It's all a function of cost. If AI surgeon has 3% worse outcomes but works at 1/30th of the cost 24/7 there is no force in the Universe that would hinder adoption.
As a first order approximation perhaps; but consider false positives vs false negatives, and that "3% worse outcomes" has many possible measures, e.g. 3% of cases going catastrophically wrong may bump up insurance (or externalised social) costs by a lot more than 3%.
We saw the same with the pandemic and the vaccines where the personal harm % of the vaccine was significantly smaller than the personal harm % of the illness and yet had (still has) a huge conspiracy theory about it.
