@UncivilServant this has nothing to do with biases; llms don't produce correct answers, they produce statistically-probable text completion. @atoponce