Hacker News Reader

Comment by inigyou | Hacker News Reader

Why is everyone talking about human analogies when LLMs are not humans?

Because the process is the same. When you read a book you don‘t save it as a brain file, you form memories from it. Some people can recall verbatim bits here and there but I have never met someone regurgitating a book word for word. And I‘m pretty sure I can not ask ChatGPT to output the first chapter of Moby Dick word for word. I think that would be, rightfully, considered copyright infringement.

https://arxiv.org/pdf/2601.02671

Why is it different other than, "just cause?" No one seems to have actual reasoning to back it up while it feels very similar the other way around, is human brains and neural nets (notwithstanding that they're both called neurons) seem to learn similarly and can act on similar classes of problems like language and mathematics.

are you asking what the diff is between a human and an LLM?

LLMs are just tools wielded by humans. There's always a human behind whatever is done.

I asked why, not what.

why is there a difference between a teapot and an old lady?

I didn't ask why two things are different, I asked what difference does it make, other than just cause? In other words, I don't ask why there is a difference between an apple and an orange (which is nonsensical to ask, just as you have discovered with your question here), I ask why it matters whether I eat an apple or an orange if I'm hungry.

What's the difference between humans and LLMs? The difference is that the creators of laws are humans, and the purpose of laws is for humans. Authors write books with the expectation that they will be read by humans, and copyright law was written with unstated assumptions, like the fact that books have an effect on a person's mind after reading it.

IMO the spirit of the law would prohibit LLMs from training, and the letter of the law leaves room for that only because nobody thought to write down "books are for people to read".

Laws are not only for humans, there are laws for bots as well, like anti spam, but that's besides the point because in reality behind LLMs there are humans and so humans still control them, therefore laws target them too, and now it looks like humans using LLMs to train via ingestion of books is deemed fair use.

I agree that laws don't only apply to humans, but they are only for humans. The legal system is for our benefit.

> therefore laws target them too, and now it looks like humans using LLMs to train via ingestion of books is deemed fair use

Which is orthogonal to my point.

AI benefits me and many other humans, so it is working to our benefit as intended, per your logic.

There are several distinct differences, I think the most pertinent one is that you can create seperate instances of the LLM that ingested that media without reingesting the media (i.e. you can copy the LLM but you can't copy a human). Copright law is screwed up for multiple reasons but it makes sense to treat LLM as different than a human consuming the media, especially if the LLM is created for profit (e.g. movie owner selling tickets to a dvd he bought in the store vs a home viewing is treated differently).

I don't understand how the first pertinent one is relevant, when the LLM can't recreate verbatim at length an entire work anyway. It is literally unable to based on information entropy, it is not big enough to contain all of the bits of the training data at maximum theoretical compression.

But if a single person "can't recreate verbatim at length an entire work anyway" why does copyright law treat playing music to guests in your house differently to playing it to a classroom? I'm not saying that copyright law isn't stupid I'm saying that these distinctions make a difference and the fact that you can't copy humans is a pertinent distinction. LLMs would be treated very differently if like humans you had to spend the training costs of each model for each running instance of that model, the fact that models can be copied is a core feature of these models and makes them distinct from showing 'training data' to a human.

Reply on news.ycombinator.com