"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis Universal Medical Intelligence: OpenAI's Plan to Elevate Human Health, with Karan Singhal
122 snips
Feb 25, 2026 
Karan Singhal, Head of Health AI at OpenAI and safety-minded researcher who led ChatGPT Health and HealthBench. He discusses building physician‑level performance, working with hundreds of clinicians, constructing the 49,000‑criterion HealthBench, multimodal medical inputs, privacy and safety safeguards, clinical trials of AI copilots, and how AI could become part of routine medical practice.
AI Snips
Chapters
Transcript
Episode notes
Collect Records And Use Integrated Tools To Advocate For Care
- If you're a patient, advocate by collecting and sharing medical records; the activation energy to use AI drops as products improve.
- Use ChatGPT Health's forthcoming integrations to connect EMR and wearables for richer context.
Models Matched Attending Oncologist In Many Cases
- Nathan reported frontier models often matched attending oncologist-level reasoning and disagreed only on fine multimodal judgments like skin pallor or breathing.
- He ran recommendations in duplicate/triplicate across GPT-5 Pro, Gemini 3, and Claude during inpatient care.
Provide Full Clinical Context To Improve AI Accuracy
- Give models as much relevant context as possible (EMR, wearables, region-specific epidemiology) to improve accuracy and tailoring.
- Use ChatGPT Health integrations to lower activation energy for uploading records and connecting wearables.