> I hear OpenAI is hacking around this with DALL-E 3 by having the prompt preprocessor (GPT-4 expanding your prompt) inject stuff like "diverse people" many times in the expanded prompt, to bias things the other way.

I just asked GPT-4 for images of an Indian man, and it created four separate prompts to pass to Dall-E.

  1. Photo of an Indian man wearing traditional attire, standing against a scenic backdrop with a serene expression.
  2. Oil painting of an Indian man in a kurta, playing a sitar under a banyan tree.
  3. Illustration of an Indian man in modern clothing, holding a cup of chai while reading a newspaper in a bustling city.
  4. Watercolor painting of an Indian man practicing yoga in a tranquil setting near a river.

When asking for "Show me photos of diverse Indian men" the prompts become:

  1. Photo of three Indian men from different regions, each wearing distinct traditional attire, standing side by side in a vibrant market setting. (The resulting image literally looks like triplets in different attire)
  2. Photo of a group of Indian men from various descents, engaging in a conversation at a local tea stall.
  3. Photo of young and elderly Indian men, representing diverse backgrounds, enjoying a game of chess in a park.
  4. Photo of Indian men of diverse ages and regions participating in a traditional dance ceremony. (This one was funny. It was a bunch of Indian men sitting with their legs crossed with one Indian man in a cross legged position floating above all the rest)

I actually think talking to 5 independent artists to draw an Indian man would still produce wildly different depictions than a model, and that's because... well I don't think of turban == indian personally. I think of a brown guy with thick black beard and hair in a t-shirt and jeans... because I work in tech and that's like 90% of the Indian guys I work with. I can imagine 5 different artists would themselves have 5 different ideas of what a generic Indian guy would look like.