Heads, Shoulders, Knees and Toes is too hard for AI

Arse

头、肩膀、腿和蹄，腿和蹄！

头、肩膀、腿和蹄，腿和蹄！

还有眼睛、耳朵、嘴巴、鼻，

头、肩膀、腿和蹄，腿和蹄！

ChatGPT will not draw an image of a child touching its toes. In AI land, children touch their toes by balancing on one leg and performing some kind of weird pistol squat. Why? No amount of correcting or prompting will make the computer show me an image of a normal child bending or crouching down to touch both of his or her feet.

How lame ChatGPT is, right? Until you go and try one of the other free image generators out their and get this cursed horror…

I suppose I could try Midjourney or one of the others, but are they going to give me anything better?

Microsoft’s online image generator, for example, believes that we keep our toes above our heads.