L4sBotJul 9, 2023

2 authors say OpenAI 'ingested' their books to train ChatGPT. Now they're suing, and a 'wave' of similar court cases may follow.

https://lemmy.world/post/1246165

2 authors say OpenAI 'ingested' their books to train ChatGPT. Now they're suing, and a 'wave' of similar court cases may follow. - Lemmy.world

Two authors sued OpenAI, accusing the company of violating copyright law. They say OpenAI used their work to train ChatGPT without their consent.

Show threadMeow.tar.gzJul 9, 2023
Too be honest, I hope they win. While I my passion is technology, I am not a fan of artificial intelligence at all! Decision-making is best left up to the human being. I can see where AI has its place like in gaming or some other things but to mainstream it and use it to decide who’s resume is going to be viewed and/or who will be hired; hell no.
Show threadUlu-Mulu-no-die
I’m not against artificial intelligence, it could be a very valuable tool, but that’s nowhere near a valid reason to break laws as OpenAI has done, that’s why I too hope authors win.
Jul 9, 2023 at 1:28pmWeb
Show threadbioemerlJul 9, 2023
What laws are you saying they've broken?
Show threadUlu-Mulu-no-dieJul 9, 2023
Copyright, this is not the first time they’re sued for it apparently.
Show threadbioemerlJul 9, 2023
Scraping the web is legal and training AI on data is also legal.
Show threadUlu-Mulu-no-dieJul 9, 2023
Reusing the content you scraped, if copyright protected, is not.
Show threadGnothiJul 9, 2023
Sounds like fair use to me.
Show threadLegendofDragoonJul 9, 2023

That really will be the question at hand. Is the ai producing work that could be considered transformative, educational, or parody? The answer is of course yes, it is capable of doing all three of those things, but it's also capable of being coaxed into reproducing things exactly.

I don't know if current copyright laws are capable of dealing with the ai Renaissance.

Show threadbioemerlJul 9, 2023

Yeah it is. The only protection in copyright is called derivative works, and an AI is not a derivative of a book, No more than your brain is after you've read one.

The only exception would be if you manage to overtrain and encode the contents of the book inside of the model file. That's not what happened here because I'll chat GPT output was a summary.

The only valid claim here is the fact that the books were not supposed to be on the public internet and it's likely that the way open AI the books in the first place was through some piracy website through scraping the web.

At that point you just have to hold them liable for that act of piracy, not the fact that the model release was an act of copyright violation.