Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!helios!bcm!dimacs.rutgers.edu!seismo!uunet!news.larc.nasa.gov!grissom.larc.nasa.gov!kludge
From: kludge@grissom.larc.nasa.gov ( Scott Dorsey)
Newsgroups: comp.ai.philosophy
Subject: Re: computer life?
Keywords: Survival, instincts
Message-ID: <1991Feb28.235517.20218@news.larc.nasa.gov>
Date: 28 Feb 91 23:55:17 GMT
References: <5375@media-lab.MEDIA.MIT.EDU> <1991Feb26.213835.27074@watdragon.waterloo.edu> <1991Feb27.134800.18153@news.larc.nasa.gov> <1991Feb28.190553.20519@msuinfo.cl.msu.edu>
Sender: news@news.larc.nasa.gov (USENET Network News)
Reply-To: kludge@grissom.larc.nasa.gov ( Scott Dorsey)
Organization: NASA Langley Research Center
Lines: 79

In article <1991Feb28.190553.20519@msuinfo.cl.msu.edu> dailey@buster.cps.msu.edu (Chris Dailey) writes:
>In article <1991Feb27.134800.18153@news.larc.nasa.gov> kludge@grissom.larc.nasa.gov ( Scott Dorsey) writes:
>>minsky@media-lab.media.mit.edu (Marvin Minsky) writes:
>>>But that shows the great joke in the attempts to "define life" that
>>>have appeared in this newsgroup.  Missing the whole point of how
>>>natural selection produces stuff.  The lesson should be, you can't
>>>define stuff, only words. [...]
>>   Granted, this is a good point.  But manmade systems (like computers)
>>are not evolved, but designed.  Knowing the path by which lifeforms
>>evolved might help us construct an artificial life form, but it's not
>>required.
>
>This is the only form of evolution of which we have some idea of how
>things work.  A manmade system which is designed would probably be
>designed so it will evolve beyond its design.  We will probably learn a
>lot if we do this.  (Some of us believe that WE were designed, and have
>since evolved on our own, but I have digressed.)

   Many learning systems exist which in a sense evolve as they learn about
their environment.  But this is evolution in a very different sense.  A
system that evolves beyond its design must not do so in the same fashion
as humans did.  Software mutation and the weeding out of bad variants is
not a good software engineering practice.  

>>   Computers play chess.  They play chess well.  But they play chess in
>>a fashion utterly unlike human beings, because they operate in a manner
>>very different from the human brain.  
>
>However, these computers are designed (in many or most cases) with the
>human's strategies.  They are algorithmic representations of human
>thoughts.  The only way (IMO) they could operate in a manner truly very
>different would be if they were the ones that taught themselves how to
>play.

No, they don't at all use the same strategies that humans use.  Many humans
can't explain the strategies they use, and if they could, they probably
would not be efficient to implement.  Many good chess programs use exhaustive
search strategies which, if used by humans, would result in many years for
each move.  On the other hand, humans can immediately discount moves using
tree-pruning-type behaviour that isn't possible to implement because computers
don't do pattern matching as well as humans do.  Because humans and computers
are better at different things, they use different strategies.  The strategies
used by computers were thought of by people whose job it is to think like
computers (ie. programmers), and while they were indeed thought of by people,
those people would not have thought of them had they not been specifically
trying to tailor a solution for a system which has limitations utterly unlike
the human brain.

>>Nevertheless, although the mechanisms
>>inside and the playing strategies might be quite different, the end result
>>is the same.
>
>But the strategies are not quite different, except that most computers
>do not have the pattern matching capabilities of humans are are usually
>made to do cost/benefit analysis on most or all possible moves instead
>of just the most promising ones.

   Computers have the disadvantage of not being able to do pattern matching
very well, but they have the advantage of being able to do calculation and
repeated testing of many possibilities very well.  This is a very significant
difference.  The way I learned to extract square roots in school is very much
unlike the way the SQRT reoutine on my Sun extracts them, because what is easy
for me may be difficult for the grey box and vice versa.

>
>>   If a computer life form is constructed, it will probably not be constructed
>>in any manner resembling the evolutionary method by which all living systems
>>we know have been formed.  This is because computers, again, operate in a 
>>very different fashion than organic systems.  But nevertheless, the result
>>will be the same.
>
>I believe a computer's evolutionary method, although modeled after the
>evolution of living (I assume you mean, organic) systems, would be
>significantly different (although not necessarily different enough that
>we could not learn more about our own evolution).  Kinda like weather
>patterns in comparison to living systems.

This sounds very much like we are agreeing.
--scott