3 ms·
Everyone knows the training data is stored in some way in the LLM. The point is the use of the copyrighted material is transformative. Remember google books, it
by johanyc 1y ago
Everyone knows the training data is stored in some way in the LLM. The point is the use of the copyrighted material is transformative. Remember google books, it literally shows photocopy of pages of books but the court ruled it’s fair use. A simplified explanation is book vs search engine and book vs ai chatbot are very different from each other.
- SideburnsOfDoom 1y ago"a photocopy of pages of books" is exactly that, pages of an existing book. It doesn't pretend to be something else. The output of a LLM, when based heavily on that same page, pretends to be something novel. IMadeThis_Meme.gif