cross-posted from: https://nom.mom/post/121481
OpenAI could be fined up to $150,000 for each piece of infringing content.https://arstechnica.com/tech-policy/2023/08/report-potential-nyt-lawsuit-could-force-openai-to-wipe-chatgpt-and-start-over/#comments
No.
.
We 100% need to ensure that automation and AI benefits everyone, not a few select companies. But copyright is totally the wrong mechanism for that.
A pen is not a creative work. A creative work is much different than something that’s mass produced.
Nobody is limiting how people can use their pc. This would be regulations targeted at commercial use and monetization.
Writers can already do that. Commercial licensing is a thing.
… Google’s proposed Web Integrity API seems like a move in that direction to me.
But that’s besides the point, I was trying to establish the principle that people who make things shouldn’t be able to impose limitations on how these things are used later on.
Why should that difference matter, in particular when it comes to the principle I mentioned?
Because creative works are rather obviously fundamentally different from physical objects, in spite of a number of shared qualities.
Like physical objects, they can be distinguished one from another - the text of Moby Dick is notably different from the text of Waiting for Godot, for instance
More to the point, like physical objects, they’re products of applied labor - the text of Moby Dick exists only because Herman Melville labored to bring it into existence.
However, they’re notably different from physical objects insofar as they’re quite simply NOT physical objects. The text of Moby Dick - the thing that Melville labored to create - really exists only conceptually. It’s of course presented in a physical form - generally as a printed book - but that physical form is not really the thing under consideration, and more importantly, the thing to which copyright law applies (or in the case of Moby Dick, used to apply). The thing under consideration is more fundamental than that - the original composition.
And, bluntly, that distinction matters and has to be stipulated because selectively ignoring it in order to equivocate on the concept of rightful property is central to the NoIP position, as illustrated by your inaccurate comparison to a pen.
Nobody is trying to control the use of pens (or computers, as they were being compared to). The dispute is over the use of original compositions - compositions that are at least arguably, and certainly under the law, somebody else’s property.
It’s not like AI is using works to create something new. Chatgpt is similar to if someone were to buy 10 copies of different books, put them into 1 book as a collection of stories, then mass produce and sell the “new” book. It’s the same thing but much more convoluted.
it’s not even close to that black and white… i’d say it’s a much more grey area:
possibly that you buy a bunch of books by the same author and emulate their style… that’s perfectly acceptable until you start using their characters
if you wrote a research paper about the linguistic and statistical information that makes an authors style, that also wouldn’t be a problem
so there’s something beyond just the authors “style” that they think is being infringed. we need to sort out exactly where the line is. what’s the extension to these 2 ideas that makes training an LLM a problem?
Except it’s not a collection of stories, it’s an amalgamation - and at a very granular level at that. For instance, take the beginning of a sentence from the middle of first book, then switch to a sentence in the 3-rd, then finish with another part of the original sentence. Change some words here and there, add one for good measure (based on some sentence in the 7-th book). Then fix the grammar. All the while, keeping track that there’s some continuity between the sentences you’re stringing together.
That counts as “new” for me. And a lot of stuff humans do isn’t more original.
The maybe bigger argument against free-reign training is that you’re attributing personal rights to a language model. Also even people aren’t completely free to derive things from memory (legally) which is why clean-room-design is a thing.
That is not even close to correct. LLMs are little more than massively complex webs of statistics. Here’s a basic primer:
https://arstechnica.com/science/2023/07/a-jargon-free-explanation-of-how-ai-large-language-models-work/
Google web integrity is very much different than what I’m proposing. “Nobody” was more in relation to regulating this.
I hold the opposite opinion in that creatives (I’d almost say individuals only, no companies) own all rights to their work and can impose any limitations they’d like on (edit: commercial) use. Current copyright law doesn’t extend quite that far though.
A creative work is not a reproduceable quantifiable product. No two are exactly alike until they’re mass produced.
Your analogy works more with a person rather than a pen, in that why is it ok when a person reads something and uses it as inspiration and not a computer? This comes back around to my argument about transformative works. An AI cannot add anything new, only guess based on historical knowledge. One of the best traits of the human race is our ability to be creative and bring completely new ideas.
Edit: added in a commercial use specifier after it was pointed out that the rules over individuals would be too restrictive.
I think that point’s worth discussing by itself - leaving aside the AI - as you wrote it quite general.
I came up with some examples:
Taking your statement at face value - the answers should be: no (I can’t decorate), yes (it’s a valid restriction), and no (I can’t use it to illustrate my argument). But maybe you didn’t mean it quite that strict? What do you think on each example and why?
