News YouTube creator sues Nvidia and OpenAI for ‘unjust enrichment’ for using their videos for AI training

[Admin]

David Millette filed a case against Nvidia and OpenAI for 'unjust enrichment' for using his uploaded YouTube videos to train their AI models.

YouTube creator sues Nvidia and OpenAI for ‘unjust enrichment’ for using their videos for AI training : Read more

 

[Integr8d]

It's a very interesting question© Everyone should keep their eye on this© Truly a question of who owns what©

 

[Blessedman]

This is like suing Ford because their son was killed in a car accident involving a Ford vehicle. OpenAI I get, Nvidia??? Not so much. This will all end badly for this creator.

 

[ThomasKinsley]

This is like suing Ford because their son was killed in a car accident involving a Ford vehicle. OpenAI I get, Nvidia??? Not so much. This will all end badly for this creator.

The lawsuit is not based on Nvidia's hardware. Nvidia has their own AI models and is scraping the Internet to train it. Philosophically I understand the YouTube creator's complaint. He makes the videos and Nvidia and OpenAI use it to sell an AI product. I'm not entirely sold by Nvidia's argument that AI is merely learning information just like humans do. They are using his videos (or videos like his) for the express intent to sell a product that will, if successful, compete against his videos. The whole point of the AI is so you don't have to search for individual sources but can use the AI instead. For that reason, I think the argument that they are guilty of unjust enrichment might just work.

 

[Findecanor]

I'm not entirely sold by Nvidia's argument that AI is merely learning information just like humans do.

First, NVidias statement said "anyone"; not "anything".

Of course, the argument that artificial neural networks would "learn" like humans — often touted by AI-bro's — would best be likened to excrement from an uncastrated male bovine.
If you have any insight into how artificial neural networks actually function, you know very well that it is false ... for now.
It is theoretically possible that machines will be able to learn like humans in the future, but that technology has not been invented yet.

As of now, there is no learning being done when data is applied into a model.
"AI" technology of today is more similar to that used in search engine or data compression than to how an actual human brain functions.

But Nvidias statement did not actually said that their model learned. That would have been factually incorrect.
It was probably chosen with the intention of being confusing to those who do not know better, and be insulting to those who do, which it is.

Also, "Fair use" is not an argument — it is the statement being contested.

 

[vanadiel007]

I am thinking that at some point Nvidia and AI are going to get "scraped". Once that happens we will see what they truly believe when it comes to "fair use".

 

[derekullo]

I am thinking that at some point Nvidia and AI are going to get "scraped". Once that happens we will see what they truly believe when it comes to "fair use".

How do you scrape a black box?
The whole point of youtube/digital artists complaining about being scraped is it took them actual effort to create those works and they aren't being compensated for an AI being trained on it.
AI/LLMs/Diffusion models can create an unlimited amount of remixed works of varying quality much faster than an artist could.

 

[palladin9479]

Yeah ... this is gonna get very interesting. Unjust enrichment is a very open ended category of law and different states have different rulings and interpretations.

Virginia's
https://www.berliklaw.com/unjust-enrichment.html

The important part is that this type of claim does not require a preexisting contractual agreement, and is completely outside of the entire Copyright / Intellectual Property Theft debate. It boils down to "I put in effort to make a thing, someone else came along and used my effort to make money without compensating me".

If Nvidia was indeed scraping lots of youtube videos while using that to create a product to derive profit from, then it's going to almost impossible to wasn't unjustly enriching itself at the expense of the video creators.

 

[bit_user]

I was hoping the article would link the Youtuber's account. It'd be funny if they licensed their videos under Creative Commons, since that would pretty much destroy their case.

 

[DavidC1]

The whole point of the AI is so you don't have to search for individual sources but can use the AI instead. For that reason, I think the argument that they are guilty of unjust enrichment might just work.

The system itself is flawed. @Blessedman is likely right in the overall picture. Nvidia is too big of a company for laws to truly affect them.

Laws are good as the people who make them, there's nothing sacred or unbreachable about it, not even the constitution. It will fail in lockstep with the fall of the citizens(moral decay) and like all previous empires will end.

Most people don't care or don't know about it either. Why would government, with being drunk off the power do?

 

[bit_user]

The system itself is flawed. @Blessedman is likely right in the overall picture. Nvidia is too big of a company for laws to truly affect them.

There are plenty of counterexamples, like people winning lawsuits over cancer caused by herbicides and talcum powder. It's tempting to be cynical, but not always warranted. In my experience, cynicism is too often simply an excuse for laziness.

I won't address the rest of your post. Not only is it quite a reach, but we're not supposed to veer into politics, here.

 

[Vanderlindemedia]

Technically it's taking work of another, and thus copyright(s) are in effect. Youtube or not. I think it's fair to say that one should give permission.

I have been dealing with these things for a while, and learned quite quickly that enabling the blocking of AI Bots to scrape websites is a good start.

Google has been hit in the past with using paid content (behind paywalls) of for example news sites that saw revenue drop since google was publishing and indexing pretty much that content.

 

[Blitz Hacker]

Private machine learning/ai data scraping should be illegal.
Using a public good, for private/personal gain seems immoral to me, if it's self observation then well and fine. But an automated process using public data should only be 'fair use' if the result remains a public good (open source).
I suppose I could make an argument for perhaps a minor royalty for the processing/transformative process of it, (much like a street vendor taking your picture and selling it to you) but it would still have to remain in the public domain while using a public good.

 

[DavidC1]

There are plenty of counterexamples, like people winning lawsuits over cancer caused by herbicides and talcum powder. It's tempting to be cynical, but not always warranted. In my experience, cynicism is too often simply an excuse for laziness.

Absolutely not when it's the truth. Lobbying and looking away because it's convenient it's full of them. Few that gets won, many others that don't.

I won't address the rest of your post. Not only is it quite a reach, but we're not supposed to veer into politics, here.

Just another word for censorship actually.

 

[bit_user]

I suppose I could make an argument for perhaps a minor royalty for the processing/transformative process of it, (much like a street vendor taking your picture and selling it to you)

This does already happen, when machine learning companies want access to non-public archives. Examples include research journals and stock photos.

For an individual author, the monetary value of such royalties would probably be insignificant. AI companies couldn't afford to pay all the authors a non-trivial amount. Also, for them, the value of an individual work is negligible, so they would just exclude it from their training set, if an author actually had some way to make them pay a non-negligible amount for it.

Ultimately, the individual authors can't really win this war. The only ones potentially doing well off it are big companies and anyone who's able to use the resulting AI models in a profitable way.

 

[bit_user]

Just another word for censorship actually.

I honestly don't know why that policy exists, but surely you've seen threads get out of hand and hopelessly derailed when politics enters the picture?

IMO, tech is contentious enough. We don't need anything else to spawn arguments and sew divisions.

Anyway, the main reason I replied is just to point out that the proper place for discussion of forum policies is here:

Forum Rules & Feedback

forums.tomshardware.com

 

[vanadiel007]

How do you scrape a black box?
The whole point of youtube/digital artists complaining about being scraped is it took them actual effort to create those works and they aren't being compensated for an AI being trained on it.
AI/LLMs/Diffusion models can create an unlimited amount of remixed works of varying quality much faster than an artist could.

That is why I put "scraped" and not scraped. Different types of scraping.
Still wondering where all those lawyers from the days of Napster are, the ones making arguments about copyright violation when "sharing" music.
In a way, this scraping is making copies for the purpose of training AI LMM or whatever it's called. the AI is making a copy in part or all of it, with the copy being altered or not. Very similar to MP3 format of music.