Re:Things are about to get a lot worse for Generative AI
by DonsDiscountGas on Saturday December 30, 2023 @08:51AM
(#38815260)
The examples given are all billion-dollar, decades old characters. The volume of material directly/indirectly referencing those characters in a random internet crawl will be fairly large. Most copyrighted works won't have that issue. If anything it means they only infringe on archetypal works and not the other 99.9%. If I write a story involving robots and spaceships (of which there are many, before and since Star Wars) DALL-E won't infringe me because it will be busy infringing on Star Wars.
[ Parent
| on HN ]
Re:Things are about to get a lot worse for Generative AI
by jprete on Saturday December 30, 2023 @10:01AM
(#38815740)
I'm opposed to my (fairly minor) copyrighted works being used in GenAI datasets as well. I just have no practical way to stop it, and there aren't clear enough damages to sue. That doesn't make it legal.
Re:Things are about to get a lot worse for Generative AI
by eropple on Saturday December 30, 2023 @10:21AM
(#38815898)
OpenAI also plays some ugly games with regards to the difference between training and search. Search requests come from the `ChatGPT-User` user-agent, and I'd like to allow those; training and scraping requests come from `GPTBot`, and I have no interest in those. But as per their own documentation, putting one in robots.txt disables the other.
Re:Things are about to get a lot worse for Generative AI
by nullstyle on Saturday December 30, 2023 @11:29AM
(#38816464)
Thats a major asshole move and I’ll reference it when people ask about bad acts from OpenAI. Thanks!
Re:Things are about to get a lot worse for Generative AI
by edgyquant on Saturday December 30, 2023 @10:27AM
(#38815955)
This is entirely because these are the companies with the recognition and resources to push back against this.
Re:Things are about to get a lot worse for Generative AI
by _jal on Saturday December 30, 2023 @09:49AM
(#38815647)
The examples were chosen by the author to make a point precisely because they are well known.
But every single copyright holder with their works online (which includes you and me) has the same legal rights as the NYT or Disney. Naturally some copyright holders have more real-world capability to go legal than others, but that does not reduce the legal risk.
> If anything it means they only infringe on archetypal works and not the other 99.9%
How on earth do you get to that conclusion? There's no "popularity" floor to copyright protection. Either a work has been infringed or it hasn't.