[temp8830]

Even though the founders of OpenAI are not exactly someone you'd root for, comparisons to theft are silly.

By that token it would be illegal to go into a library, read a book, and actually remember what was in it. Except in this case the reader is a robot.

LLMs are such a fundamentally different thing that existing laws don't really make sense. Wait! Put the pitchfork down! I know, I know, stealing is stealing, and OpenAI founders are slimy. But what about derivative works? Why is a human making a hip-hop track allowed to sample, and a robot is not? Again, LLMs are such a fundamentally different thing that existing laws don't really make sense.

It's actually surprising in retrospect that nobody did this sooner. Even back in the 80s books about computers would gush about how a computer has enough memory to store an entire library's worth of books. It's just that someone finally figured out how to put an index on it.

Where I agree: given that this is basically the sum of all humanity's knowledge, the company should have been a non-profit. It was a non-profit. And then greed won.

[jcheng]

I think you make a good point but the use of samples in hip hop doesn’t support it; those samples need to be licensed.

[temp8830]

This is very much untrue, and the debate about exactly how much sampling constitutes fair use has gone one for many years and court cases.

[overgard]

There are a lot of things that are fine individually that are extremely problematic at scale. Like "reading a book" vs "ingesting all the books and art that ever existed into a plagiarism machine"

[mannanj]

If they illegally pirated books, i.e. downloading pdfs. thats illegal and piracy by no other name. If any of us do it, it's called pirating, why are you being disingenuous and saying it's not theft when a company does it?