Found in NeMo.
the authors said the takedown reflects Nvidia’s having “admitted” it trained NeMo on the dataset
Oooh so they’re fucked.
THOUGH… let’s say I bought a book, why am I not allowed to learn from that book, and write in a similar style to that book. I am? Well why can’t I train an AI to use that book and have it write in a similar style? I’m not sold on “I must give you permission to use my book to train an AI.” Maybe if I agreed to those terms BEFORE buying the book, but it seems odd that someone can bar me from doing that AFTER buying the book. And just because “we never thought about that” isn’t really a good excuse to change the rights for someone who bought the book.
Though if anything this basically proves the old adage. “Don’t tell anyone what’s in your AI’s training data”
Humans are not generally allowed to do what AI is doing! You talk about copying someone else’s “style” because you know that “style” is not protected by copyright, but that is a false equivalence. An AI is not copying “style”, but rather every discernible pattern of its input. It is just as likely to copy Walt Disney’s drawing style as it is to copy the design of Mickey Mouse. We’ve seen countless examples of AI’s copying characters, verbatim passages of texts and snippets of code. Imagine if a person copied Mickey Mouse’s character design and they got sued for copyright infringement. Then they go to court and their defense was that they downloaded copies of the original works without permission and studied them for the sole purpose of imitating them. They would be admitting that every perceived similarity is intentional. Do you think they would not be found guilty of copyright infringement? And AI is this example taken to the extreme. It’s not just creating something similar, it is by design trying to maximize the similarity of its output to its training data. It is being the least creative that is mathematically possible. The AI’s only trick is that it threw so many stuff into its mixer of training data that you can’t generally trace the output to a specific input. But the math is clear. And while its obvious that no sane person will use a copy of Mickey Mouse just because an AI produced it, the same cannot be said for characters of lesser known works, passages from obscure books, and code snippets from small free software projects.
In addition to the above, we allow humans to engage in potentially harmful behavior for various reasons that do not apply to AIs.
- “Innocent until proven guilty” is fundamental to our justice systems. The same does not apply to inanimate objects. Eg a firearm is restricted because of the danger it poses even if it has not been used to shoot someone. A person is only liable for the damage they have caused, never their potential to cause it.
- We care about peoples’ well-being. We would not ban people from enjoying art just because they might copy it because that would be sacrificing too much. However, no harm is done to an AI when it is prevented from being trained, because an AI is not a person with feelings.
- Human behavior is complex and hard to control. A person might unintentionally copy protected elements of works when being influenced by them, but that’s hard to tell in most cases. An AI has the sole purpose of copying patterns with no other input.
For all of the above reasons, we choose to err on the side of caution when restricting human behavior, but we have no reason to do the same for AIs, or anything inanimate.
In summary, we do not allow humans to do what AIs are doing now and even if we did, that would not be a good argument against AI regulation.
Mickey Mouse is public domain…
Only the very early version, “Steamboat Willie”. And only insofar as it is not a trademark. You can, however, perfectly legally use copyrighted characters. EG The copyrighted Mickey Mouse appears multiple times in South Park. Copyright ideologues always try to make people forget that. They want every human thought and every idea to be owned as “property” and used to extract rent money.
deleted by creator
Everyone here seems to be missing the point and haven’t read the article. Nvidia isn’t being targeted because they make the hardware that enables training ai. They are being sued because they trained an ai using the authors books.
There’s no difference. There’s also no infringement. Show me the copy.
There’s is a huge difference though.
That being one is making hardware and the other is copying books into your training pipeline though
The copy occurs in the dataset preparation.
Why do you hate AI? It’s a useful tool to allow humans to work less? Or have you bought into the capitalist lie that we must work 9-5 5 days a week for an employer to have purpose?
I hate it because it’s 80% hype garbage and 10% suck. The remaining 10% varies between “instant code check” and “might be cool someday”, so it’s in the “we’ll see” column. Not worth the garbage though.
You’ve set up a false dichotomy. There are reasons to dislike AI besides capitalist propaganda. For example, moral concerns with training on data without explicit approval
They may as well sue bic for all of those copyright infringement enabling pens the company made
Except the pens weren’t covered in snippets of copyrighted text.
