The US is placing a big bet on AI to break the economic secular stagnation of recent decades. If IP is an obstacle to that, it would be great if IP laws are loosened.
The NYT could prolly win under current law - due to the interpretation of how GPT is reproducing the articles. The judge that may ultimately decide the future of how AI will fall for now in the US, will likely know very little about AI/LLMs.
I don't think reproducing a current article is that big deal tbh - it just retrieved the info from itself rather than the NYT bc it had already saved the info before and can't forget things. That's humanely possible by anyone with a photographic memory - it's more similar to recollection than infringement. The NYT is pissed the LLMs learned from their data - you can't stop something from learning - especially after the fact.
Obviously, the user intent not machine intent is key also. No infringement can happen without a prompt that suggests copyrighted material - easy to do or not, requires a person nonetheless.
The technical aspects of this is where they have a chance, taking a step back, it all just looks kind of silly.
You use the notes I gave you verbatim in your paper and are found guilty of plagiarism.
You obviously didn't realize the notes I gave you on the spot were from a textbook - that wasn't what you were looking for. I didn't realize you were just going to use whatever I wrote in your paper.
Did I commit plagiarism or did you?