Hacker News Reader

Comment by krainboltgreene | Hacker News Reader

It seems really unethical to train a model on someone else's words without their permission.

When I read your comment I trained my own mental model on your words. How is that any different? When a human reads words they apply a sophisticated theory of mind to contextualize the writing and the mental state of the author. If anything, LLM fine tuning is far less invasive than having a person read your writing.

This is an unserious argument and no one is swayed by it.

The idea that reading a piece of text constitutes copyright infringement is ridiculous. Copyright isn’t some infectious thing. Reading copyrighted text doesn’t give the copyright holder a claim to the future creative work of the reader.

You want to restrict model training, I get it. The debate is still ongoing, but I’m confident when these “copyright” claims work their way through court the AI companies will come out on top.

> The idea that reading a piece of text constitutes copyright infringement is ridiculous.

No man, it's not ridiculous. If I write a program that copies someone's book and try to sell it I'm infringing on that copyright. I cannot sell a zipped version of the Harry Potter books. I feel like there's so many people weighing in on this discussion who haven't actually done any real world copyright related stuff.

I see the source of your confusion. LLMs are not actually zips of the training dataset.

It is an incredibly common refrain amongst experts in the field that it is a compression of the dataset.

But that doesn't matter, because you clearly didn't understand what I was writing which doesn't shock me considering your position on LLMs.

> It is an incredibly common refrain amongst experts in the field that it is a compression of the dataset

This was a common idea three years ago. No one in the field seriously believes this today.

You're going to have to let a lot of scientists know that, because they're still publishing papers with that understanding. I guess they should have consulted you first.

Not their fault, hindsight is 20/20.

It may be just about as unethical as someone discussing something they overheard in a crowd.

Sure if you have a really warped understanding of context.

Sounds horrifically unethical to try to censor people like you’re doing.

That's not what that word means and I'm not surprised you don't know that.

Not to me.

Isn't that basically what OpenAI did?

It's what the original dataset did, and then they further compounded on that.

Reply on news.ycombinator.com