70
Why the Training of Generative AI Models Is a Violation of Copyright Law
(www.futuristiclawyer.com)
this post was submitted on 06 Oct 2024
70 points (85.7% liked)
Futurology
1752 readers
108 users here now
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Thought experiment: What if it an extremely intelligent person memorised every Metallica song on a free Spotify account, down to every small detail. Later that person writes a new song, it's heavily inspired by Metallica, it sounds like them, you might even mistake it for them except for the vocals, but the lyrics are new, the chords are new, etc. Did that person then violate copyright, even though it's a completely new song?
I know the AI techbros are just scraping every datapoint they can get their grubby little hands on, but it makes me think.
Imagine for a sec that all that AI buzz and hype leads to something that is indistinguishable from that extremely intelligent person (however unlikely).
We're not anywhere close to a scenario like that, but at what point is a regular artist that spent their youth listening to Britney Spears violating copyright law, philosophically speaking, when they decide they want to make music like that?
Penny for your thoughts.
I feel like the important difference between an intelligent human and an AI is that the AI can be replicated easily whereas we are yet to be able to clone a human with their memories. Imagine if we added a restriction to AI training in that only 1 copy of the model can ever be run at a time, like a human. In that situation, all the trouble with copyright would go away.