How do you reconcile that you interact with AI when you may not even realise? This could be in the form of a recommendation feed, a newsreader delivering a story on the 6 o'clock news or using software that was built by it?

There are two different meta-uses of AI that we're talking about here.

   #1 AI as final step

   #2 AI as upstream step, with subsequent human review/adjustment

In the case of medical AI, we're almost always talking about the latter: models preclassify or review results, which are then subsequently reviewed by a human.

The cases people seem to have problems with are where no human eyes/hands provide oversight to probabilistically-correct model output*.

It's an important enough distinction that I'd be in favor of mandating companies declare which they're using (for a specific use case).

* As distinct from previous deterministically-correct expert systems / rules-engines

I find that wild, but to each their own!

i find it wild that you would trust technology that has no problem with giving you wildly wrong information!

We're not talking about a doctor logging onto ChatGPT and asking it a question then blindly following what ChatGPT says.

But we _are_ talking about developers working on critical path software, like medical devices, who would do exactly this.

(There have also been some reports of doctors making diagnosis decisions based on ChatGPT results.)