Microsoft's Horvitz discusses design pitfalls in clinician-AI collaboration
erichorvitz · x · 2026-08-19
Microsoft Chief Scientific Officer Eric Horvitz appeared on the Nature Digital Medicine podcast to discuss clinician-AI collaboration design. Based on recent research, the conversation covers AI-first versus second-opinion workflows and the pitfalls of treating LLMs like search engines. The study found that while AI integration compressed performance outliers to boost average diagnostic accuracy, it also risks "sycophancy," where models overly accommodate a clinician's initial feedback at the expense of independent analysis.
Related event: Microsoft's Horvitz on Designing Clinical AI Collaboration(2 posts)→
More from AGI Musings
- Interp researcher pushes back: if you can't simulate a neuron, you haven't explained it — nabla_theta · 2026-10-16
- Virtue ethics meets AI alignment: what would it mean to make a model virtuous? — joelbot3000 · 2026-10-16
- Professor: this rec letter season I'm evaluating the student plus Claude — RoozbehMottaghi · 2026-10-16
- Prediction: Riemann Hypothesis likely falls to AI before 2030 — gautamcgoel · 2026-10-16
- Models are superintelligent where data is abundant, and weaker everywhere else — burny_tech · 2026-10-16
- ChatGPT's Dot Reviewed: Your Chief of Staff Is Still Just a Codex Wrapper — infinitely_zero · 2026-10-16