Like Deep Learning (DL), [Evolutionary Computing] was introduced decades ago, and it is currently experiencing a similar boost from the available big compute and big data. However, it addresses a distinctly different need: Whereas DL focuses on modeling what we already know, EC focuses on creating new knowledge. In that sense, it is the next step up from DL: Whereas DL makes it possible to recognize new instances of objects and speech within familiar categories, EC makes it possible to discover entirely new objects and behaviors—those that maximize a given objective.
I used evolutionary computing (specifically genetic algorithms) in both my bachelors project and my PhD. So I can confirm that it works incredibly well, and can throw up some surprising solutions to the problems you throw at them. Even so, I think there’s definitely some hyperbole at play here.
That said: I think this work is incredibly cool. Evolving new recurrent neural net topologies is pretty fascinating. Genetic algorithms can throw out solutions which seem genuinely creative. I can’t wait to see the art evolved neural networks might be able to produce.
On the other hand: we already have some issues with knowing why neural nets makes the decisions they do in cases where the topology is designed by humans. Having them be designed by machines is not going to make that better.