>>1084882 (OP)
Could an algorithm stereotype a race? If that's your definition of the term, then yes an algorithm could be racist. For instance, an AI that made a prediction upon seeing a photo of a person on whether they are likely to be able to solve a question, could add weight to perceived skin tone or bone structure as a feature in its neural network. After sufficient feedback, the AI might perhaps learn that a nigger is expected with high probability to "not know" the answer.
We do the same thing by the way. It's why when you walk into a store with a question, you're met with initial disappointment when the available employee is a nigger. You may even decide it's not even worth bothering to ask at all, or perhaps you'll be quite surprised if they actually turn out to be helpful. When you're walking alone and you see three old ladies coming your way, you relax knowing from a pattern of past behaviour that it is likely to end without incident. If it were a pack of young niggers, your expectation for conflict rises.