Microsoft says AI system outperforms doctors in simulated disease diagnosis

Microsoft said Thursday that its artificial intelligence system diagnosed complex medical conditions more than twice as accurately as physicians in simulated tests, citing research published in the journal Science.

The company’s AI Diagnostic Orchestrator (MAI-DxO) achieved 83% diagnostic accuracy on complex medical cases, compared with 36% for physicians using textbooks and online resources, according to a Microsoft statement.

Physicians working without external reference materials achieved 22% accuracy.

Researchers evaluated 304 complex cases drawn from The New England Journal of Medicine, using a system in which multiple AI agents work together to assess symptoms, select diagnostic tests and reach conclusions.

The system also achieved 80% accuracy on more typical emergency department cases, compared with 47% for physicians, Microsoft said.

The company said MAI-DxO could also reduce diagnostic testing costs by selecting fewer or less expensive procedures while maintaining high accuracy.

In one cost-optimized configuration, the system achieved 81% accuracy on complex cases at less than half the estimated cost of comparative systems.

Microsoft described the findings as progress toward “medical superintelligence,” its vision of AI systems capable of surpassing leading clinical teams in medical reasoning and care coordination.

However, the results were based on simulated cases rather than real-world clinical practice, and the physician comparison did not include specialist teams.

The company said further evaluation in clinical environments is necessary to establish how the technology performs in actual patient care.

Microsoft plans to evaluate the system through a prospective study with Beth Israel Deaconess Medical Center in Boston, comparing its diagnostic assessments with those of physicians in hospital settings.