montana.edu!news.u.washington.edu!ogicse!reed!news
>From: rseymour@reed.edu (Robert Seymour)
Newsgroups: comp.ai.philosophy
Subject: Re: Artificial Life Processing Power
Message-ID: <1992Nov7.232417.7945@reed.edu>
Date: 7 Nov 92 23:24:17 GMT
Article-I.D.: reed.1992Nov7.232417.7945
References: <1992Nov3.062019.21928@wixer.cactus.org>
Sender: news@reed.edu (USENET News System)
Reply-To: rseymour@reed.edu
Organization: Reed College, Portland, OR
Lines: 92

In article <1992Nov3.062019.21928@wixer.cactus.org> sparky@wixer.cactus.org  
(Timothy Sheridan) writes:
> In article <1992Nov2.080921.9421@netcom.com> lamorte@netcom.com (R. Scott  
LaMor
> te) writes:
> >
> >I'm doing some research for a story I'm writing on Artificial Life,
> >and this seems like the best place to get a little info.
> >
> >It is my understanding that the level of complexity of existing AL is
> >several factors simpler than even the simplest amoeba. And although
> >I'm not sure what the relationship is between the complexity of a
> >amoeba and a human is, I'm sure it's astronomical.
> >
> >So here's my main drift: how much processing power would it take to
> >simulate a human being down to the cellular level? And how long until
> >we have a computer powerful enough to process this info in real time?
> >Are we talking 50 years? 500 years? What?
> >
> >Perhaps another approach is this: how much simplification can we get
> >away with, and still "grow" and simulated human from a simulated
> >zygote?
> >
> >What I am assuming here (and this is open to debate) is that a
> >sufficiently detailed simulated (physical) brain, in a simulated body,
> >raised in a simulated environment, will develop intelligence without
> >the need for top-down programming.
> >
> >-Scott
> 
> 
> If you keep your simulated friend simple (max headroom was a good
> aproximation) then 25-40 years should due just fine by some seasoned
> predictors..
> 
> Hans Morevek (sp) "mind Children" is a good scorce of the complexity isues
> and trends in computer developmet.
> 
> If you think about it though, the amount of data that we pou put is very
> small but its complexity is large.  Todays computers put out lots of data but
> at lower complexity.   So the inteligence of the future will be measured in
> its bulk performance as well as its congeaneality.....
> 
> Also 3 dimentional holograms have perhaps the highest storage capabilities of
> any up-n-comming technology...they should be quite havdy.
> 
> Tim

	At least from the perspective of Artificial Life (as opposed to  
Artificial Intelligence), the important factors which will lead us to simulate  
life are harnessing the power of nature (i.e. evolution, genetic codes,  
mutation, crossover, natural/unnatural selection, etc.). Most (if not all)  
projects in the field of Artificial Life seek not to copy the human form  
through a brute duplication, but to develope the important factors of life  
through simpler models, and allowing the natural developement of these models. 
	Artificial Life has focused on "emergent behaivior" more than  
simulating what a human does. Typical projects begin with simulated beings  
created from random code. These creatures are actually small computer programs  
which typically compete for resources, seek to complete tasks, or fulfill a  
"fitness" criterion in some other manner. Usually very large populations (from  
several hundred to tens of thousands) are competing against each other to  
increase fitness, and therefore to survive and distribute its genetic code.
	For instance, in the project I work on, we simulate a population of a  
few hundred "bugs". These bugs have a genetic code which tells them how to act  
given various combinations of environmental data. The bugs all start out with  
completely random reactions. Eventually the bugs reproduce either sexually or  
asexually, with crossover and/or mutation. A gene which is useful to the bug  
will tend to propagate through the population, and increase the overall  
fitness. Our project has been concentrating on measuring diversity, entropy,  
fitness, and other measures throuhgout the evolution of the creatures.
	The fact that these creatures learn to find food well is an example of  
emergent behavior. We never program the bugs, yet they learn to find food with  
increasing efficiency. This is the manner in which ALife hopes to develope  
living systems. A being created by ALife methods is a product of evolutionary  
forces, not the power of a computer (though a very powerful computer is  
necessary to use the evolutionary forces).
	As for the ability to model a human, I have serious doubts as to  
whether or not this is really possible. While chip technology is advancing to  
single electron transistors, and optical technology is progressing, many  
intrinsic limitations will crop up in the future. Since photons don't interact  
with each other, it is going to be very difficult if not impossible to utilize  
this technology to the extent necessary for modeling a human. It may be that we  
will develope a system for modeling a complex system of the magnitude of a  
human (just think of how many nuerons are simultaneously active in your brain,  
let alone the interaction your whole body), but I think that Artificial Life  
will yeild significant results long before this is possible.

--
Robert Seymour				rseymour@reed.edu
Departments of Physics and Philosophy	Reed College
Artificial Life Project			Portland, OR
Reed Solar Energy Project (SolTrain) - Sunrayce 1993


