Path: utzoo!news-server.csri.toronto.edu!rutgers!dimacs.rutgers.edu!mips!sdd.hp.com!spool.mu.edu!munnari.oz.au!ariel!ucsvc.ucs.unimelb.edu.au!u1365281
From: u1365281@ucsvc.ucs.unimelb.edu.au
Newsgroups: comp.ai.philosophy
Subject: Re: computer life?
Message-ID: <1991Mar14.170618.1731@ucsvc.ucs.unimelb.edu.au>
Date: 14 Mar 91 07:06:17 GMT
References: <2179@cluster.cs.su.oz.au> <1991Mar11.165932.19507@news.larc.nasa.gov> <4149@aipna.ed.ac.uk> <1991Mar13.225321.14042@Think.COM>
Organization: The University of Melbourne
Lines: 39

In article <1991Mar13.225321.14042@Think.COM>, kurt@think.com (Kurt Thearling) writes:
 
> 
> You might be interested in "Self-Reproducing Automata and the Origin of
> Life," Robert C. Newman, Perspectives on Science and Christian Faith,  
> vol. 40, no. 1, pp. 24-31.  In this article the self-reproducing automaton
> designed by Chris Langton is analyzed to show that life is designed.
> 
> From the abstract: "... Langton's very simple self-reproducing automaton is
> described in detail.  The complexity of Langton's automaton strongly
> suggests that life is designed rather than accidental."
> 
> Newman computes the time to required consider all of the possible designs
> for such a self-reproducing automaton.  This time is (in his estimate) 
> 3 * 10**139 years.  He then assumes that the universe is 20 billion years
> old to determine that the probablility of producing the self-reproducing
> automaton via random formation is 10**-129.
> 
> 
> kurt
> 

Of course Lord Kelvin thought he showed Darwin was wrong by showing that
the Sun would have cooled long ago if Darwin were right. Kelvin did not
know about nuclear reactions.

In the life case, it is wrong to take all possible combinations, since
early combinations severly restrict future combinations. We call this
historical constraint (see Brooks and Wiley, _Evolution as Entropy_
Chicago: 1988).

Relevance to AI? Historical constraints play an important role in
restricting future behaviour. Behaviour evolved oer a long time may
have many constraints that some one starting from a "top-down"
approach would have trouble thinking of. This raises Dreyfus's old
objection to the strong AI program.

John Collier
Collier@HPS.unimelb.edu.au