So, you can take a neonatal human brain, connect its visual cortex to a carousel of Cubist images, and get new Cubist paintings through the motor output, is what you're saying?
The point here is that learning style in humans is more involved than just ingesting images. Humans train to know what an artistic style is, for starters. In end effect, the process is considerably more divorced from any "source" images than what training an ANN looks like. The real issue at hand is what is, and what isn't, a derivative work. With humans, there is, when done professionally, less of an evidently demonstrable connection between any applicable, copyrighted source, and the output.