Nvidia Allegedly Scraped YouTube, Netflix Videos for AI Training Data 37
Nvidia scraped videos from YouTube, Netflix and other online platforms to compile training data for its AI products, 404 Media reported Monday, citing internal documents. The tech giant used this content to develop various AI projects, including its Omniverse 3D world generator and self-driving car systems, the report said. Some employees expressed concerns about potential legal issues surrounding the use of such content, the report said, adding that the management assured them of executive-level approval. Nvidia defended its actions, asserting they were "in full compliance with the letter and the spirit of copyright law" and emphasizing that copyright protects specific expressions rather than facts or ideas.
It's amusing when Big Data players (Score:5, Funny)
steal from each. And it's a refreshing change: usually they steal from us.
Re: (Score:2)
_________ scraped youtube for _________
Fill in the blanks. Preferably with highly lucrative lawsuit targets.
Re: (Score:2)
They stole from media companies too and those do tend to register their copyrighted works ... even NVIDIA isn't big enough to pay for all statutory damages they are risking. NVIDIA was insane to expose themselves to this, leave the piracy to the customers.
Re: (Score:1)
Even the world's most valuable company can't afford to pay what are (to summarise in a way connected to reality) imaginary damages caused by an AI 'seeing' copyrighted material.
Truly sobering.
Re: (Score:3)
Re: (Score:1)
Perhaps one day an AI will be able to untangle the system known as Law, identify those aspects which are a result of sponsorship by those with power and filter them out to make "Law 2.0 - payback's a bitch" - in theatres now (also on BT :-)
Reading above, I seem to be this guy with a different focus:
* https://www.youtube.com/watch?... [youtube.com]
Re: (Score:2)
What the model is doing is entirely irrelevant to my argument.
What NVIDIA did is make copies. Not transient, not transformative, just plain old semi-permanent copies to a storage array.
umm so what? (Score:1)
It's all in the public domain.
Your ownership rights ended when you posted your video on youtube.
Re: (Score:2)
Google probably figures it belongs to them.
Re: (Score:3)
No they dont, they explicitly say that in EULA. However in the same breath they also forbid you to do anything else with the content on youtube besides watch it:
from youtubes EULA:
"You retain ownership rights in your Content. However, we do require you to grant certain rights to YouTube and other users of the Service, as described below."
License to Other Users
You also grant each other user of the Service a worldwide, non-exclusive, royalty-free license to access your Content through the Service, and to use
Re: umm so what? (Score:2)
"and to use that Content, including to reproduce, distribute, prepare derivative works, display, and perform it, only as enabled by a feature of the Service (such as video playback or embeds)."
Only emailed by a feature SUCH AS blah blah. But another feature of the service is serving the video stream to you. I think that's vague enough to take to court.
Re: (Score:2)
For clarity, this license does not grant any rights or permissions for a user to make use of your Content independent of the Service.
Making use of is not a copyright issue. Copyright only concerns the reproduction and performances of copyrighted works. It says nothing about use beyond that.
Re:umm so what? (Score:5, Informative)
Uh, what? You own videos even if you upload them.
The agreement you enter into when you do use a service like YouTube is you grant them non-exclusive license to use, reproduce, distribute, etc. You certainly diminish the exclusivity of your content the moment you upload it to YouTube. But it isn't the same as being "public domain"
You can still sell and distribute your creations, even if YouTube has the right to do pretty much whatever they want with your video. That's the "non-exclusive" part of the agreement.
It's a terrible deal of course. But the only deal in town. It means we've essentially fallen for the classic artist trope of doing free art/music for "the exposure"
Re: (Score:2)
legally, no. Everything is automatically copyrighted until it enters Public domain, which is a very small edge case, long after the useful life of the work. All published work will be protected for the life of the creator plus 70 years.
Morally and ethically? Yes knowledge is free and should be enshrined as such in law. But it really isn't possible. You have to license it as such. (eg GPL https://en.wikipedia.org/wiki/... [wikipedia.org]).
