Steve Jones on Computational Evolution

This forum is for the discussion of the evidence for evolution. Anyone is welcome to post, however, scripture is not allowed. As the title says, Science Only please!

Moderator: Moderators

Steve Jones on Computational Evolution

Postby Brian Jordan » Wed Jul 18, 2007 4:20 pm

Darwin's ideas are being used by scientists to develop new drugs and plan phone networks, says Steve Jones
Telegraph, 17/7/07 http://tinyurl.com/2uqwu2
Steve Jones wrote:Computers have long been used to model biological evolution (and Dawkins himself has played a part in this) but Darwin would have been amazed at the ways his ideas are being used by computer scientists to solve non-biological problems.
advertisement

Evolution works, in the factory as much as the field; and military tactics, automatic music transcription and marine architecture - just some of the topics discussed - prove that Darwin's notion of unintelligent design can sometimes beat the most expert engineer.

The theoreticians use evolutionary robotics, genetic algorithms and their relatives to mimic the notion of descent with modification.

The equivalents of mutation, sex and natural selection crack challenges too complex for the fine scalpel of pure mathematics.
"PPSIMMONS is an amorphous mass of stupid" - Rationalwiki
User avatar
Brian Jordan
Forum Admin
 
Posts: 4215
Joined: Sat Oct 07, 2006 4:59 pm

Postby jon_12091 » Wed Jul 18, 2007 5:12 pm

Cool, they've permeated quite away into society these days. Some guy on a forum for a popular on-line (time-wasting) flash game used a genetic algorithm to design mazes for playing the game. I also think they've been used in software for games, CGI and such.
User avatar
jon_12091
 
Posts: 1476
Joined: Sat Oct 07, 2006 4:59 pm

Postby jon_12091 » Fri Aug 10, 2007 8:50 am

There's a good article on the subject in a recent New Scientist (26 July edition)
User avatar
jon_12091
 
Posts: 1476
Joined: Sat Oct 07, 2006 4:59 pm

Postby TheFallibleFiend » Fri Jan 11, 2008 7:28 pm

"The theoreticians use evolutionary robotics, genetic algorithms and their relatives to mimic the notion of descent with modification. "

This is very true; however, there are MANY applied scientists and engineers who are using evolution to solve real-world problems. There are entire companies that make good money solving problems using these techniques:

John Koza owns the company Genetic Programming Inc. that develops programs that evolve.
http://en.wikipedia.org/wiki/John_Koza
Here's an article in Popular Science: http://tinyurl.com/lxc3u

Eric Bonabeau and Icosystem use these techniques http://www.icosystem.com/about.htm
Eric has written articles for Harvard Business Review and Scientific American.

Dave Davis and George Danner at VGOAssociates use these techniques to solve VERY difficult real world problems.

There's not a lot of COTS stuff available yet, but the idea promulgated by some few creationists that this stuff is "out there" is preposterous. OTOH, IDers and other creationists sometimes attack these facts by noting that the programs that implement the genetic algorithms were written by human beings. But most probably see the futility of this kinda of argument, because the essential element is not the code, but the "random" variation of the agents acting under selection. A different criticism is that the program has a selection criterion, while nature does not. This is not correct - it isn't that Nature doesn't have a selection criterion, it's that it's not a purposeful or intentional selection criterion.

One thing I think we could REALLY learn from this is how we might present evolutionary ideas to children, many of whom have preconceptions about the limitations of selection over a randomly varying population. "Eyes on the prize," as I've heard said. Winning the battles is good, but to really win, we want the children coming along to understand.

As a beginning, we could start by show kids a video like this: http://www.youtube.com/watch?v=M2SVMKZhV2g
It would be easy to make a model program that kids could play with themselves. Such programs already exist, but we might want to make them more engaging to children. Finally, in some cases, we could have kids at HS level implement these sorts of algorithms. Creationists can rant to their hearts' content. After kids see the algorithms work - and particularly for those few who might code them and clearly see the simplicity of the principles and the inevitability of the effect - creationists will have a much more difficult time peddling snake-oil
---
"The greatest obstacle to discovering the shape of the earth,
the continents and the ocean was not ignorance but the illusion
of knowledge." -- Daniel J. Boorstin
TheFallibleFiend
 
Posts: 6
Joined: Thu Jan 10, 2008 7:48 pm
Location: US

Postby George Jelliss » Mon Jan 21, 2008 12:18 pm

I find that video unconvincing. It uses an evolutionary algorithm to converge on the image of a south park cartoon character which has been specified in advance as the goal to be reached. This is pure teleology.

More convincing demonstrations on the same lines should be possible. For instance if the requirement is added that any two adjacent rows or columns must have a high proportion of their cells either all matching or all clashing this might lead to the evolution of the types of patterns called "anallagmatic pavements".

It would be interesting to experiment with other possible restrictions, representing environmental constraints. Probably someone has already done this.
GPJ
User avatar
George Jelliss
 
Posts: 204
Joined: Sat Oct 07, 2006 4:59 pm
Location: St Leonards on Sea (UK)

Steve Jones on Computational Evolution

Postby TheFallibleFiend » Tue Jan 22, 2008 2:59 pm

The technique you mention sounds more like cellular automata (CA) than genetic algorithms (GA) or evolutionary programs (EP), but I reckon there might be some benefit to merging the two ideas. I agree that there could be better examples, although most applications of GA have specific goals in mind in advance: parameters that optimize some function. GA techniques tend to have explicit goals, while actual evolution has an implicit 'goal', i.e. survival. I use 'goal' in quotes, because it's not a goal, per se. It's more of a fitness or objective function. A more important distinction, I think, is that the environment (to include members of the same population, as well as other populations of organisms) is constantly changing in real evolution. This means that there is no universally optimum fitness.

Here's some examples of GAs that had more realistic explicit objective functions.
http://www.archive.org/details/sims_evo ... tures_1994
This is interesting, because the strategies evolve in response to those developed by competitors.

It would be interesting to see whether one might have a GA that could figure out its own objective function - a sort of GA version of abiogenesis. Perhaps we would call it a pre-GA. I'm not aware of any work done on that, but I haven't looked for it either.
---
"The greatest obstacle to discovering the shape of the earth,
the continents and the ocean was not ignorance but the illusion
of knowledge." -- Daniel J. Boorstin
TheFallibleFiend
 
Posts: 6
Joined: Thu Jan 10, 2008 7:48 pm
Location: US


Return to Science Only

Who is online

Users browsing this forum: No registered users and 1 guest

cron