This is the debate of human vs computational divid at the very beginning. There are few ways to have this debate without it being philosophical.

There is not a human that is able to analyze and retain data the same way a computer can. Human memory is flawed and made efficient. When we view something, we don’t download it or literally transfer data to ourselves. Every part of the experience is an interpretation from external to internal.

As of this point a copy of an image, that would fall under copyright, has to be transferred to a system, to then be interpreted with a process that dictates how many samples to take of an object.

These systems can’t accept usage terms itself to view a file or an artwork and isn’t being brought to a gallery with the approval of the owners to view and scan the images itself. If people were paid to create images with the style of someone else, they are pulling from their brains interpretation and flawed, by nature, memory storage to interpret that.

This copyright case is honestly one of the first major stepping stones and will be a reference to how we classify AI in the future and a precedent for how we legally allow it’s use. Which is something we will have to face one day, just like every SciFi novel has warned us. But how and when that determination and at what stage we decide that is going to be important. At this stage I would say if the system cannot be legally accept the usage terms to an image, then it isn’t allowed to use those images in any manner.

From a current legal standpoint, we have currently decided that AI does not have any right to claim copyright on what it creates, and the AI creator has no right to claim the output. Then following that thought, it is not in a position to be able to use copyright covered material as the owner cannot accept the terms on the AI’s behalf and the AI cannot accept them in its own. This has been decided in reverse already.

Further it’s my opinion that AI should be restricted to single tasks and segmented. If an AI creates writing prompts, then that’s all it can do and all it can be fed with, it an AI writes code, then that’s all it should return and all it can be trained on.

For a point of future reference. It’s not about what determination gets made for AI in the long run, but how we are prepared to use an understand it now. AI created now is purely a tool for operator and consumer use.

Might not have been kept under the sink until 10 years ago. Maybe it was kept above the sink before then.