The art of sampling from Kanye to machine learning models

As I dive deeper into machine learning, one question always comes up. Who is the artist? You or the machine? #AIArtist There's a ton of similarity in music sampling, but we'll focus on visual sampling.

Let's look at Rene Magritte work "The Lovers", what makes it Magritte? is it the bushes? the faceless human? the mask? We can use a ML model called visual-importance created by Adobe/MIT to find out ( http://web.mit.edu/zoya/www/docs/predImportance_final.pdf )

It generate a gradient map of what's important, what make this piece recognisable. So in theory, if we remove that part of the painting, it loses it meaning or its rights to the artist, right?... or does it create new ones. I then used content-aware by Adobe to complete the painting.

I like that the Machine literally removed the faces, bringing another depth to the conversation, emptiness. And when It tries to fill in that emptiness, the subject is erased from history.