Microsoft's AI Beats Doctors at Diagnosis, But Not Ready for Your Hospital Yet

Microsoft announced that its AI tool, called Diagnostic Orchestrator, significantly outperformed human doctors in diagnosing medical conditions. In tests using published case reports and emergency room data, the AI achieved about 83% accuracy, compared to 36% for doctors using references. The system uses multiple AI agents to mimic a panel of experts, asking questions and ordering tests before reaching a diagnosis. It also tended to be cheaper. The research was published in the journal Science.

However, this doesn't mean you'll see AI doctors anytime soon. The AI was only tested on historical cases, not real patients, so its real-world performance is unknown. Microsoft admits it needs clinical validation and plans to start hospital testing in the next couple of months. There's also no timeline for when it might be available to the public, and strict regulations must be followed. So, while the results are impressive, they're a milestone, not a revolution yet.

For everyday people, this could eventually mean faster, more accurate diagnoses and lower costs. But for now, it's a glimpse of what's coming. Microsoft is also working on other AI health tools, like coordinating care and making treatment decisions. The company calls this push 'medical superintelligence'—AI that can outperform the best human-led healthcare. But don't expect to see it in your doctor's office next year; it's still early days.