Genetic algorithms, heuristic hill-climbing, simulated annealing, etc, etc, etc. These are all basically equivalent ways of tackling an optimization function that can only be point-wise queried efficiently, and they are all subject to same problems (most notably local optima, curse of dimensionality, etc.).
I've seen these used rather sporadically in academia, where people generally prefer to reformulate a difficult optimization problem as a convex (or similar) relaxation that is provably solvable, and then throwing solvers at it.
However, I get the impression that when you just need to get some reasonable solution quick and dirty in practice (especially with discrete domains), one of these is the method of choice. I'm not aware of some particularly good arguments for GA's specifically though.
The power of EAs (Evolutionary Algorithms) is that they are able to escape local optima through random mutation and recombination.
EAs are suitable for finding solutions to complex problems, as long as you state the problem correctly and have a decent fitness function. Besides this, parameter tuning is vital for creating a decent EA. This goes so far that there has been research in tuning parameters for an EA using an EA (creating an EA Inception ;)).
> I've seen these used rather sporadically in academia
It really depends on the area. In some areas, like graphics and control theory, convex optimization is the norm. In others, like parts of AI, randomized optimization is a lot more common. Traditionally GAs were more strongly planted in the IEEE-flavored parts of AI, sometimes called "computational intelligence"; for example, the IEEE Transactions on Evolutionary Computation has one of the higher impact factors of AI-related journals. I agree there's no great reason to treat GAs as categorically different from other randomized optimization, though; I suspect they get more press because of the biology metaphor.
I've seen these used rather sporadically in academia, where people generally prefer to reformulate a difficult optimization problem as a convex (or similar) relaxation that is provably solvable, and then throwing solvers at it.
However, I get the impression that when you just need to get some reasonable solution quick and dirty in practice (especially with discrete domains), one of these is the method of choice. I'm not aware of some particularly good arguments for GA's specifically though.