What if Evolution is Compatible with Design After All?

Hi Rope and welcome.

I hope you don’t mind a couple of quick comments while I digest the rest:

I don’t think anyone has ever rejected the idea that design and evolution are compatible, not least because they can be shown to be compatible by looking (as you do) at implementation of genetic algorithms.

Thanks for the link.

Perhaps it’s my background in implementing evolutionary algorithms, but that paper seemed rather shallow.

Some specific issues I found with it are:

  • It doesn’t mention one of the most vexing issues when implementing evolutionary algorithms - their tendency to take advantage of loopholes in either the fitness function or the environment and reach ‘solutions’ that are nothing like the implementer’s goal[1];
  • It refers frequently to Dawkins’ ‘weasel’ program, but not to any more recent or more sophisticated evolutionary algorithms, which generally don’t share 'weasel’s simplifications;
  • It seems to assume that selection takes place by looking at the fitness values of the entire simulated population, but there are many evolutionary algorithms that implement selection by directly comparing two (or sometimes three) individual ‘organisms’, without ever calculating a fitness value.

  1. Such as the one intended to produce a signal generator, which produced an aerial and amplifier instead. ↩︎

4 Likes