Microsoft Says AI Diagnoses Illnesses With 83% Accuracy, Beating Doctors by Up to 4 Times
Updated
Updated · The Verge · Oct 22
Microsoft Says AI Diagnoses Illnesses With 83% Accuracy, Beating Doctors by Up to 4 Times
3 articles · Updated · The Verge · Oct 22
Summary
Science published Microsoft’s study showing its Diagnostic Orchestrator hit about 83% accuracy on New England Journal of Medicine cases, versus 36% for generalist physicians using references and 22% without them.
The system also scored 80% on emergency-department cases from Beth Israel Deaconess Medical Center, compared with 47% for human physicians, while reaching diagnoses at lower estimated cost.
Microsoft said the tool uses multiple AI agents powered by OpenAI’s GPT-5 to question symptoms, choose tests and reason through cases like a panel of specialists; it plans to open-source the SDBench benchmark behind the study.
Real-world use remains unproven because the tool has only been tested on historical cases, and executives said it still needs clinical validation before wider deployment.
Hospital testing with Beth Israel Deaconess is expected to begin in the next couple of months as Microsoft pursues broader 'medical superintelligence' and possible commercialization.