Art for AI's sake

In which I continue to experiment with AI so you don't have to...

I read recently about a website called Night Cafe that enables users to generate all manner of pictures, in all manner of styles, without ever picking up a pencil or paintbrush. I had to have a play, of course, in my continuing but most likely futile attempt to be ready for the eventual takeover of our robot overlords.

It's pretty good, actually. You provide it with a text prompt of what you want to draw (the irony here being that you don't actually do any drawing), select a rendering model and processing algorithm, and click 'Create' (the irony here being that ... oh, you get the idea by now).

I thought I would try to create something that would never be seen otherwise, so entered a prompt of "Morrissey eating a hambuger". Here is the result:

There are, of course, a couple of things to note here. I cannot draw this well, for starters. But it's not perfect. For a start, it hasn't followed the brief: SPM is holding the burger here, not eating it. And finally, from this and other experimentations I can tell you that Night Cafe, at least, struggles with hands, their angle, their positioning, even (sometimes) the number of fingers and having two right hands and no left. Look at SPM's left hand here, for example. So it struggles with complexity ... for now, at least. I might try this again in a year's time...