My analogy was going to be: suing the company that makes the glass for photocopiers.
deleted by creator
This feels like suing gun manufacturers over murder. They made the tool but they’re not the ones responsible for the crime.
https://www.theguardian.com/us-news/2023/may/27/gun-lawsuits-manufacturer-sellers-crimes
"That led to the 2005 Protection of Lawful Commerce in Arms Act (PLCAA), which effectively granted gun manufacturers and sellers immunity from lawsuits related to harm “resulting from the criminal or unlawful misuse” of guns.
Unlike other industries, the gun industry was now largely free to act as it chose."
I doubt the kitchen knife industry is worried about getting sued, but they are readily available to use in stabbing.
Yeah that “unlike other industries” is a load of crock. Nobody goes after car manufacturers for all those deaths either unless there’s an actual flaw in the car itself that caused it.
Hell we barely go after the pharmaceutical industry and they KNOWINGLY cause harm when they downplay all the risks that they do.
+1 for the pharmaceutical industry. Those motherfuckers are actively causing harm. They pay doctors to over prescribe, which leads to addiction, which leads to more pharmaceutical sales. They are basically with the tobacco industry 
It’s more like charging the iron ore mining companies over gun murders.
NVIDIA doesn’t have any say over how their GPUs are used.
Creating material that is copyright infridgement is not a desired output, the purpose of guns is to kill (when used). AI manufacturers depend on copyrighted material to “train” the AI, the method of creation makes it more likely to infridge.
Creating material that is copyright infridgement is not a desired output
Agreed.
the purpose of guns is to kill (when used).
Guns is a term with varied definitions of which not all are intended to kill. There are rubber bullets, air soft, small caliber, and even paint ball guns. These MAY be lethal but were made with other goals in mind.
Nvidia on the other hand made GPUs for applications that revolve around video, the G literally stands for graphics. Some people found out that they are also efficient at other tasks so Nvidia made a new line of products for that workload because it was more lucrative. Gamers usually only buy 1 graphics card per machine, a few years ago some would even buy up to 3. In contrast, AI researchers/architects/programmers buy as many as they can afford and constantly buy more. This has made Nvidia change their product stack to cater to the more lucrative customer.
AI manufacturers depend on copyrighted material to “train” the AI
With everything I said, these AI creators CHOOSE what to feed into these new tools. They can choose to input things in the public domain or even paid-licensed-content but instead using copyrighted and pirated content is the norm. That is because this is a new field and we are collectively learning where the boundaries are and what is considered acceptable and legal.
Reddit recently signed a deal to license it’s data (user generated content like posts and comments) for use with AI generation. Other companies are using internal data to tailor their AIs to solve field-specific problems. The problem is that AI, just like guns, is a broad term.
the method of creation makes it more likely to infridge.
Nvidia has given us the tools but until we define what is considered acceptable, these kinds of things will be inevitable. I do believe that the authors had their copyrights infringed but they are also going after the wrong people. There have been reports of AI spitting out full books on command, clearly proving that those works were used to train. The authors should be going after the creators of those specific AIs, not Nvidia.
There is a long and bumpy road ahead.
deleted by creator
I think it’s more like the gun manufacturers themselves sueing the US army
Since some folks in this thread can’t be bothered to click on the article, here is the first sentence:
Nvidia, whose chips power artificial intelligence, has been sued by three authors who said it used their copyrighted books without permission to train its NeMo AI platform.
In this case, “it” is Nvidia. They are training their own model, not just selling chips.
I didn’t even knew they had a model themselves.
of course they have trained models in house. they had ai write a song during a meme conference last year
they also use ai to train a model for DLSS image upscaling in house.
Here is an alternative Piped link(s):
they had ai write a song during a meme conference last year
Piped is a privacy-respecting open-source alternative frontend to YouTube.
I’m open-source; check me out at GitHub.
Found in NeMo
Guys! He found him!
Several comments making the some gun manufacturer analogy didn’t bother to read 3 sentences from article.
Lemmings showing their Redditage (Reddit heritage)
Not in a million chance this goes through. Can you sue Intel/AMD for making processors that could be used to develop malware? No.
Intel/AMD created a bunch of malware to test their processors that infringed on copyright? I mean. . . yeah if they did then sure.
They did have those people trying lawsuits against firearm manufacturers for any misuse of their firearms, until Congress blocked that in 2005.
Ah, the wonderful “Child Safety Lock Act of 2005” lol. Love politician names for this shit.
Seems to me like a reasonable criterion would be to determine if the trained model outputs copyright-infringing text in response to non-infringing prompts.
The ability to create a copy of a copyrighted work from ai shouldn’t be any more relevant than with a copy machine.
Good
