Nasty little bureaucratic tyrants. EU needs to get their shit together or they're going to be quibbling over crumbs while the rest of the globe feasts. I'm not inclined to entertain any sort of bailout, either.
Here in the states, we routinely let companies fuck us up the ass and it's going great! Right, guys?
Reproducing a copyrighted work 1:1 is infringing. Other sites on the internet have to license the lyrics before sending them to a user.
So far all I've tried are willing to return a random phrase or grammar used in a song, so it is only getting to asking for a line of lyrics or more that it becomes troublesome.
(There is also the problem that the LLMs who do comply will often make up the song unless they have some form of web search and you explicitly tell them to verify the song using it.)
I know no one wants to hear this from the cursed IP attorney, but this would be enough to show in court that the song lyrics were used in the training set. So depending on the jurisdiction you're being sued in, there's some liability there. This is usually solved by the model labs getting some kind of licensing agreements in place first and then throwing all that in the training set. Alternatively, they could also set up some kind of RAG workflow where the search goes out and finds the lyrics. But they would have to both know that the found lyrics where genuine, and ensure that they don't save any of that chat for training. At scale, neither of those are trivial problems to solve.
Now, how many labs have those agreements in place? Not really sure? But issues such as these are probably why you get silliness like DeepMind models not being licensed for use in the EU for instance.
As for searching for the lyrics, I often have to give it the title and the artist to find the song, and sometimes even have to give context of where the song is from, otherwise it'll either find a more popular English song with a similar title or still hallucinate. Luckily I know enough of the language to identify when the song is fully wrong.
No clue how well it would work with popular English songs as I've never tried those.
2. LLMs typically don't produce content verbatim. Some LLMs do provide references but it remains a pasta of sentences worded differently.
You are asking for gpt to publish verbatim content which may be copyrighted, it would be deemed infringement since non verbatim is already crossing the line.
Not for any particular reason, it flat out refuses. I asked it whether it could describe the picture for me in as much detail as possible, and it said it could do that. I asked it whether it could identify a movie or TV series by description of a particular scene, and it said it could do that, but that if I'd ever try or ask it to do both, it wouldn't do that cause it'd be circumvention of its guide lines! -- No it doesn't quite make sense, but to me it does seem quite indicative of a hard-coded limitation/refusal, because it is clearly able to do the sub tasks. I don't think the ability to identify scenes from a movie or TV show is illegal or even immoral, but I can imagine why they would hard code this refusal, because it'd make it easier to show it was trained on copyrighted material?