Sarah Silverman is suing OpenAI and Meta for copyright infringement

Post by dragonwriter » Mon, Jul 10, 2023, 1:50 AM UTC

CydeWeys wrote: post #318

verelo wrote: post #79

Earlier quoted context omitted.

I feel like thats one of the many questions regulators and law makers are going to be asked long term. I'm sure buying the book for "commercial purposes" like that would't be appropriate, but then again, does that mean if I read it and then summarize it in my work, or regurgitate its info as part of my job...I'm violating a license? A world where humans have special permissions but LLMs don't seems pretty interesting…

There's nothing illegal about reading a book and then circulating your summary/review of it. This doesn't even get into issues of fair use because you aren't redistributing any of the copyrighted material in the first place, merely facts and your own opinions about it. The separate issue that's concerning is that GPT can't be trusted to accurately summarize anything obscure at all, but it'll sure throw text at you no…

Post by listenallyall » Mon, Jul 10, 2023, 1:52 AM UTC

holmesworcester wrote: post #262

pessimizer wrote: post #86

> The complaint lays out in steps why the plaintiffs believe the datasets have illicit origins — in a Meta paper detailing LLaMA, the company points to sources for its training datasets, one of which is called ThePile, which was assembled by a company called EleutherAI. ThePile, the complaint points out, was described in an EleutherAI paper as being put together from “a copy of the contents of the Bibliotik private t…

Let's take a second to remember that copyright is the reason ~every child doesn't have access to ~every book ever written. While it might be too disruptive to eliminate copyright overnight, we should remember that our world will be much better and improve much faster to the extent we can reduce copyright's impact. And we should cheer it on when it happens. A majority of the world's population in 2023 has a smartphone…

Post by RcouF1uZ4gsC » Mon, Jul 10, 2023, 1:52 AM UTC

Post by cgriswald » Mon, Jul 10, 2023, 1:56 AM UTC

TheDong wrote: post #294

int_19h wrote: post #285

> The lawsuit against OpenAI alleges that summaries of the plaintiffs’ work generated by ChatGPT indicate the bot was trained on their copyrighted content. “The summaries get some details wrong” but still show that ChatGPT “retains knowledge of particular works in the training dataset," the lawsuit says. Setting aside the whole issue of whether LLM constitutes a derived work of whatever it's trained on, this sounds l…

That isn't firm evidence, but courts don't need firm evidence to start a case and discover new facts. They very well can ask LLM experts, and openAI themselves, whether that output is highly likely to have been derived from the copyrighted work in question. Anyway. If the argument is "No, it's not from the book, it's from someone else's copyrighted summary", that just means the person who wrote such a summary needs t…

Post by helsinkiandrew » Mon, Jul 10, 2023, 2:00 AM UTC

phoehne wrote: post #305

Natsu wrote: post #282

Earlier quoted context omitted.

This isn't completely new, similar issues came up with search engines and this may be seen as 'transformative'. But there may be issues with models that happily reproduce copyrighted texts in their entirety along with other novel issues like models that hallucinate defamatory things or other such problems. Still, I doubt this particular genie can be stuffed back into the bottle, so we'll probably see a lot of litigat…

I agree it's not an entirely new issue. But it's a little different from search results. Say I use the generative paint brush in photoshop. It reproduces a portion of the copyrighted work. I then use the image on an advertising campaign, other merchandise, or post the final product as my own work. Would I be responsible? Would Adobe? Given that retraining these models is not simple, or cheap, would this be just 'cost…