Hard disagree. Any AI version of something is not the vision of anyone. Sure, the prompt engineer worked on it, and someone signed off on which output to use, but unless a lot of retouching is done, it will always be just noise.
If the answer to “why was the curtain blue?” is “the random seed in the image generator dictated it so”, that sucks.
Not sure. They don’t mention 4 being in beta or anything. Though they don’t list much advantage of using it over 3 either.