I can see your argument it’s just your metaphor wasn’t very strong and I think it just made things a bit confusing
You made two arguments for why they shouldn’t be able to train on the work for free and then said that they can with the third?
Did openai pay for the material? If not, then it’s illegal.
Additionally, copywrite and trademarks and patents are about reproduction, not use.
If you bought a pen that was patented, then made a copy of the pen and sold it as yours, that’s illegal. This is the analogy of what openai is going with books.
Plagiarism and reproduction of text is the part that is illegal. If you take the “ai” part out, what openai is doing is blatantly illegal.
Just now, I tried to get Llama-2 (I’m not using OpenAI’s stuff cause they’re not open) to reproduce the first few paragraphs of Harry Potter and the philosophers’ stone, and it didn’t work at all. It created something vaguely resembling it, but with lots of made-up stuff that doesn’t make much sense. I certainly can’t use it to read the book or pirate it.
Openai:
That doesn’t mean the copyrighted material isn’t in there. It also doesn’t mean that the unrestricted model can’t.
Edit: I didn’t get it to tell me that it does have the verbatim text in its data.
Here we go, I can get chat gpt to give me sentence by sentence:
Most publically available/hosted (self hosted models are an exception to this) have an absolute laundry list of extra parameters and checks that are done on every query to limit the model as much as possible to tailor the outputs.
This wasn’t even hard… I got it spitting out random verbatim bits of Harry Potter. It won’t do the whole thing, and some of it is garbage, but this is pretty clear copyright violations.
Maybe it’s trained not to repeat JK Rowling’s horseshit verbatim. I’d probably put that in my algorithm. “No matter how many times a celebrity is quoted in these articles, do not take them seriously. Especially JK Rowling. But especially especially Kanye West.”
It’s not repeating its training data verbatim because it can’t do that. It doesn’t have the training data stored away inside itself. If it did the big news wouldn’t be AI, it would be the insanely magical compression algorithm that’s been discovered that allows many terabytes of data to be compressed down into just a few gigabytes.
Do you remember quotes in english ascii /s
Tokens are even denser than ascii. simmlar to word “chunking” My guess is it’s like lossy video compression but for text, [Attacked] with [lazers] by [deatheaters] apon [margret];[has flowery language]; word [margret] [comes first] (Theoretical example has 7 “tokens”)
It may have actually impressioned a really good copy of that book as it’s lilely read it lots of times.
You are reading my comment right now. In my comment, I am letting you know that Sidehill Gougers come in both clockwise and counterclockwise breeds.
Oh no! You just learned that fact for free! I didn’t give you permission to learn from my comments, even though I deliberately published it here for you to read. I demand that you either pay me or wipe that ill-gotten knowledge from your mind.
Don’t you dare tell anyone else about Sidehill Gougers. That’s illegal.
Keep on that strawman my goodman.
Good job demonstrating you don’t understand the underlying point.
Computer manufacturers aren’t making AI software. If someone uses an HP copier to make illegal copies of a book and then distributes those pages to other people for free, the person that used the copier is breaking the law, not the company that made the copier.
All of the examples you listed have nothing to do with how OpenAI was created and set up. It was trained on copyrighted work, how is that remotely comparable to purchasing a pen?
Would a more apt comparison be a band posting royalties to all of their influences?
i think that’s a pretty good analogy that i haven’t heard before!
A pen manufacturer isn’t repurposing other peoples’ work to make their pens.
A computer manufacturer has to license the intellectual property that they use to make their computers.
They didn’t pay the writers though, that’s the whole point
True - but I don’t think the goal here is to establish that AI companies must purchase 1 copy of each book they use. Rather, the point seems to be that they should need separate, special permission for AI training.
100% this! there are separate licenses for personal listening, public performance, use in another work (movie and TV)… there will likely be a license added for AI training to which some authors will opt into, some will opt out of… it’ll likely start very expensive, nobody will pay, someone will offer up
old works that aren’t selling well for bargain basement prices, make a killing, then others will see the success and slowly prices will follow and eventually prices will sit at a happy medium where AI companies can tolerate and copyright holders aren’t feeling screwed… well, i mean, they’ll be being screwed but their publishers will be making bank
that’s my totally out of thin air prediction anyway
I believe this is where it’ll inevitably go. However I’m not sure it’ll be just AI, rather hopefully more protections around individual creative work and how that can be used by corporations for internal or external data collection.
This really does depend on privacy laws as well and probably data collection, retention and usage too.
There is no law that requires them to be paid for this.