Maybe it’s taking a picture of Godzilla’s face and shoulders, converting it to black and white (not grayscale), and then doing an ASCII representation of that. A dark Godzilla on a light background would look sort of like that after an unsophisticated conversion to black and white.
I wondered if it was starting with the top but then it’s length limits kicked in and it decided the response was getting too long, so it jumped to the conclusion.
Maybe it’s taking a picture of Godzilla’s face and shoulders, converting it to black and white (not grayscale), and then doing an ASCII representation of that. A dark Godzilla on a light background would look sort of like that after an unsophisticated conversion to black and white.
I doubt it’s doing any of that. It just predicts text
I’ve lost track of which versions just generate text and which ones generate pictures too.
GPT-4 can accept images, but not generate them. However it can go to Dall E 3 to generate and help you prompt it
I wondered if it was starting with the top but then it’s length limits kicked in and it decided the response was getting too long, so it jumped to the conclusion.