Technology Review article offers a good introduction to the use of
evolutionary algorithms for hardware design. By replacing natural
selection with a user-defined fitness function genetic algorithms use
random mutation and recombination of the most fit designs to produce
continually better designs. The article includes plenty of real-world
examples of the results including "a corkscrew contraption small
enough to fit in a wine glass, yet able to send a wide-beam radio wave
from space to Earth. It resembles nothing any sane radio engineer would
build on her own." The article also touches on John von Neumann's
"complexity barrier" and the origins of genetic algorithms in the 1950s.
It sounds like a lot of trust is being placed in the software running
these genetic algorithms. Theres only one thing I can think of though
when relying on a computer that heavily - Garbage in, Garbage Out...
(I haven't yet RTFA)
I'd be very skeptical of a genetically-grown digital device with
asynchronously cleared flip-flops, for instance, but a genetically-grown
antenna should be pretty straightforward to verify.
Genetic algorithms are (somewhat) notorious for finding flaws in the
model and then exploiting them, such as the aforementioned noiseless
environment. Noise can be added to the performance metric, but there's
always the chance something else will be exploited.
None the less, IMHO, they can create very good designs and are a
valuable computer aided design tool.