Remix.run Logo
ToucanLoucan a day ago

I can't speak for everyone obviously, but my anti-AI sentiment in this regard is not that IP law is flawless and beyond reproach, far from it. I'm merely saying that as long as we're all required to put up with it, that OpenAI and company should also have to put up with it. It's incredibly disingenuous the way these companies have taken advantage of publicly available material on an industrial scale, used said material to train their models "for research" and as soon as they had something that vaguely did what they wanted, began selling access to them.

If they are indeed the output of "research" that couldn't exist without the requisite publicly available material, then they should be accessible by the public (and arguably, the products of said outputs should also be inherently public domain too).

If they are instead created products to be sold themselves, then what is utilized to create them should be licensed for that purpose.

Additionally, if they can be used to generate IP violating material, then IMHO, makes perfect sense for the rights holders of those IPs to sue their asses like they would anyone else who did that and sold the results.

Again, for emphasis: I'm not endorsing any of the effects of IP law. I am simply saying that we should all, from the poorest user to the richest corporation, be playing by the same rules, and it feels like AI companies entire existence is hinging on their ability to have their IP cake and eat it too: they want to be able to restrict and monetize access to their generative models that they've created, while also having free reign to generate clearly, bluntly plagiarizing material, by way of utilizing vast amounts of in-good-faith freely given material. It's gross, and it sucks.

flats a day ago | parent | next [-]

Very well put. I’m open to a future in which nothing is copyrighted & everything is in the public domain, but the byproduct of that public domain material should _also_ be owned by the public.

Otherwise, we’re making the judgement that the originators of the IP should not be compensated for their labor, while the AI labs should be. Of course, training & running the models take compute resources, but the ultimate aim of these companies is to profit above & beyond those costs, just as artists hope to be compensated above & beyond the training & resources required to make the art in the first place.

loki-ai 18 hours ago | parent [-]

as an artist, I totally agree with this approach. the whole idea of trying to pay artists for their contributions in training data is just impractical.

if the data’s pulled from the public domain, the model built from this human knowledge should be shared with all creators too, meaning everyone should get access to it

Kim_Bruning a day ago | parent | prev [-]

Beware of pushing for rules that you don't personally believe in. You just might succeed a little too well, and have to live with the consequences.