Study: Meta AI model can reproduce almost half of Harry Potter book
The research could have big implications for generative AI copyright lawsuits.
https://arstechnica.com/features/2025/06/study-metas-llama-3-1-can-recall-42-percent-of-the-first-harry-potter-book/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social
@arstechnica So, ChatGPT is either half of an infinite number of monkeys with typewriters and an anti-trans bias, OR it is a 1980s Xerox photocopier with a malfunctioning duplexer.

@Plumbert

Ok and here's the interesting question.

If a company lets monkeys memorize the whole Internet and sells their service,

and then a person gives them a pen and asks them to write down as much of Harry Potter that they can remember,

who broke the copyright? The company that fed the Monkeys, the Monkeys, who did as they were asked, or the person?

@arstechnica

@fxnn
beware of the false dichtomy: the answer may end up being "nobody broke the copyright". See https://ansuz.sooke.bc.ca/entry/23 for nice explanation why identical copies of data might or might not violate the copyright...
@Plumbert @arstechnica
What Colour are your bits? - Ansuz - mskala's home page

As of Summer 2024, my article "What Colour are your bits?" has been online 20 years, and people are still linking to it as a benchmark. It's clear that people still care about intellectual property in general and copyright in particular, and the difference, if any, between id...

Ansuz - mskala's home page
@arstechnica So I'm pretty sure that Harry Potter is coverwd by UK copyright law.. which doesn't have a fair use defence... and so the Berne Convention should come in to play here as the US did sign up to that.
@arstechnica nice article and interesting for explaining how this generative model work..