10 things about recent AI art responses that blew me away 2024
Yesterday I made a list of silly superheroes and asked ChatGPT 4 to “draw a funny picture of a superhero named (x)”. It created some funny drawings that made a LOT of accurate intuition:
1. The art was all in the same style
Not necessarily from the same story, but I did not request that. If you had been told that a human had done this, you probably wouldn’t question that it was the same person doing it. I have found this stage that I can usually guess whether an image was created by mid journey for example.
2. Anatomy was pretty accurate
they were cartoons but almost everyone had the correct number of limbs and fingers.
3. Lots of associated items were included
For example when I asked for “the Obesity Squad
the AI got that there was a group, they all had the same outfits on, their belts did not fit. The characters looked like cartoon fat people that were fat in the expected proportions: not just an obese arm. These were cakes and spoons and forks as weapons.
4. The drawings were funny
ChatGPT four seems to somehow understand silliness and they crack me up.
5. 40 year old that lives with their parents
Nailed so much about this: so many funny details that make this character pathetic: how did it associate all of these things together? This image will be its own list
6. These were all first take results from very minimal prompts
The panhandling super villain, was placed outside on a mat with a cardboard sign that said, instead of will work for food will work for world domination. I did not prompt any of that.
7. Waffle woman
I was running out of ideas, and I just gave this one to see what would happen. It made an accurate waffle, made her outfit waffle-ish, but really like a superhero outfit and gave her a spatula to use And syrup.
8. AI can make images 24/7 with unlimited revisions
9. As I keep creating these images more usage ideas may come up
It is good to have reason to practice working with AI. I avoided using Siri for years but realized that I should get used to it.
10. Large language models have a very deep “understanding “ of many things
. You just have to learn how to ask the right questions to get the answers.