The total disruption of medicine is in the air.

Microsoft AI researchers recently published a paper showing that recent AI models outperform doctors in generalized diagnostics - not only in a specific area but overall on a difficult 304-case benchmark.

See this:

"The farther up-and-left you are, the better. Humans sit low-and-middle; off-the-shelf chatbots are scattered; the orchestrated AI (blue) carves out a whole new skyline—often matching or beating top models’ accuracy while paying less for tests, and utterly eclipsing human doctors." (source)

While multi-step diagnostic accuracy of doctors was 20% on average, the best AI agents reached 80% accuracy with the same access to information - at a reduced total cost per case.

Costs cover physician-visit equivalents plus lab and imaging CPT fees. Model inference (~$0.15) is ignored as immaterial.