AI Diagnosed Complex Medical Cases Better Than Experienced Physicians. That’s Not the Most Interesting Part.
This week, Microsoft released one of the most fascinating studies I’ve read in a long time. Using 304 exceptionally complex cases from the New England Journal of Medicine, their new AI diagnostic system, MAI-DxO, correctly diagnosed 85.5% of cases, while a group of experienced physicians achieved just 19.9% accuracy. The physicians had between five and twenty years of clinical experience, and the AI reached its conclusions while ordering roughly 20% fewer diagnostic tests.
Before anyone concludes that AI is about to replace physicians, it’s worth understanding what this study actually measured. These weren’t everyday primary care visits. They were some of the most diagnostically challenging cases published in one of the world’s leading medical journals. The participating physicians weren’t allowed to consult colleagues, references, or AI tools as they normally would, and the system has not yet been validated in routine clinical practice.
