The idea is simple. Art and science has gone together a long time. Imagination is important and comes from advancements in art. Now in machine learning I guess the situation is the same. We could use the strategy to advance art using machine learning to get some more useful science.
So the idea is to mix and come up with new machine intelligences to create music for instance. That is. Better art might require new algorithms that then could be used to advance science.
I think the strategy is faster since there exist much art material that we can compare with which is important in machine learning.