So it's a bit like a mini skirt in the 60's - how short the thing is allowed to be - depends on your set of morals
I suppose we all draw the line somewhere
Aweful?
I absolutely disagree. While AI is by n long shot not perfect, it can - and does- generate/produce stuff you would not be able to tell from the real thing in 7 months of Sundays and even then - AI 1, you 0
I am very pessimistic about our ability to distinguish things generated by AI from reality and so I agree in broad terms, but where the particular version of AI running behind Adobe's generative fill is concerned what I have seen thus far has been awful unless it was just filling in a background, in which case it has been various degrees of impressive.
For example, here are several I quickly had it generate which are all uniformly bad.
You can only see EXIF info for this image if you are logged in.
You can only see EXIF info for this image if you are logged in.
You can only see EXIF info for this image if you are logged in.
You can only see EXIF info for this image if you are logged in.
You can only see EXIF info for this image if you are logged in.
You can only see EXIF info for this image if you are logged in.
You can only see EXIF info for this image if you are logged in.
That last one, by the way, is supposed to be a toad. The ones which actually do look like the things they're supposed to be, at least sortof, all have a plasticky, very artificial look to them, and often you get something that goes to the extreme of looking more like a Toy Story era CGI creation than anything else. Often, like the last one, the thing it makes doesn't even look close to right. For instance, I can't post it because I'm at the limit for this post but I just asked it to put a flock of birds in front of a photo I have of the sky and a early-evening moon and it came up with a collection of eldritch horrors that look something like what a bird might look like in a hellish nightmare.
Now, ask it to extend a background or add some ground to part of your photo and it can be varying degrees of convincing. Like I said, it seems to do a lot better working on a photo at the forest level than at the individual tree level.