🚀 go-pugleaf

In article <4029b61d.ffb5@bumper.engin.umich.edu> offutt@caen.engin.umich.edu
(daniel m offutt) writes:
>
>A genetic algorithm simulates a population of linear chromosomes,
>crossover, and fitness-based differential reproduction over a period
>of hundreds or thousands of generations of simulated evolution.
>The algorithm was originally intended as a model of evolution;
>it is quite interesting that it just happens to also be a very
>efficient function optimization method.
>
>So to come to the point:  If creationists are right and evolution is
>nonsense, then how can it be that when one implements a computer
>simulation of evolution (of the right type) the simulation turns
>out to be an algorithm that has tremendous practical value?
>
The characterization of genetic algorithms may have reversed the cart and the
horse.  Genetic algorithms may be said to have been inspired by chromosomal
behavior, but I think it would be an exaggeration to call them a simulation,
of any organic situation.  Nevertheless, genetic algorithms do appear to
exhibit some rather impressive performance in the implementation of
optimization techniques.  The trouble is that, if you want them to
"solve a problem" for you, you have to formulate that problem in terms
of a function to be optimized.  This is all very well and good if you
are, for example, analyzing flows through networks;  but I, for one,
am not yet ready to believe that any life form constitutes an "optimal
fit" of some yet-to-be-discovered "life function."  Thus, impressive as
they may be, genetic algorithms are probably not an appropriate standard
bearer for the virtues of evolution.