Compensating the creators of the training data might sound like a good idea, but it breaks down in a few key areas: -Generative models can/will be created/modified/used locally by individuals and small groups, not just large companies, making enforcement difficult/impossible. -If you go so far as to ascribe authorship rights to generated works, it creates issues for licensing and copyright expiration. Additionally, the creator of the first instance of that element has likely been dead for over 70 years, which suggests the element should be in the public domain. -If the same element of a generated work is tied back to thousands (or more) of individuals, you can argue that element isn’t substantial enough to warrant copyright protection due to a lack of uniqueness.
![Chris Langan - The Interview THEY Didn't Want You To See - CTMU [Full Version; Timestamps]](http://i.ytimg.com/vi/9miVG2xT5jY/mqdefault.jpg)
Compensating the creators of the training data might sound like a good idea, but it breaks down in a few key areas:
-Generative models can/will be created/modified/used locally by individuals and small groups, not just large companies, making enforcement difficult/impossible.
-If you go so far as to ascribe authorship rights to generated works, it creates issues for licensing and copyright expiration. Additionally, the creator of the first instance of that element has likely been dead for over 70 years, which suggests the element should be in the public domain.
-If the same element of a generated work is tied back to thousands (or more) of individuals, you can argue that element isn’t substantial enough to warrant copyright protection due to a lack of uniqueness.