And what it isn't.
Another interview I did recently: doctors have started believing AI diagnoses and not questioning them because they're rushing around and not taking the time to verify the information that's being surfaced. Meanwhile insurance wants to opt out of coverage for incidents involving AI because they KNOW people are going to be harmed by sloppy use of it. Is it good at dredging up the 50 articles on using X medication off label so you can review those articles and make an informed decision? Honestly, it really is. Can it diagnose the patient in your chair correctly? Absolutely fucking not. But the problem is HUMANS not knowing the difference.
(The expert in this article has started asking his own doctors to turn off ambient listening because it's so bad at getting the information correct - cases where ambient listening diagnosed a patient with diabetes because the doctor asked the patient about their MOM's diabetes, which then impacted the patient's insurance rate later on... we're not ready for this technology, and it's doing so much harm by not realizing we're not ready for it.)