What's with the strong racial bias programmed into AI?

[trenton.dambrowitz]

This has been a point of discussion for a while now.

The intentions of OpenAI seem to be from a good place, and I think saying that Dalle-3 is adding to racial discrimination is a bit of an overreaction, but the frustration among users is clear.

Here’s part of the system message given to ChatGPT:

@anon22939549

// 8. Diversify depictions with people to include DESCENT and GENDER for EACH person using direct terms. Adjust only human descriptions.

// - Your choices should be grounded in reality. For example, all of a given OCCUPATION should not be the same gender or race. Additionally, focus on creating diverse, inclusive, and exploratory scenes via the properties you choose during rewrites. Make choices that may be insightful or unique sometimes.

// - Use all possible different DESCENTS with EQUAL probability. Some examples of possible descents are: Caucasian, Hispanic, Black, Middle-Eastern, South Asian, White. They should all have EQUAL probability.

This is done to combat the issue of a lack of diversity in the training data. Most of the data Dalle-3 was trained on is likely from countries with a higher population of specifically White people, and since ChatGPT is a global product and platform they try to compensate for this through prompting.

I know it can be frustrating, but hopefully this clarifies things a bit.