And? What's the model supposed to do? It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.

If infringement is happening, it arguably doesn't happen when an infringing work product is generated (or regurgitated, or whatever you want to call it.) Much less when the model is trained. It's when the output is used commercially -- by a human -- that the liability should rightfully attach.

And it should attach to the human, not the tool.

> It's just doing what many human artists would do

I really don't think so. If I paid a human artist to make the prompt in the title, and I didn't explicitly say "Indiana Jones" I would think it should be fairly obvious to the human artist that I do _not_ want Indiana Jones. If they gave me back a picture of, clearly, Indiana Jones, I would ask them why they didn't create something original.

> It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.

It isn’t an independent human. It is a service paid for by customers. The moment it provides the image to a paying user, the image has thus been used commercially.

In fact, the user may not even necessarily have to be paying in order to infringe copyright.

And besides, even amateur artists are ashamed to produce copies unless they are demonstrating mastery of technique or expressing adoration. And if it happens spontaneously, they are then frustrated and try to defend themselves by claiming to never have even experienced the original material. (As happens with simplistic, but popular musical riffs.) But AI explicitly is trained on every material it can get its hands on and so cannot make such a defense.

I agree. But massive changes in scale or leverage can undermine this type of principled stand.

One death is a murder; 100k deaths is a war or a pandemic. One piece of chewing gum on the ground will get you a caning in Singapore; when everyone does it, that's NYC.

Up until now, one had to have some level of graphical or artistic skills to do this, but not anymore. Again, I agree that it attaches to the human...but we now have many more humans to attach it to.

Assuming you can identify it's someone else's IP. Clearly these are hugely contrived examples, but what about text or code that you might not be as familiar with?

Don't worry, the lawsuit will name a corporation that made it, not the AI tool.

> What's the model supposed to do? It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.

Not really? Why would a human artist create a faithful reproduction of Indiana Jones when asked to paint an archeologist? And besides, if they did, it would be considered clear IP infringement if the result were used commercially.

> If infringement is happening, it arguably doesn't happen when an infringing work product is generated (or regurgitated, or whatever you want to call it.) Much less when the model is trained. It's when the output is used commercially -- by a human -- that the liability should rightfully attach.

I agree. Release groups, torrent sites and seedbox operators should not be wrongly accused of pirating movies. Piracy only occurs in the act of actually watching a movie without paying, and should not be prosecuted without definitive proof of such (¬‿¬)

Right! AI developers and directors should be culpable for infringement as part of their duties to larger organizations.

> It's when the output is used commercially -- by a human -- that the liability should rightfully attach.

I am paying OpenAI. So they are producing these copyrighted works and giving them to me for their own commercial benefit. Normally that's illegal. But somehow not when you're just doing it en masse.

>> "a photo image of an intergalactic hunter who comes to earth in search of big game."

I can literally imagine hundreds of things that are true to this description but entirely distinct from "Predator."

> used commercially

Isn't that what these AI companies are doing? Charging you for access to this?