Key Takeaways
- Microsoft’s AI system, MAI-DxO, diagnosed complex medical conditions up to four times more accurately than doctors.
- The AI system achieved 83% accuracy in complex cases, compared to 36% for doctors with reference materials and 22% without.
- Microsoft plans to test the technology in real hospital settings through a partnership with Beth Israel Deaconess Medical Center.
Microsoft has developed an artificial intelligence system that outperforms doctors in diagnosing complex medical conditions, according to research published in the scientific journal Science.
The system, called Microsoft AI Diagnostic Orchestrator (MAI-DxO), uses multiple AI agents to analyze symptoms, review medical histories, order tests, and identify diseases.
In tests using 304 complex medical cases published in The New England Journal of Medicine and 325 emergency department cases from Beth Israel Deaconess Medical Center, the AI system performed significantly better than human doctors.
The AI system achieved 83% accuracy in complex cases without reference materials, compared to 22% for doctors without references and 36% for those with references.
For emergency department cases, the AI system reached 80% accuracy, compared to 47% for doctors.
Microsoft designed the Diagnostic Orchestrator to function like a team of medical specialists working together to diagnose a patient, updating its conclusions as new evidence becomes available.
The system uses OpenAI’s GPT-5 model and coordinates several AI agents that challenge each other’s conclusions before reaching a diagnosis.
Microsoft also developed a testing framework called Sequential Diagnosis Benchmark (SDBench) to evaluate how AI systems and doctors diagnose conditions through multiple steps.
The company plans to make the benchmark available to researchers so they can evaluate other AI systems.
Despite the promising results, Microsoft acknowledged that further clinical testing is necessary before the technology can be widely deployed.
The research relied on previously documented medical cases rather than direct interactions with patients, and compared the system with individual generalist physicians, not specialist teams.
Microsoft plans to begin testing the Diagnostic Orchestrator in real hospital settings through a partnership with Beth Israel Deaconess Medical Center in Boston.
Dr. Dominic King, Microsoft’s AI health lead, told The Verge that testing could begin within the next couple of months.
