Actually no that would be copyright infringement. Badly signing a recent pop song in public also qualifies as copyright infringement. Public performances count as copying here.
For commercial purposes only. If someone sells a recreation of the Harry Potter book, it’s illegal regardless whether it was by memory, directly copying the book, or using an LLM. It’s the act of broadcasting it that’s infringing on copyright, not the content itself.
But just for clarification, selling a recreation isn’t required for copyright infringement. The copying itself can be problematic so you can’t defend yourself by saying you haven’t yet sold any of the 10,000 copies you just printed. There are some exceptions that allow you to make copies for specific purposes, skip protection on a portable CD player for example, but that doesn’t apply to the 10k copies situation.
Having said that I think the cat is very much out of the bag on this one and, personally, I think that LLMs should be allowed to be trained on whatever.
This kind of argument keeps popping up usually to justify why training LLMs on protected material is fair, and why their output is fair. It's always used in a super selective way, never accounting for confounding factors, just because superficially it sort of supports that idea.
Exceptional humans are exceptional, rare. When they learn, or create something new based on prior knowledge, or just reproduce the original they do it with human limitations and timescales. Laws account for these limitations but still draw lines for when some of this behavior is not permitted.
The law didn't account for a computer "software" that can ingest the entirety of human creation that no human could ever do, then reproduce the original or create an endless number of variations in a blink of an eye.
Traditionally tools that reduce the friction of creating those transformations make a work less “transformed” in the eyes of the law, not more so. In this case the transformation requires zero mental or physical effort.
This supposed failure to see the difference between the human mind and a machine whenever someone brings up copyright is peformative and disingenuous.
Maybe you've been following a different conversation, or jumping to conclusions is just more convenient. This isn't about "legal status of AI" but about laws written having in mind only the capabilities of humans, at a time when systems as powerful as today's were unthinkable. Obviously the same laws have to set different limits for humans and machines.
There's no law limiting a human's top (running) speed but you have speed limits for cars. Maybe you're legally allowed to own a semi-automatic weapon but not an automatic one. This is the ELI5 for why when legislating, capabilities make all the difference. Obviously a rifle should not have the same legal status or be the same thing as a human, just in case my point is still lost on you.
Literally every single discussion on this LLM training/output topic, this one included, eventually has a number of people basing their argument on "but humans are allowed to do it", completely ignoring that humans can only do it in a much, much more limited way.
> is peformative and disingenuous
That's an extremely uncharitable and aggressive take, especially after not bothering to understand at all what I said.
To be clear, my intent wasn't to say you were the one being performative and disingenuous. I was referring to the sort of person you were debating against, the one who thinks every legal issue involving A.I. can be settled by typing "humans are allowed to do it."
Since I replied to you, I can see how what I wrote was confusing. My apologies.
The parent you replied to claimed LLMs are using "mechanism similar enough to what humans do and what humans do is fine."
Parent probably doesn't want his or her brain shredded like an old hard drive despite claiming similar mechanisms whenever it is convinient.
I'm arguing nobody actually believes there are "similar mechanisms" between machines and humans in their revealed preferences in day to day life.
>There's no law limiting a human's top (running) speed but you have speed limits for cars. Maybe you're legally allowed to own a semi-automatic weapon but not an automatic one.
I don't believe this analogy works. If we're talking about transmitting the text of Harry Potter, I believe it would already be illegal for a single human to type it on demand as a service.
If we are talking about remembering the text of Harry Potter but not reciting it on demand, that's not illegal for a human because copyright doesn't govern human memories.
I don't see what copyright law you think needs updating.
Claims like this demonstrate it, really: it is obviously not copyright infringement for a human to memorise a poem and recite it in private; it obviously is copyright infringement to build a machine that does that and grant public access to that machine. (Or does anyone think that's not obvious?)
Not fair use. No one would ever prosecute it as infringement but it's not fair use.
The LLMs I've used don't randomly start spouting Harry Potter quotes at me, they only bring it up if I ask. They aren't aiming to undermine copyright. And they aren't a very effective tool for it compared to the very well developed networks for pirating content. It seems to be a non-issue that will eventually be settled by the raw economic force that LLMs are bringing to bear on society in the same way that the movie industry ultimately lost the battle against torrents and had to compete with them.