Greg Rutkowski, a digital artist known for his surreal style, opposes AI art but his name and style have been frequently used by AI art generators without his consent. In response, Stable Diffusion removed his work from their dataset in version 2.0. However, the community has now created a tool to emulate Rutkowski’s style against his wishes using a LoRA model. While some argue this is unethical, others justify it since Rutkowski’s art has already been widely used in Stable Diffusion 1.5. The debate highlights the blurry line between innovation and infringement in the emerging field of AI art.
That’s true, but only in the sense that theft and copyright infringement are fundamentally different things.
Generating stuff from ML training datasets that included works without permissive licenses is copyright infringement though, just as much as simply copying and pasting parts of those works in would be. The legal definition of a derivative work doesn’t care about the techological details.
(For me, the most important consequence of this sort of argument is that everything produced by Github Copilot must be GPL.)
That’s incorrect in my opinion. AI learns patterns from its training data. So do humans, by the way. It’s not copy-pasting parts of image or code.
AI doesn’t “learn” anything, it’s not even intelligent. If you show a human artwork of a person they’ll be able to recognize that they’re looking at a human, how their limbs and expression works, what they’re wearing, the materials, how gravity should affect it all, etc. AI doesn’t and can’t know any of that, it just predicts how things should look based on images that have been put in it’s database. It’s a fancy Xerox.
It’s actually not copyright infringement at all.
Edit: and even if it was, copyright infringement is a moral right, it’s a good thing. copyright is theft.
Aside from all the artists whose work was fed into the AI learning models without their permission. That art has been stolen, and is still being stolen. In this case very explicitly, because they outright removed his work, and then put it back when nobody was looking.
Let me give you a hypothetical that’s close to reality. Say an artist gets very popular, but doesn’t want their art used to teach AI. Let’s even say there’s even legislation that prevents all this artist’s work from being used in AI.
Now what if someone else hires a bunch of cheap human artists to produce works in a style similar to the original artist, and then uses those works to feed the AI model? Would that still be stolen art? And if so, why? And if not, what is this extra degree of separation changing? The original artist is still not getting paid and the AI is still producing works based on their style.
So you hire people to trace the original art, that’s still copying it, and nobody is learning anything. It’s copying.
Strictly speaking it wouldn’t exactly be stealing, but I would still consider it as about equal to it, especially with regards to economic benefits. It may not be producing exact copies (which strictly speaking isn’t stealing, but is violating copyright) or actually stealing, but it’s exploiting the style that most people would assume mean that that specific artist made it and thus depriving that artist from benefiting from people wanting art from that artist/in that style.
Now, I’m not conflicted about people who have made millions off their art having people make imitations or copies, those people live more than comfortably enough. But in your example there are still other human artists benefiting, which is not the case for computationally generated works. It’s great for me to be able to have computers create art for a DnD campaign or something, but I still recognize that it’s making it harder for artists to earn a living from their skills. And to a certain degree it makes it so people who never would have had any such art now can. It’s in many ways like piracy with the same ethical framing. And as with piracy it may be that people that use AI to make them art become greater “consumers” of art made by humans as well, paying it forward. But it may also not work exactly that way.
