More Than 10,500 Artists Unite in Fight Against AI Companies' 'Unjust' Use of Creative Works (aitrainingstatement.org) 64
More than 10,500 artists and creators -- including ABBA's Bjorn Ulvaeus, actress Julianne Moore, actors Kevin Bacon and F. Murray Abraham, as well as former Saturday Night Live star Kate McKinnon, author James Patterson and Radiohead's Thom Yorke -- signed a statement condemning AI companies' unauthorized use of creative works for training their models. The initiative, led by former AI executive Ed Newton-Rex, demands an end to unlicensed training data collection amid mounting legal challenges against tech firms. "The unlicensed use of creative works for training generative AI is a major, unjust threat to the livelihoods of the people behind those works, and must not be permitted," reads the statement.
The protest comes as major artists and publishers battle AI developers in court. Authors John Grisham and George R.R. Martin are suing OpenAI, while record labels Universal, Sony and Warner have filed lawsuits against AI music creators Suno and Udio. The signatories reject proposed "opt-out" schemes for content scraping, calling instead for explicit creator consent.
The protest comes as major artists and publishers battle AI developers in court. Authors John Grisham and George R.R. Martin are suing OpenAI, while record labels Universal, Sony and Warner have filed lawsuits against AI music creators Suno and Udio. The signatories reject proposed "opt-out" schemes for content scraping, calling instead for explicit creator consent.
Farts in the Supreme Court need to legislate (Score:2)
Congress isn't going to do shit. The EU is completely paralyzed waiting for the US to tell them which way to jump ... until this gets to the US Supreme Court the status quo will stand.
It could get hilarious, we could see the first trillion+ dollar judgements if copying into a training set requires a license. The statutory damages on registered works alone will get there for the big boys.
Re: (Score:3)
Re:Farts in the Supreme Court need to legislate (Score:4, Interesting)
My best guess is the courts will (eventually) rule that content creators must be paid for their work.
There's hardly any basis for it. The most likely outcome would be that downloading data and training a model with that data does Not infringe upon copyright at all. You are not reproducing and distributing Nor are you publicly performing a work by training a model.
On the other hand: when using the trained model to generate images based on it - the resulting image might be infringing. This Infringement could occur when you prompt an AI model and then download and distribute the result of the generative AI -- IF the result is probatively similar to a copyright work, then the AI created a reproduction, and the user's act of downloading and sharing a copy of that reproduction would be copyright infringement. Also, the Online service provider who provided the model that the end user prompted Could be vicariously liable, because their online Subscription as a service "AI generator tool" will have created the reproduction most likely on its own: Therefore the online AI service provider would be Liable for assisting the end user of their online service in committing any acts of copyright infringement.
Re: (Score:3, Insightful)
Re: (Score:2)
Re: (Score:2)
>While it's true that the information from the original work is in some sense copied into the AI model, that information is also thoroughly mixed with information from the rest of the vast training data set.
So you're saying it's okay because they do it to everybody? What? So their vast training set was also mostly copyrighted, so it's fine as long as we mix it all up? Even though it can be "unmixed"?. Nope. Sorry. Try again.
Re: (Score:3)
>While it's true that the information from the original work is in some sense copied into the AI model, that information is also thoroughly mixed with information from the rest of the vast training data set.
So you're saying it's okay because they do it to everybody? What? So their vast training set was also mostly copyrighted, so it's fine as long as we mix it all up? Even though it can be "unmixed"?. Nope. Sorry. Try again.
Being inspired by a copyrighted work isn't typically a copyright violation if a human does it, so logically it shouldn't be perceived as one if a computer does it, either, and the weights being mathematical tensor weights instead of changes to protein structures at the connections between neurons doesn't really change anything, IMO.
Therefore, training the model is no more making a copy than you reading a book is making a copy. If you inadvertently write a novel and start it with "It was the best of times,
Re: (Score:2)
If you can "untransform" the transformed work back into the original, your argument falls apart.
