If you are talking about CAMD, then the so-called "genetic algorithms" are merely a restricted form or monte-carlo search algorithms--that is, they rely on luck. There are millions of AI search algorithms out there that could easily stand in with similar if not better results. One reason why is because genetic algorithms suffer from the same difficulties as neural networks, in that they get stuck in local maxima as the "population" is pruned. Increasing your mutation rate is analogous to larger weight adjustments with neural networks and the result is the same: a much longer search period.
In my graduate AI class, my professor remarked that genetic algorithms simply moved the difficult part of the search to the fitness measure. If you had a good enough fitness function, then many search algorithms could do just as well or better. The algorithm itself was inconsequential by comparison.
Google "stochastic."