These image generating AIs don’t need to have been trained on what you want it to output. If you tell it to generate a banana car, it doesn’t need to have seen a real banana car before, it just knows what a banana and a car look like, and combines them. Similarly, such AIs would know what naked humans look like and what kids look like.
These image generating AIs don’t need to have been trained on what you want it to output. If you tell it to generate a banana car, it doesn’t need to have seen a real banana car before, it just knows what a banana and a car look like, and combines them. Similarly, such AIs would know what naked humans look like and what kids look like.
I see, thanks for the explaination