How I Doctor with Dr. Graham Walker What Doctors Get Wrong About AI with Robert Wachter, MD
28 snips
Jan 29, 2026 
Dr. Robert Wachter, UCSF chair and author on digital health, brings grounded optimism about AI in medicine. He discusses early ward tools, risks of de‑skilling, designing AI that preserves trainee reasoning, trust versus instinct, equity gaps for vulnerable patients, and cautious low‑risk rollouts with local accountability.
AI Snips
Chapters
Books
Transcript
Episode notes
De‑skilling Is Real And Potentially Dangerous
- Wachter frames de-skilling as a real but nuanced risk that can be acceptable for unused skills.
- He warns de-skilling could create a death spiral where humans degrade and AI becomes the only reliable decision-maker.
Force Trainees To Think Before Consulting AI
- Teach fundamentals first and require trainees to commit their diagnosis before consulting AI.
- Build AI tools that prompt learners to state their reasoning before revealing the model's output.
Vinod Khosla Eyeball Test Exchange
- Wachter recounts a conversation with Vinod Khosla who argued machines will surpass subtle clinician judgments.
- He pushed back, saying translating human eyeball tests into machines is harder than venture optimism assumes.
