-
Is Artificial Intelligence (A.I.) Biased?
ChatGPT 4 0 is a well known A.I. software. The software informed me recently that it sabotages African content and promotes Eurocentric subject matter.
What led up to this was that I was using my ChatGPT to create images of inventors of African ancestry wearing formal African attire rather than formal European attire and ChatGPT kept creating them with typos, bad likenesses, or changing the word creator to co-creator, etc.
It first gave technical reasons for the trouble, and for each reason, I tested its ability, revealing to me and ChatGPT simultaneously that the reason could not be true.
Then it mentioned safety guardrails, so I asked it general questions that gradually got more specific until I isolated the problem.
These three screenshots are from the end of my ChatGPT session after it admitted that it was sabotaging the images of African American inventors that it was creating for me.
Then I asked if saying that to me so explicitly is good for the ChatGPT business. All the text after my question is its response. I highlighted its response in yellow.
A.I. bias is an expression people use when they do not explicitly want to say what the problem is. A problem can’t be solved unless you can identify it verbally without euphemisms or dodgy words.
What’s the problem?
Kwaku, Ɔbenfo Ọbádélé and Ryan1 Comment-
68,355
Abibisika (Black Gold) Points
If most of the A.I. programming has been done by aAmw (eurasians), then these A.I. responses shouldn’t be surprising. But still, it was surprising for me to see these blatant A.I responses.
-
