Mate, nobody's saying that every single black person or woman has it worse in every way than every single white person. We could play the 'Ah, but this person has this malady that's worse than the last one mentioned!' game for hours and it really means nothing. It's got nothing to do with my understanding of statistics, it's got to do with the fact that I'm trying to make a generalized statement to keep the flow of conversation going, which we kind of have to accept or else having any kind of short-form discourse is impossible.

Here, let me try again.

"Generally speaking, doctors tend towards having implicit biases against black people and women. On average, medical treatment given to white men is more reliable than what's given on average to those two marginalized groups."

Is that really adding any meaningful clarity to a casual discussion? I argue that it does not.

To return to the topic at hand, I don't think, in the vast majority of medical cases, diagnosis is something AI couldn't handle. It's not like every doctor is Dr. House, solving a complicated puzzle to figure out what someone has... it's more or less a flow chart. Patient complains of X, most common causes are A, B and C. Order a test to confirm. Test comes back negative, next most common causes are Y and Z, order a test to confirm.

If AI could solve the majority of cases like this - even if there was a doctor whose whole job it was to take the AI diagnosis, review the symptoms and test results, and say "Yes, this seems correct", it would presumably leave more time for doctors to spend on cases like yours where more attention is perhaps warranted. Or, alternately, maybe AI would have solved it more quickly by not being beholden to the specialist blindspots and institutional inertia that prevented them from correctly identifying it.