The courts need to settle this: Do we treat AI models like a Xerox copier or an artist?
If it’s a copier then it’s the user that’s responsible when it generates copyright-infringing content. Because they specifically requested it (via the prompt).
If it’s an artist then we can hold the company accountable for copyright infringement. However, that would result in a whole shitton of downstream consequences that I don’t think Hollywood would be too happy about.
Imagine a machine that can make anything… Like the TARDIS or Star Trek replicators. If someone walks up to the machine and says, “make me an Iron Man doll” would the machine be responsible for that copyright violation? How would it even know if it was violating someone’s copyright? You’d need a database of all copyrighted works that exist in order to perform such checks. It’s impossible.
Even if you want OpenAI, Google, and other AI companies to pay for copyrighted works there needs to be some mechanism for them to check if something is copyrighted. In order to do that you’d need to keep a copy of everything that exists (since everything is copyrighted by default).
Even if you train an AI model with 100% ethical sources and paid-for content it’s still very easy to force the model to output something that violates someone’s copyright. The end user can do it. It’s not even very difficult!
We already had all these arguments in the 90s and early 2000s back when every sane person was fighting the music industry and Hollywood. They were trying to shut down literally all file sharing that exists (even personal file shares) and search engines with the same argument. If they succeeded it would’ve broken the entire Internet and we’d be back to using things like AOL.
Let’s not go back there just because you don’t like AI.
riskable@programming.dev 3 days ago
The courts need to settle this: Do we treat AI models like a Xerox copier or an artist?
If it’s a copier then it’s the user that’s responsible when it generates copyright-infringing content. Because they specifically requested it (via the prompt).
If it’s an artist then we can hold the company accountable for copyright infringement. However, that would result in a whole shitton of downstream consequences that I don’t think Hollywood would be too happy about.
Imagine a machine that can make anything… Like the TARDIS or Star Trek replicators. If someone walks up to the machine and says, “make me an Iron Man doll” would the machine be responsible for that copyright violation? How would it even know if it was violating someone’s copyright? You’d need a database of all copyrighted works that exist in order to perform such checks. It’s impossible.
Even if you want OpenAI, Google, and other AI companies to pay for copyrighted works there needs to be some mechanism for them to check if something is copyrighted. In order to do that you’d need to keep a copy of everything that exists (since everything is copyrighted by default).
Even if you train an AI model with 100% ethical sources and paid-for content it’s still very easy to force the model to output something that violates someone’s copyright. The end user can do it. It’s not even very difficult!
We already had all these arguments in the 90s and early 2000s back when every sane person was fighting the music industry and Hollywood. They were trying to shut down literally all file sharing that exists (even personal file shares) and search engines with the same argument. If they succeeded it would’ve broken the entire Internet and we’d be back to using things like AOL.
Let’s not go back there just because you don’t like AI.