Re: (Score:2)
from youtubes EULA: https://www.youtube.com/static... [youtube.com]
Important parts:
Rights you Grant
You retain ownership rights in your Content. However, we do require you to grant certain rights to YouTube and other users of the Service, as described below.
License to Other Users
You also grant each other user of the Service a worldwide, non-exclusive, royalty-free license to access your Content through the Service, and to use that Content, including to reproduce, distribute, prepare derivative works, display, and perform
Re: (Score:2)
While it's clear it mandates grants of certain rights, this notably *explicitly* would exclude the alleged behavior.
only as enabled by a feature of the Service
this license does not grant any rights or permissions for a user to make use of your Content independent of the Service.
Re: (Score:2)
Morally and ethically? Yes knowledge is free and should be enshrined as such in law. But it really isn't possible.
Knowledge is not copyrightable.
Re: (Score:2)
It's all in the public domain. Your ownership rights ended when you posted your video on youtube.
You are either trolling or colossally naive. YouTube creators own the copyright on their masters. However, YouTube can do whatever they want with the videos because creators grant "...YouTube a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to use, reproduce, distribute, prepare derivative works of , display, and perform the Content in connection with the Service and YouTube’s (and its successors’ and affiliates’) business, including without limitation for p
So AI streams, views, not copies ... (Score:2)
Re: (Score:2)
So we train AI by letting it view public content on YouTube through the standard streaming protocols. No copyright violation. The AI is just recording its interpretation of what it just saw, just like a person.
Sure, it's not like that's not going to take waaaaayyyyy more time that scraping.
It will however deliver more data than zero, which is what they will have once unable to scrape.
Furthermore, intention matters. A person watching a YouTube video does not intend, as a rule, to make tons of money off of it, ...
Untrue, people often watch videos to learn things. Sometimes things that can be commercialized, used at a job.
... they just wish to be entertained or informed. An AI bro doing the same to train up his AI, make tons of money doing it and possibly even with the intention to put the original content creators out of business which is a completely different set of motivations and definitely something you can sue them over.
And then you lose the lawsuit. If you want to stop an AI bros you need something in your terms of use prohibiting viewing by AI.
Re: (Score:2)
That is so far from reality it is farcical. Please go home and come back when educated.
Re: (Score:2)
There are entire generations now that do not understand copyright, why it was invented or what public domain even means.
I hope you never write a book, software, movie or create an artwork.
Everyone is scraping everyone (Score:3)
Everyone will gripe that their data is getting scraped while furiously scraping everyone else. It's really that simple.
There is lots of content on YouTube (Score:2)
There is lots of content on YouTube that has a public domain license. This is not telling us if Nvidia did anything illegal sense it makes no mention of the license of the said content.
Re: (Score:2)
Even if the work itself is licenced as such i think an argument can be made that NVIDIA may not access it FROM YOUTUBE because YouTube TOS still forbids it. and since by uploading to youtube, you agree to their TOS, it licenses the work to YouTube for YOUTUBE to redistribute, under THEIR terms, not yours.
Re: (Score:2)
You can make any argument you want, however, the argument that matters is the one that holds up in court.
The specific expressions are on their storage (Score:3)
The training sets stored by most AI companies are no different from the big public shadow library (direct copies of which they also contain). All copied without a license.
Ahoy matey.
Scraping? (Score:2)
When an intelligence is doing it, it's called 'watching'.
Don't impede progress (Score:1)
any actual evidence of copying? (Score:2)
Has anyone been able to link a specific piece of human art as the 'original' for a specific piece of modern 'AI' art?
If I can view it freely, so should AI (Score:3)
The double standard being applied to AI applications is ridiculous. Those video are up on youtube for anyone to see and learn from, even AI. This is another cash-grab in disguise. Greedy artists and corporations who already got paid, are trying to get paid again based on anti-AI hype.