Re: Complex computer programs

From: Richard Loosemore (rpwl@lightlink.com)
Date: Mon Sep 12 2005 - 18:38:27 MDT


Phil Goetz wrote:
> --- Richard Loosemore <rpwl@lightlink.com> wrote:
>
>
>>Did your agents evolve their behavior substantially in the course of
>>interacting with other agents? This would be the crucial factor to
>>look
>>for. Not just a small change of parameters, but the kind of change
>>that
>>makes them almost become different programs? Now, if they did, we
>>might
>>expect some complex behavior. What I was trying to say was that most
>>programmers do not put that level of evolution (adaptiveness) in
>>their code. And because they don't, it is hard to get any of the
>>interesting facets of complexity to show up.
>
>
> The cells in a cellular automata don't change their behavior,
> and don't evolve.
>
> - Phil

You have me hoist by my own petard there. :-)

My first thought is that they are a bit of a peculiar case of a CAS,
almost too simple to be intersting. For example, when you set off a
cellular automaton like Conway's Life, you have give it some initial
data or it sits there, lifeless. Then, we observe regularities in the
course of many, many runs with different data. (Wolfram did this a
great deal, if I recall correctly). But the regularities we observe are
of a fairly simple nature.

I would almost say that CA are a class of CAS by themselves.

And to be sure, there are all sorts of other things that show emergent
behavior without the adaptive component (the sand piles beloved of Per Bak).

So I would revise my position: yes, you can get different kinds of
complexity occurring under a range of conditions, even when there is not
adaptation. But some of those emergent regularities are fairly minor,
and I was overlooking them. For example, Hofstadter tells the story of
the multiuser computer system that became horribly slow after a critical
number of users logged on, but was fine below that number (somebody
suggested curing the problem by swapping all occurrences of the critical
number, inside the OS, with a number ten times as big). That critical
number was an emergent (or at least a non-localized) property of the
system, but from my point of view that kind of emergence is small
potatoes, so I was neglecting it.

Richard



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:52 MDT