It happens you can do so.
Re: (Score:2)
If you can "untransform" the transformed work back into the original, your argument falls apart.
It happens you can do so.
If you can "untransform" it to a meaningful amount of the original work (i.e more than would be laughed off as obviously fair use), then maybe, but that also is a strong indication that the model doesn't have enough training data in a specific area, and that the code that runs the model isn't doing enough to reject queries that go down a path that hits nodes with limited training or whatever, i.e. it is a fundamental bug in the model or surrounding code if that can happen in practice.
Re: (Score:2)
There is no distribution or performance requirement for copyright infringement.
Re: (Score:3)
Isn't that a false dichotomy that misses (rather badly) that copyright is automatic? IF they use say appropriately licensed creative commons works,. those are still copyrighted worjs if created where copyright is automatic, for instance.
Maybe
Re: (Score:1)
Seriously, the RIAA and the like have done enough to fuck up people's understanding of copyright as it is.
To borrow from Mr. Adams: "A bunch of mindless jerks who'll be the first against the wall when the revolution comes."
Re: (Score:3)
"train LLMs with Public Domain media for free"
So we could get the level of literacy of Jane Austen and the Brontë sisters? James Fenimore Cooper? Daniel Defoe? Jack London? Edgar Allen Poe? H.G. Wells?
I can think of ever so many worse outcomes.
Of course when Chandler's and Christie's works hit public domain the LLM may get just a little paranoid.
The difference (Score:5, Insightful)
Re: (Score:1, Insightful)
It's one thing to be inspired by something, it's another to copy and paste their work into yours. "Generative AI" as it is implemented right now does not create anything. It merely copies.
Re: (Score:2)
"Generative AI" as it is implemented right now does not create anything. It merely copies.
Point one, IMO, seems hardly substantiated if one wants to get really pedantic in that a specific combination of elements not existing before exists now - seems in that basic context it'd be created whether man or machine does it.
Second one IMO DEFINITELY needs a citation. My understanding of LLMs and AI is rudimentary at best, but it seems like this is not how they are supposed to work, not even "as they are currently implemented. IMO if that is correct (that this notion is incorrect)... can we not?
Re:The difference (Score:5, Insightful)
It's one thing to be inspired by something, it's another to copy and paste their work into yours. "Generative AI" as it is implemented right now does not create anything. It merely copies.
No, not really. It is more like a series of probabilities. If you train on a book that includes the line "The quick brown fox jumped over the lazy dog," it might know that there's a 1% chance that "the" is followed by quick or lazy, a 1% chance that fox is preceded by quick and brown, a 1% chance that lazy is followed by dog, etc.
The original data is not present in the training set, except in very limited circumstances where there are not very many items in the training set that match against the query weights, and even then, it is still a statistical description of the original data, not the original data.
That set of weights basically turns into a beefed up version of what happens when you tap the middle button above your iPhone keyboard. If you feed it "The quick", it might sometimes follow a path in which several of those words appear in the expected order, or it might favor other contextual clues, and you might get "The quick farm tractor ate three brown eggs for middle school students in Germany."
Calling it copying is a rather severe misnomer. To the extent that anything resembling copying occurs, it is a fluke, and means that the model is too small.
Re: (Score:2)
The thing that everyone seems to forget is that in order to train an AI model on a work, you have to have a copy of that work. If said copy was not purchased, that's the copyright infringement. Training an AI on the infringing copy is irrelevant.
Whether the AI can produce similar output sufficient to be considered a derivative work is a separate quest
Re: (Score:2)
The thing that everyone seems to forget is that in order to train an AI model on a work, you have to have a copy of that work. If said copy was not purchased, that's the copyright infringement. Training an AI on the infringing copy is irrelevant.
