logoalt Hacker News

protocolturetoday at 3:40 AM2 repliesview on HN

>It's clear to me that AI training is transformative fair use under existing law.

I wouldn't even go that far. Its an entirely new product. Its like the guy who sold you the keyboard demanding royalties for the software you built.

That the person who wrote the book couldn't predict a new use case for the book in training LLMs, is irrelevant. The book isn't in the LLM. Its not being sold with the LLM. Its one of billions of tools used to create the LLM.

People try and sell this as the AI companies extracting value from the poor little IP holders like Disney. Its maddening. That content is your cultural heritage. It already belongs to you, just some idiot has been granted a lifetime of exclusive exploitation. An LLM is trained on data you already own. Disney et al wants to exploit the new technology to extract even more money out of stuff created often decades ago.

At absolute worst its reverse engineering, which was supposed to be fair use protected in the US but apparently that's been somewhat eroded.


Replies

xigoitoday at 4:28 AM

> The book isn't in the LLM.

An LLM is essentially a lossy compression of the training data. The book absolutely is in there, it’s just mangled to the point of unrecognizability.

show 1 reply
gizajobtoday at 4:56 AM

If my book isn’t in your LLM, then prove it and don’t use my book to train your LLM.

show 1 reply