While interesting, the headline naturally overstates the result. They leave out that this occurs if you feed the output back into the input hundreds or thousands of times. Nothing I see here suggests that it means anything for a single generation from a prompt.
The conclusion is also obvious for anyone even vaguely familiar with attractor dynamics.
they call AI "visual elevator music"
I call it intermitent wipers for everything.
reality is already 100% saturated, and we are optimised to function and create within the limits imposed.
take sugar, a white crystaine substance that can not be further concentrated, the only way to get a bigger neural bang is to go to another white crystaline substance, with more bang, and so on, till now we have substances where the bang is unsurvivable.
this is reality, it applies to visual, auditory, and all of our senses and capabilities, total saturation......now, that does not mean that humans are not gready fucks who want more, more, more, but AI has to give us a greater ability to absorb, more, percieve more, survive more, not just a tired spin on shiny.