OpenAI wants to bend copyright rules. Study suggests it isn't waiting

theregister.com

19 points·pseudolus··7 comments

even if you want to accept the ‘copyright is less important than getting to AGI before a government we deem less moral than us’ argument, i find it hard to believe things like emulating studio ghibli art styles, are a necessary part of that path.

Yeah. The studio ghibli thing is brazen, in-your-face theft. Especially when considering all the threats of legal action we once endured just to watch a DVD. But when it’s theft on a grand scale by wealthy people, well that’s ok then. /s

We truly are in the second age of the robber barons.

I don't consider stylistic imitation, or even outright copying or obvious copyright infringement, to be theft per se, since the original is not taken away.

But training automated systems on the works of human artists in order to create dirt cheap imitations at scale in their (formerly) distinctive style could certainly make it harder for active artists to find paid commissions.

Flooding the world with inferior imitations seems like it might have some negative consequences as well.

If US companies can't train on copyrighted material China will.

> If US companies can't train on copyrighted material China will.

Where are BSA, MPAA and RIAA when you need them ?

> Tech textbook tycoon Tim O'Reilly claims OpenAI mined his publishing house's copyright-protected tomes for training data and fed it all into its top-tier GPT-4o model without permission.

"I am shocked, shocked, to find that copyright infringement is going on in here."

"Your LLM, sir."

"(Oh, thank you very much.)"

Beauty of Capitalism: everything has a price.

Next time please pay me $5 when you sell my data.

Reply on news.ycombinator.com