New Microsoft AI surpasses doctors in diagnosing difficult medical cases

Microsoft has developed an artificial intelligence system that it claims surpasses human doctors in diagnosing complex medical cases, potentially paving the way for a new era of superintelligent medical AI.

According to The Guardian, Microsoft’s AI division, led by technology pioneer Mustafa Suleyman, created a system designed to emulate a team of doctors managing cases that require sophisticated reasoning skills.

When integrated with OpenAI’s advanced AI model o3, the system correctly diagnosed eight out of ten complex cases. By contrast, practising doctors working alone—without consulting colleagues, textbooks, or AI tools—correctly diagnosed only two out of ten cases.

Microsoft highlighted that this AI system is not only more accurate but also more cost-effective than relying solely on human physicians, particularly in ordering appropriate diagnostic tests.

However, the company was quick to reassure medical professionals that AI is intended to complement, rather than replace, doctors. “We believe AI will augment, not substitute, the role of physicians,” Microsoft stated in a blog post. “The clinical role of human doctors goes far beyond diagnosis; they build trust with patients and families in ways AI cannot replicate.”

Mustafa Suleyman told The Guardian that the technology will become increasingly reliable in the coming years. “We are on a path where these systems will be nearly error-free within the next five to ten years,” he said. “This will significantly ease the burden on healthcare systems worldwide.”

Nonetheless, Microsoft acknowledged the system’s limitations. The research was conducted under controlled conditions using 304 complex cases from the New England Journal of Medicine and has not yet undergone peer review or live clinical testing. The company also noted that AI systems sometimes rely on memorised answers rather than deep understanding—particularly in multiple-choice medical licensing exams—potentially leading to an overestimation of their real-world competence.

While the AI’s performance in this controlled study is impressive—outperforming human doctors by a wide margin—experts caution that further validation and clinical trials are needed before such systems can be safely and effectively integrated into everyday medical practice.