re: George RR Martin suing ChatGPT/OpenAI after fan uses it to finish the last 2 novels

I’ve actually given presentations on this issue. There are some similar suits filed last year and this year alleging the same kind of fact patterns. No precedent so it’ll be interesting on how the courts come down but this isn’t cut and dry if you think Martin has a slam dunk case.

The legal climate that evolves over the next year or two will have a huge impact on the future of intellectual property as a whole.

There aren’t really any easy analogies for how generative AI works, so it’s hard to make direct comparisons to other technological advances. The AI (whether it’s an LLM like GPT or a diffusion image model like MidJourney) “learns” styles in a way that’s not all that dissimilar from how people learn styles - by consuming media.

If your art is available on the internet, you wouldn’t sue someone from viewing that art and learning your style. But it’s a little different when the AI can learn it in a matter of hours, and then be sold to people all over the world as a service.

The GRRM thing has the added twist of someone using ChatGPT to actually create a derivative work (sequel), which is copyright infringement in and of itself. But that’s a fairly easy thing to work out. The real battle is over the use of training data.