I don't think they are dead. Their popularity for NNs is much lower for sure.

In general, GAs can theoretically solve any problem (if you can formulate a fitness function), given long enough time. Because of that, I think they will always have some use cases.