Not sure if you are asking a serious question or not, but in the image generation process, local data is weighted way more highly than distant data, so in effect patterns that repeat a set number of times are its Kryptonite, it has a loose understand of when to stop repeating the pattern. Fingers are basically a repeating pattern. (Yes I understand that I’m incorrectly using language that anthropomorphizes a machine learning algorithm.)
Is AI hand generation trained off pictures of thalidomide poisoning?
Not sure if you are asking a serious question or not, but in the image generation process, local data is weighted way more highly than distant data, so in effect patterns that repeat a set number of times are its Kryptonite, it has a loose understand of when to stop repeating the pattern. Fingers are basically a repeating pattern. (Yes I understand that I’m incorrectly using language that anthropomorphizes a machine learning algorithm.)