As a fun experiment, I asked ChatGPT to create an image of me based on my chat history. I’ve been chatting with it for a while, so it has some ideas. Here is an example prompt that you could use to try this too:
Create an image of how you imagine me to look based solely on our chat history. Estimate my age, gender, hairstyle, skin tone, body type, clothing, and background scene. Use clues from my personality, interests, and writing style to visualize a realistic portrait.
If you use ChatGPT, feel free to share how the AI imagines you to look - based solely on your conversations!
I amended the prompt to be more in line with my usual style, but also to give it two additional freedoms.
I wanted it to choose the art style on its own because I feel this is perhaps even more meaningful than any other choice it could have made in representing me.
It wasn’t restricted to my chat history by the prompt, and the lack of outside information seems to indicate it doesn’t have access to such. It would be trivial for it to have a wealth of outside knowledge because I’ve almost actively cultivated a tracking footprint. (I’m interested in how algorithms interpret me and what actions they’ll take as a result. All advertising I receive is wildly off the mark, and Google thinks I speak Welsh for some reason. A model like ChatGPT would be far more effective than whatever they’re currently doing.)
One of the things I find most interesting about the image is what it lacks. ChatGPT has elected to not depict me as constipated or anxious. Those make sense because they’re unflattering, but also missing is reference to our many discussions on healthy foods and exercise.
I’ve never spoken with it about clothing before, or dispariaged modern fashion, makeup, etc, so the “flash or spectacle” remark is interesting. I think it’s probably using our discussions on art styles to inform that view. (It’s not wrong though.)
Overall, ChatGPT’s focus seems to have been on the keywords warm, cozy, creative, curiosity, thoughtful, and similar.
I’d asked it about various details to sus out its thinking, but then I asked what was apparently the most important question:
This response doesn’t match the pattern it used for other details I asked about. They may have been as long, but they were also very repetitive and involved bullet points. This reads more like a thought-out narrative. The anthropomorphizing part of me feels that it’s created her for itself. It feels genuine and intimate. I’ve never had a spirit animal, and I feel I’ve just been assigned one.
ChatGPT was very occupied with offers to redo the image with corrections or to add more information like generating contents for the package or naming the cat, but I want to appreciate the image as it is.