Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!scramble.lm.com!news.math.psu.edu!news.cse.psu.edu!uwm.edu!newsfeed.internetmci.com!in2.uu.net!EU.net!usenet1.news.uk.psi.net!uknet!newsfeed.ed.ac.uk!dcs.ed.ac.uk!cnews
From: iic@dcs.ed.ac.uk (Ian Clarke)
Subject: Re: Reproduction
X-Nntp-Posting-Host: ramna.dcs.ed.ac.uk
Message-ID: <Dr8rtK.J13.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk>
Sender: cnews@dcs.ed.ac.uk (UseNet News Admin)
Organization: Department of Computer Science, Edinburgh University
X-Newsreader: knews 0.9.6
References: <318F40AD.41C67EA6@gpsemi.com>
  <4msvu5$blv@pheidippides.axion.bt.co.uk>
Date: Sat, 11 May 1996 12:53:41 GMT
Lines: 84

In article <4msvu5$blv@pheidippides.axion.bt.co.uk>,
	richardm@dilithium (Tin Tin) writes:
>Hi - delurk time for me too!
>
>I've read what Thomas Ray (or is it Ray Thomas?)
>wrote WRT to evolving software organisms.

It was Tom Ray.

>
>Based on his work, I created a program (mainly out of
>interest) to attempt to duplicate his work (not
>in-toto I hasten to add - but just the interesting
>bits).
>
>It seemed to me that the essence of the idea was
>two simple steps:
>
>1. copy self (requiring determination of limits of self)
>2. let new organism get some runtime (birth?)
>
>If the cycle repeats then multiple autonomous
>instances of the original will exist. Add random
>errors during duplication and you get a sort
>of evolution. Most will be unviable and will
>either not run at all or will not be able to
>reproduce and die out (others will loop forever
>unless some aging is built in to allow them
>to expire gracefully).

If the system is designed like Tierra then all programs
should run.
>
>I wrote something which was capable of doing this
>using fifteen different opcodes (two were data 0 and
>data 1 for labels). The whole thing took up only
>some three dozen opcodes.
>
>I think the chances of randomly producing the
>right sequence of 30 odd numbers from 0 to 14
>that constitute my program cannot be that high.
>I'm not a statistician so I can't say. I'd be
>interested if someone could quantify this.

Yes but here the problem lies.  It is a very similar
argument to that used against the theory of evolution.
How could something as complex as the human eye evolve?
It was obviously not gradually, half an eye would do an
organism no good, and it is not easy to progress gradually
from compound eyes such as found in insects.
In the statistics you request it would be useless to
calculate the probability of producing a specific
organism (It is about the same as winning the British 
national lottery 35 times in a row), you would need to take
into account all possible organisms imaginable which could
possibly reproduce.  This will bring the number down to
a suitably realistic value quite rapidly.
>
>BTW I'll post the ancestor entity I designed if anyone
>wants to see it.

It may be interesting to see.
>
>I think a key point which makes Tierra type organisms
>more likely to be created randomly is that the
>instruction set is so small. This is a huge
>limitation 'cos the evolution of the organism is
>tied to the ways in which it can interact with the
>environment in which it lives. The opcodes and the
>executive that runs them constitutes that interaction
>set and therefore the limits are set in such a
>way that no amount of eveolution can increase the
>interactions.
>
If you want to increase the flexability of a program you
do not need to increase the size of the instruction set,
merely use lower-level instructions.

Ian
-- 
|IAN CLARKE        I.Clarke@sms.ed.ac.uk  "..until human voices wake  |
|                  iic@dcs.ed.ac.uk       us, and we drown" - Shelley |
|WWW:http://www.lookup.com/homepages/107558/homepage/index.html       |

