Comment by atlantic

If diseases manifest differently for different races and genders, the obvious solution is to train multiple LLMs, based on separate datasets for those different groups. Not to mutter darkly about bias and discrimination.