Nah... this "dude" is just naive...
AI generated art is 100% dependent on human art. It's basically a math trick that allows users to appropriate attributes of human art, at varying levels of abstraction, for the creation of novel digital images. There are definitely some exciting developments ahead in the world of computer-based art-making, but it won't replace good-old-fashioned human artistic endeavour. It will just add a new layer of practitioners to one sector (i.e., digital imagery/media).
What I do find disturbing is that the proceeds of literally millions of hours of human artistic endeavour are going to be shovelled into the open mouths of a handful of venture-backed companies. And of course, human artist will continue to be marginalized and to endure perpetual financial insecurity, all the while providing the content that makes these technologies function. Something should be done to redress the balance, but it seems unlikely it will happen any time soon.
Mind you, the visual art world lacks a strong, centralized industry. Once these tools step into the ring with human music, film, or games, the lawyers will be out in full force and legislation will start to appear. Putting it bluntly, any technology capable of generating infinite new Beyoncé tracks won't escape regulation—the majors will see to that.
Honestly, I think any digital artist who generates an image from a prompt including only one non-public-domain human artist's name is basically plagiarizing. The output they generate would be stylistically impossible without the thousands of hours effort put in by the named artist to develop the style. That's theft and is frankly fucked, imho. (Actually, no self-respecting human artist would do it, and the art "system", such as it is, would call it out right away.) To be clear, I don't think traditional copyright is a good way to deal with this, but I do think that any generation carried out in this way should be subject to some kind of licensing fee, allowing some portion of the proceeds to go to the human artist for their work (since the digital image would be impossible without them).
Basically, these models represent a form of temporally asynchronous (and partially non-consensual) collaboration between artists, entrepreneurs, data scientists, and end users. What's the best way to compensate the collaborators?