Nope. Copyright law provides an automatic fair use exception for reading a book, looking at a website, playing a movie, etc. The things that a user would do when consuming some types of media are technically making a copy, but not in a way that is considered unlawful infringement. And browser caches and other similar behavior are also protected. The term for these in Title 17 is an "ephemeral copy". So barring any reason to treat AI training differently, an AI model "reading" or "watching" should also
Re: (Score:2)
Nope. "Fair use" is specifically and only for "quoting [or excerpting] brief segments for review." Reading a book has nothing to do with copyright. After you buy a book, whether you actually read it is irrelevant. You could use it to put under the short leg of your desk for all the publisher cares. But by purchasing, you have acquired an authorized copy, so no copyright infringement.
"Loo
Re:The difference (Score:4, Insightful)
associatively relevant parts of the overall model and synthesizing a result) rather than a (passive) query (for directly stored information).
Re: (Score:2, Insightful)
Remember back in the early days when ChatGPT became open to the public and we were all amused for a few weeks how inputing random hashes would bring up rather curious output? I once had it regurgitate the copyright page from a specific book, verbatim.
Re: (Score:2)
Re: (Score:2)
The model is irrelevant, pretending the model was of primary importance was an early and successful disinformation campaign which is now losing its effectiveness. As they say, the primary issue is "use of creative works for training".
Human reading and viewing is generally not considered copying, whereas copying is considered copying. JJ Abrams could view Spielberg's movies without license and without fair use, the AI companies need fair use for their training software to view Spielberg's movies ... because
Re: (Score:2)
Technically to view a movie on the internet I need a computer to copy it as well. Its up to the courts to decide if its fair use or not.
Re: (Score:2)
The transient copies are exempted by DMCA and before that implied license doctrine, which judges are never going to apply to allow AI training.
They copied a ton of stuff behind terms of service and paywalls too ... fair use is their only shot. Fair use or bankruptcy.
Re: (Score:2)
That's not different in kind to what the computer is doing copying it to be represented as many additions/substractions to many weights between the nodes of an artificial neural network.
That's the crux of the issue.
We're going to need some new understanding, in the legal system, of what is really going on with AI learning and output generation/synthesis.
Re: (Score:3)
They could change decades of jurisprudence and common sense of what constitutes a copy because it's convenient to AI companies ... or they could just not. I'm going to assume the judges will just not.
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Re: (Score:3)
The difference is the AI can do it for less than JJ Abrams. And that's why Hollywood loves it - because JJ Abrams might accidentally get creative and threaten the profits of the Hollywood studio.
I understand where the misunderstanding lies. While artists do learn from others, and even sometimes incorporate elements of others' work into their own, an artist is fundamentally different from computer equivalent of Mad Libs not in the volume produced, but in the creativity of what they produce. If you scal
Re: (Score:1)
How is this any different than Thomas Kinkade?
Re: (Score:2)
Bob Ross introduced an entire generation to the notion that anyone could paint, and Kinkade showed them they could build a successful business selling mediocre art. While both were very influential in the art world, neither produced work that was regarded as exceptional.
Re: (Score:2)
The reason it's cheaper is because it's worse. If you look at what these AI "artists", the guys who write the prompts and think they are creating this century's great masterpieces, it's 99.999% shit. It's derivative, obviously AI generated, and devoid of the creativity that a human artist would put into it. It's like bad CGI.
It's not just the fact that it is produced by taking artists' work as training data without recompence, it's that they are trying to eliminate the artists entirely. "Train your replacem
Artists... (Score:2, Insightful)
Re: (Score:3, Interesting)
The more I think about it, I wonder how much copyright being as long as it is has contributed to the perception of a lack of creativity not just in regards to companies milking existing IPs, but also in terms of fear of being sued over some stupid thing being too similar in w
Re: (Score:2)
In addition, no retroactive extensions would be a good fix as well. Authors who copyrighted their stuff pre-CTEA knew what they were signing up for. There was no reason to extend their copyrights.
