
Accelerating the AI Revolution in Medicine
Peter Lee, President of Microsoft Research, joins Stanford hosts Justin Norden and Matthew Lungren for a conversation on artificial intelligence in clinical practice. Lee discusses post-training compute and what current benchmarks do and do not capture about model performance, then turns to Microsoft Research's work on sequential diagnosis, where AI systems reason through a case step by step rather than producing a single answer. The discussion covers AI agents built to reduce administrative burden on clinicians, the difficulty of evaluating medical AI in ways that reflect real-world use, and what it would take to build systems that collaborate with physicians rather than replace their judgment. Part of Stanford's AI in Healthcare series, the talk stays close to concrete examples of diagnostic reasoning and model testing rather than general predictions about AI, giving a working researcher's view of where the technology is actually useful in medicine today.