Over on EQD, Sethisto is having a fit over his inability to distinguish DALL-E 3 images from human artwork (
https://www.equestriadaily.com/2023/10/the-huge-struggle-of-pony-posting-in.html). I find DALL-E 2 totally incompetent. See the attached responses to a request for a picture of "Lemuel Gulliver and Princess Luna from My Little Pony drinking milkshakes in a cafe." I don't have access to DALL-E 3, but the contrasting v.2 vs. v.3 specimens on the OpenAI website don't show an overwhelming improvement in the latest version — at least, not to my eyes.
Why is DALL-E such a virtuoso for Sethisto and such a klutz for me? Can anyone out there sort this out? What does DALL-E 3 make of Gulliver and Luna?
This post was edited by its author on .