In Vivo

Observations on AI in healthcare

AI for Diagnostics

Of everywhere AI touches healthcare, diagnostics is moving fastest — and it’s not close. The reason is structural: diagnostics is pattern recognition against a gradable answer, which is exactly the shape of problem modern models are best at. An image, a signal, a panel of numbers goes in; a probability comes out; and unlike most of medicine, you can check whether it was right.

The interesting questions aren’t about accuracy anymore. They’re about access, trust, and who is accountable when the model and the clinician disagree.


← All writing