This is example why neural network generators aka "AI" can't draw good rope bondage. "Ropes" look like clinging wet noodles, zero knots is present or zero effect on figure is shown. That's bad even if I igonore callistic side of the picture.

There are layers upon layer of complexity in situation and implementation which even half of humans cannot comprehend, but may understand intuitively.
And even if director would understand that and somehow will explain every detail and knot to generator, most likely generator wouldn't "understand" it, because it wouldn't be able to find analogs among stuff it was educated on. And most of bondage stuff, that is not photography, is wrong too.