AI isn't the be all end all of knowledge and insight; in particular, it can exhibit strong biases, depending on its weighting and its training data. For example, all references to the Tiananmen Square massacre were missing in the data on which Chinese chatbot DeepSeek was trained, hence, it couldn't answer prompts about it even without any explicit censorship toggle in the code.
Same story with a lack of diversity in AI-generated images of founding fathers or astronauts, which is based on the corresponding lack of diversity in the historical data. (The much maligned "black founding fathers/black nazis" which Google's Gemini produced for some time were the result of a hamfisted way of forcing diversity in the results of such prompts... by woke Google engineers manually messing with the weights.)
Regarding the face of Jesus: if an AI face reconstructor program is predominantly trained with pictures of white faces, it will have a tendency to reproduce white facial features. So unless we know the specifics of the program used, this really isn't definitive proof of anything.
This post was edited by Black XistenZ on Feb 23 2025 04:15pm