Wings

[Mark Walker]

Kenelm wrote:


> This procedure is currently in commercial use. Computer programs are now
> being made by applying a form of natural selection to strings of code
> in which random mutations are made to occur. A working computer program
> would normally be considered to meet Paley's criterion (as in his famous
> watch) for having had to have been designed by a mind--but we now know
> that natural selection is such a powerful (and rapid) process for
> producing
> design from random changes that complex and useful programs can literally
> be evolved within computers. These programs, by the way, are more robust
> than those written by humans, and it is sometimes difficult to understand
> how they function in detail, since they are organized very differently
> from code written by humans.
	 
	Sorry, but this is what I do for a living.  I know, this may seem to
be a conflict of interest, but it really isn't, since there is nothing at
all that clever about genetic algorithms.  They are nothing more than an
attempt to find good or even optimal results in an accelerated amount of
time (beyond exhaustive search).  They involve nothing more clever than a
guided approach to re-jumbling available options into new combinations,
where the guidance is provided by some well-conceived cost function.  The
point that they sometimes find surprising results is only a consequence of
the fact that many more combinations can be evaluated.  It would be very
wrong to suggest that this technology in any way provides evidence that all
of life came into being by such a process.

	Mark Walker.