Ants unite on the beach to stop the tide. (Score:2)
Based on the amount of money being thrown around when the buzz-bomb "AI" gets mentioned? There is no stopping the hoovering of creative work into the computer replacement for creative work. And even if *WE* stop it here, there's a whole world of countries out there that will just let their version keep chugging along, until our creative industries fade into oblivion due to the wash-over of the spilling tides, nay tsunami, of AI generated creative works. And as much as our creative industries have lowered th
Re: (Score:2)
Don't underestimate the ants. ;)
Give them their fair compensation (Score:3)
Most of their works ... (Score:5, Insightful)
... should already be public domain.
The founding fathers had it right. Seven years, extendable by seven years, once.
Re: (Score:2)
Copyright-infringement and plagiarism are not the exact same thing.
If you copy something, you often copy the name of the author/arist as well.
If you plagiarise something, you detach the attribution from the work.
There are many movie scripts based on old literary works but updated for more modern settings and situations.
But in no way does a serious script-writer hide the fact that their script was based on a work by Shakespare or Jane Austen, and try to pass the work off as their own, because that would be p
Accurate number (Score:3)
Actually fewer than 100 artists signed it. The rest were AI-generated bots.
Re: (Score:2)
The problem is that most artists think about money, there is little open source culture among them.
Open source minded people would rather see a strong open AI with less regulations and strict rules for companies commercialising things.
A lot of artists think mainly about whether they will get less commission and whether someone can make money from AI products. They don't see that open AI (not OpenAI) is a great opportunity, and that the companies trying to get laws that restrict everyone but them are taking
Re: (Score:1)
hear hear
it is rather dystopian to head in that direction
i for one hope we can collectively pull our global civilization out of this corrupt profit driven uneconomic darktimes
as an outsider... (Score:1)
Tuber (Score:2)
YouTube videos tracking all the movie "tributes" and paeans to what went before are some of the best viewing.
AI just cuts out the middle man.
Human artists steal too (Score:2)
Everything is based on what people previously did, otherwise all art would look like cave paintings. To claim someone can’t use your works decades after you’ve made money of it is stupid, because it’s not unreasonable to conclude someone else may have made something similar enough within that time frame.
Re: (Score:3)
AND btw we know ABBA steals melodies too: https://m.youtube.com/shorts/f... [youtube.com]
I notice a lot of the copyright Karen’s made money of public domain material. I mean if you were around in the 90s you will recall Disney was the most vicious copyright Karen. They got laws written to extend their copyright of Mickey Mouse in spite of the fact that some of Disney’s biggest hits such as Snow White and Sleeping Beauty came from the public domain.
Just following the money. (Score:2)
There is a simple fix for this, you own your own identity, performance, and art until you die + 5 years. Then it is public domain. For this to work corporations can not hold copyright, which fixes a lot of shit like the 70+ year Disney copyright shit.
Re: Get a grip, morons (Score:3)
That's not correct. The concept or copyright emerged in the 18th century. It originally applied only to written works such as books, and later to music.
Many composers re-arranged each other's melodies, such as Bach, Handel and Vivaldi, in a way that just wouldn't be allowed today.
The life expectancy was also somewhat less in earlier periods and there was plenty of art in previous centuries as well, long before copyright.
Opportunity Knocks - Art is human (Score:1)
Art doesn't stand still. And its always been a human thing since there have been human things.
I don't think the Artists have anything to worry about, if anything the value of their work will go up if the examples of AI art are to go by. It's programmer art. Looks synthetic, mass produced, a too perfect plastic cup next to hand painted porcelain, imperfection is style.
Looks like an opportunity for the lawyers.
Looks like an opportunity to hype AI
An an opportunity for us. Recorded music is getting cheaper, bec
Children should be banned from watching anything (Score:2)
"Allowing children to view our works is an unjust use of intellectual property. They might learn how to do this themselves!"
Re: (Score:2)
You're an ignorant idiot. Maybe the rest of us didn't learn about the letter untli after it was published?