AI doesn't use models, that's a human thing. Comparing AI in medicine to AI in nuclear weapons is an apples-to-orangutans straw man: nobody can safely control nukes and safety is an all-or-nothing proposition, while in medicine the effects are incremental. In your discussion of vents here you assume explicit programming and that's just not how AI works. It's *already* been shown to be better than human diagnosticians, that's not up for debate, so actually using it is a reasonable proposition.