It becomes more and more apparent that AI creates serious problem(s) with copyrights. I believe the main issue is anachronistic law that does not specify how we should treat it. This lack of clarification creates two problems: whether training AI on copyrighted stuff is legal and whether AI's work could be copyright infringing or is it, as it is an AI, always transformative enough.

(2/X)

After thinking about problems I have few ideas/solutions I would like to share. Let's start with training. This one is simpler. AI companies should not be/have been allowed to use copyrighted works. I honestly think that scraping and then feeding AI works without permission/license from creator is theft and should not be qualified under fair use or any similar law. They must be punished (reasonably; I am against destroying them with fines or other penalties).

(3/X)

The second one is a bit more complicated. On the one hand I think of AI as a tool for creating images (as I am writing about Midjourney) and people who create copyrighted illustrations should be punished. They came up with an idea, they instructed AI how to achieve the result they wanted and finally they accepted the final product. This clearly shows that the originator ought to be held responsible.

(4/X)

To illustrate my point of view: You would not penalise Adobe for creating Photoshop, which is used to commit crimes (copyright infringement, etc.), would you?
On the other Midjourney is not quite like Photoshop, people do not control every step of creation, they do not "create" images, they do not control how it turns out (they might have some control) - AI does it for them and therefore should be punished.

(5/X)