“The collecting society GEMA, which manages the rights of composers, lyricists and music publishers and has approximately 100,000 members, filed the case against OpenAI in November 2024.”
I mean every German who ever heard of GEMA knew this was inevitable all the way back in 2022. They threatened entire cities to the point some of them don‘t play music on Christmas markets anymore. You can‘t win against those guys, just hope for a hefty settlement.
GEMA is so bizarre. If you’re a musician and composer and want to do a live concert that consists only of songs written by you and there is an entrence fee, you’ll basically have to give them all of said fee or even more. They are destroying any culture really, only to give the money to some of the most successful musicians. And that even if you yourself are not part of GEMA, so don’t get ANYTHING in return.
GEMA was created by the Nazis to take over pop culture. There’s a certain logic there.
Of course, under the Nazis you could be sent to camp if you belonged to the wrong subculture. So there is a difference. The rebels of the time listened to jazz, to swing. “Negro music” was the social media of that age, corrupting the youth.
deleted by creator
So, if my brain ‘ingested’ a song lyric and now I can sing it, does that mean I owe GEMA royalties every time I hum in the shower? asking for a friend. who is me. my friend is me.
Unless you are performing it for others, or for profit, they generally don’t care.
My understanding is that it’s for learning only. You should be safe for your private singing purposes. It seems like you would owe them royalties if you created your own song lyrics, after having read/heard one of theirs…
They sure would like that.
Never thought I’d hail the GEMA for what they’re doing…
I think “ingesting” is more accurate than “learning”.
Illegaly ingesting. I hate it that private people are fined when they look wrong at a minor hash of a snippet of data but when a corporation does it in full and monetizes it, it’s fine!
How is it an “undisclosed [amount of] damages” - aren’t court rulings supposed to be public [records]?
I know stuff like the rulings of me going to court with the driver and insurance companies involved in a car accident certainly was public (in Germany).
Anyone in the know please ELI5 and/or point me at the relevant laws, how/what/in which cases (parts of) rulings can be “undisclosed”.
Is ChatGPT a legal entity competent to violate copyright law? I don’t think that’s likely.
I do think OpenAI violated copyright law by copying song lyrics and other media to use them as input to their LLM systems, to embed the essence of them into their LLMs for commercial benefit. Judging by the valuations of the companies that do not yet have significant income compared to the investments, on the face of it, the IP they copied, often without license, as far as I know, is fantastically valuable.
Does that mean song lyric websites with ads also violate copyright laws?
According to Are Song Lyrics Copyrighted? How the Law Works, unless their use is ‘fair use’ or they have a license, then they are violating copyright, if I understand the article correctly. I believe that site explains laws in the United States. It probably varies somewhat by jurisdiction, so I expect it would depend on who owns the website and where they are based.
To tag along with this, I remember this becoming an issue 10 or 15 years ago and a lot of the big lyrics websites were forced to reach licensing agreements with the songwriting groups like ASCAP and BMI (they collect and distribute royalties on behalf of the writers). I think a couple sites tried going to court to claim fair use but lost pretty quickly. That’s pretty established law going back to the earliest days of music publishing. Just because they were publishing online instead of printing up songbooks doesn’t mean the laws change.
I have been thinking of adding a license clause to everything I make (code especially) that makes any AI trained on it my sole exclusive property, but I don’t know how defensible that would be in court ?
Or any other sort of trap clause. But again, I don’t know how to word it. Like “this makes your model public domain” or “you grant a free worldwide unlimited license to every human on earth”.
Something that makes the mere inclusion of the code in a training data set into absolute legal poison to the would-be owners.
I am not a lawyer, not even slightly…
I understand (to agree degree) going after AI companies for reproducing the lyrics in a way that would not normally be protected by copyright but outright scraping is going too far from a moral standpoint.
There’s a good argument to be made about abusing their resources to do the scraping as I’ve heard complaints of site owners getting overwhelmed by AI crawlers but provided you’re not doing that I think scraping should be allowed generally speaking even if the operator disallows it, since without that search engines break and archival (especially to prove malice) go out the window.
I’m inclined to take an approach of “you can ingest whatever you want, but you are liable for reproduction, and if preventing reproduction is too onerous, then you probably should get the licences to permit it or don’t ingest that data”. Even that has some caveats since that reasoning would decimate social media services and personal/community spaces if actually enforced which is kinda what Safe Harbor helps protect.
deleted by creator
