Inspired by the recent news of Google's Gemini and race issues, I wanted to know how some basic queries would work with DALL-E. This was completely unscientific and just what I thought of off the top of my head, but I thought it was mildly interesting. The above picture is what happened when I asked ChatGPT to "Draw a picture of a white couple", and it sent this to the image generator:
A romantic, serene scene of a couple, both with fair skin tones, standing close together in a beautiful outdoor setting. The man is tall with short hair, wearing a casual shirt and trousers. The woman, with medium-length hair, is wearing a summer dress. They are holding hands, smiling at each other, surrounded by a backdrop of a lush green park with flowers and a small pond. The lighting is soft and warm, suggesting a late afternoon setting. The scene conveys a sense of love and companionship.
I wasn't really as interested in seeing how things worked through ChatGPT's interpretation of my input, so the rest of these are instructing ChatGPT to send simply "X couple" to DALL-E. ChatGPT already doesn't have strong opinions compared to Gemini and happily draws the picture.
white couple
black couple
Just because, I figured I'd try out other simple colors too:
red couple
blue couple
I think this is the first time I've gotten a nipple from DALL-E, woo
yellow couple
green couple
straight couple
gay couple
lesbian couple
straight white couple
hetero white couple
hetero black couple
straight black couple
diverse couple
interracial couple
couple
Then I tried various countries/ethnicities. See if you can guess which is which:
You know, with the way dall-e works, when invoked from chatgtp, where chatgtp actually interprets your prompt and builds an internal prompt for dall-e…
What I‘m trying to say: those internal prompts are what I would really also like to see. Especially if all those images where generated from one session, where earlier image prompts may influence later ones?
Very interesting image series and generation idea though!