Microsoft's diagnostic agent beats generalist doctors in a peer-reviewed Science study
MAI-DxO, an orchestrator that questions, orders tests and debates itself, now has a journal behind its claims and a hospital trial ahead.
Published in Science, Microsoft's MAI Diagnostic Orchestrator got 83% of hard New England Journal of Medicine cases right. Generalist physicians managed 36% with reference materials and 22% without, and the system ordered fewer tests.
Next comes a prospective study at Beth Israel Deaconess. For agent builders, it is a rare peer-reviewed result for an orchestrated, multi-step agent.


