[arendtio]

Maybe the problem is not the artificial intelligence, but the thing we call copyright, and AI training is just showing us how weird the concept of copyright is.

I mean, it seems to be okay to replicate information if it has been remixed with other information enough, but not okay if the remix was too little. But then again, there does not seem to be a clearly executable definition of where this line is.

There are good reasons why we have copyright (like protecting/fostering artists/education/entertainment in our modern society), but the concept itself is a bit weird and artificial after all.