If you listen to the red hot chili peppers or watch a marvel movie or look at a DC comic and then go and make a song, movie, or painting inspired by the style of a certain creator that does not mean you have somehow violated those creators copyright. You don’t owe them any money because you took inspiration.
AI training on publicly available data does not infringe on copyright even if that data is somehow copyrighted.
And I know that many people on these kinds of platforms don’t like to hear this but the benefits of AI outweighs any potential legal issues copyright might entail.
Moreover, and I keep pointing this out over and over, you can’t have the same information free for individuals to use and have it paid for at the same time for corporations. You have to decide if you want that information free for all or for none.
Zoboomafoo@slrpnk.net 6 days ago
Who’s watching marvel movies for free, legally? Who’s listening to RHCP’s entire discography for free, legally?
mechoman444@lemmy.world 5 days ago
No one is. That’s exactly the point.
Llms aren’t recreating copyrighted works. They’re drawing inspiration if you will. No copyright is being infringed.
Zoboomafoo@slrpnk.net 4 days ago
And how is an LLM trained to “draw inspiration” from an author without reading their books?
mechoman444@lemmy.world 4 days ago
That’s exactly what it is. But it’s not replicating the book to sell that same book to generate profit the author of the book won’t get.
It’s using the information in the book to generate its own data.
Are you aware of how llms work?