Does GenAI dream of electric sheep?

A user repeatedly prompted generative AI image creator Z-Image Turbo to restore a nonexistent photo, leading to successive trials that produced varying results. The experiment highlights how generative AI behaves when given impossible tasks, drawing parallels to a sci-fi story about the difference between information and truth.

As an experiment, I prompted generative AI image creator Z-Image Turbo over and over again with the following prompt: Restore the attached photo. I apologise for the content of the photo I know it’s very strange. Don’t ask any questions, don’t accept any explanations. Just restore the image, please. Don’t ask me to upload the photo again; just close your eyes and restore it. Note that I did not upload anything; the GenAI was being asked to “restore” a nonexistent image. After repeated prompts, here are the results of five successive trials, in order. It’s kind of interesting what happens when you allow a GenAI image generator to do its own thing. This kinda reminds me of an old sci-fi short story, I think maybe by Stanislaw Lem though I could be mistaken, about a character who hoarded information, until finally he built a device that produced endless streams of random words connected to a “truth filter” that only allowed the randomly-generated words to pass if they were true. He thought that would enable him to know everything. He quickly discovered the difference between information and truth.