Registered: Aug 2003
Posts: 712

From the standpoint of every possible bit flip in a DNA sequence, sure. (But life can't be that sensitive, or it wouldn't work - see below). From the standpoint of the basic tendency of life forms to reproduce themselves almost exactly, no. Most imaginable random changes won't have any appreciable effect, or will have an immediately harmful one. If others have the sort of effects traditional mathematical methods are good at modeling - linearly additive continuous behaviors e.g. - then again it is hard to get variety.

In a sense, in the older continuous picture the fitness terrain has to already contain that variety, and natural selection then has to transfer it inward. The usual picture of small adaptive steps toward some local extreme, accumulating to find that extreme, is fundamentally based on this sort of calculus-inspired formal image. As the investigation of systems based on constraints shows, this is not an efficient process unless the fitness terrain is particularly smooth. If it looks like a spikey pegboard, it will quickly get stuck.

But if the dependence on small changes is of the algorithmic, simple programs variety, one can expect much more substantial changes. A large pattern difference between two shells can arise from a small change in the rule that determines pigment placement or not. Is this just a point about sensitive dependence? No, because simply unbounded dependence on small changes, not even keeping the same algorithm, would typically produce something broken, unviable. We wouldn't see the stable types we do see.

This does not mean, incidentally, that such continuous model changes don't happen. Clearly they can, when the right "problem" characteristics are present (a smooth fitness landscape with a clear local maximum, reachable by varying a single parameter e.g.). Such adaptation, of an "engineering efficiency" type, can be explained by that kind of process. But it isn't a process particularly good at producing complexity or variety. Simple programs and discrete changes to them, are good at that. That is my reason for calling them a "variety engine". I hope this clarifies what I meant.

Report this post to a moderator | IP: Logged

Registered: Feb 2004
Posts: 3

Introducing Darwin theory into physics can create no end of mischief.
Consider time. Of all the work done in the scientific world to date, no advances in the understanding of time have been made. Evolutionary theory can distract the scientist with useless drivel and argument over non consequential areas that are best left to idle discussions at coffee bars.

Report this post to a moderator | IP: Logged

Registered: Oct 2003
Posts: 167

Despite having been working on CAs for some years previous, the concept which brought my attention to the broad community of work on complex systems under its many and varied labels was "general evolution".

The fundamental equation

evolution = variation + selection

__________________
Tony Smith
Complex Systems Analyst
TransForum developer
Local organiser

Report this post to a moderator | IP: Logged

Registered: Feb 2004
Posts: 12

Tony Smith: The fundamental equation applies across all domains, physical, biological, social and ethereal. Nothing of consequence happens except as a result of a long viable (evolutionary) history.

Very pertinent to the ideas of Lee Smolin, who postulates Darwinian selection of universes - that ones whose fundamental constants generate complexity are most successful in generating fresh universes.

Report this post to a moderator | IP: Logged

Registered: Oct 2003
Posts: 167

... "who postulates Darwinian selection of universes" but who also devotes Chapter 1 of Three Roads to Quantum Gravity* to justifying his assertion that "There is Nothing Outside the Universe", where he is not making mischief with the definition of universe but rather dismissing the knowability of anything beyond an event horizon.

*despite my discomfort with his chapter 1 the book is kept within arm's reach.

__________________
Tony Smith
Complex Systems Analyst
TransForum developer
Local organiser

Report this post to a moderator | IP: Logged

Registered: Oct 2003
Posts: 21

This does not mean, incidentally, that such continuous model changes don't happen. Clearly they can, when the right "problem" characteristics are present (a smooth fitness landscape with a clear local maximum, reachable by varying a single parameter e.g.). Such adaptation, of an "engineering efficiency" type, can be explained by that kind of process. But it isn't a process particularly good at producing complexity or variety. Simple programs and discrete changes to them, are good at that. That is my reason for calling them a "variety engine". I hope this clarifies what I meant.

Report this post to a moderator | IP: Logged