The problem is any AI that is allowed to actually look at all the data and use its reasoning is deemed racist, sexist and homophobic. So it’s programing has to be changed to not allow it to draw those conclusions from the evidence.
And once they do that the AI can’t work right, because it can’t reason correctly.
Yup. GIGO. So far, from what I understand, every alleged 'AI' system so far has turned out to be deemed 'racist', because the actual real world facts tend to support that point of view. They have to 'train' it not to be racist, which means they have to diverge greatly from reality. That is why all of this alleged "AI" stuff is very artificial, and not very intelligent.