I love the first one in the sequence. Why would it choose hairy toast? It reminds me of a certain toasted-looking politician who despises windy days.
You know, I have a lot of observations about AI's quirks. It reminds me. . . . and this might be part of what you find unsettling about AI. It reminds of writing poetry (and sometimes visual art, in different ways), where the act of making things fit rhyme or meter causes me to say things that weren't originally intended, and then I have to check and see if I really want to say that. Sometimes it leads to amazing discoveries, serendipitous wisdom, and sometimes I have to cut out the nonsense. The AI throws out this random "poetic" element, and sometimes it's nonsense and other times it's not.
I'm probably not explaining it very well. Let me show you an example:
When I was generating my image for "Chef Cat Makes A Shrimp Cocktail" (OpenAI's DALLE-2) - the fourth image is the one I used (and the only usable result here) - but look at how shrimp images, shapes, and colors crept into the body of the cat in each image. Only #4 was "fixable." I see that a LOT in my generated images. Unintentional echoes.
Another:
The puzzle pieces were supposed to be in the blender but see how they crept out into the counter and the backsplash, etc. Don't get me started on the macarons - I have NO IDEA how those got there (reported to OpenAI as unwanted result).
There are, of course, other AI engines, and I haven't examined them all closely, but I have seen what I consider to be flaws/defects in all the ones I have looked at closely. Some of them apply heavy style changes to cover for defects in the rendering, for example.
Anyway, what I meant to say is that some of these issues are actually possible opportunities for serendipitous discovery ("happy accidents"), but what I seem to be saying now is that AI has a long way to go. Both are true
Edited to add: I lost my point. If I find it, I'll post again. I had a point that I failed to make, just can't remember what it was exactly.