That third example is a grey area. The photographer has no control over the subject of the photo, which is arguable a large part of the creative part of the photo. I am not convinced that they would automatically get copyright in that situation.

And the third and forth touch on a lot of points that modern smartphones already do - automatically set up the camera, can adjust the frame of the photo, take several pictures and blend the best ones together, stabilise the shot, auto adjust exposure etc etc. When does taking a photo with your smart phone end up being like those examples?

But regardless, this copyright claim has some good arguments for why a prompt for an AI generated image should not result in the image being copyrightable - far better then the case the article in this post points to.

The core of the argument seems to be that you cannot predict the output of the AI, so the prompts are more suggestions and not strong enough to be considered creative enough for a claim to copyright. And that generating lots of images until one matches your idea is also not enough - as that is like doing an image search until you find something that matches what you want.

Which feeds back into that third example, if the photographer cannot predict what the people will do in the frame then given the above I would argue they have a weak claim to copyright over the image. No matter how much work they did to set up the shot.