Hacker News Reader

Comment by evdubs | Hacker News Reader

> If my college textbook's author demanded a percentage cut of my future earnings I'd be much more peeved than content creators are now.

If you were writing highly derivative textbooks, your college textbook's author would be peeved, too. That is the AI analogy, not that you read a chemistry textbook and are then breaking some law when you go do chemistry.

View on HN

You think chemistry textbook authors haven't read chemistry textbooks? Heck, you think they haven't read chemistry textbooks as references, while writing their own textbooks?

Again, it's not about reading. It's about either acquiring in a way that is infringing or producing output that is infringing. If you write a chapter with something like, "Ionic bonds are bonds that occur between two elements where one or more electrons transfer to a particular element," it's not copyright infringement for another author to discuss ionic bonds. It's copyright infringement to copy the material. It's copyright infringement to produce output that is the same or substantially similar.

Are you assuming that all or most AIs generate derivative passages from textbooks when you ask them Chemistry questions? I'd be surprised if that was the case for even a single modern LLM.

I assume that all generative AI output is derivative of the training material (and the prompt). Because, what else would it be?

Copyright does not protect ideas, just specific expressions of ideas. So even if something sounds like a derivative work, it doesn't mean it's infringing. You can write stories about a fictional 19th century detective living on Boulanger Street in London without infringing on Sherlock Holmes. You can write reviews of Disney movies without Disney's permission.

AI does not understand what ideas are. AI does not think. It just takes source material and regurgitates it.

Reply on news.ycombinator.com