TechStuff How Soon Until AI Out-Diagnoses Your Doctor? - The Story
22 snips
Feb 25, 2026 Dr. Dhruv Khullar, a physician-journalist who writes on medicine and AI, explores how chatbots can both catch elusive diagnoses and make dangerous errors. He contrasts dramatic diagnostic wins with routine, long-term care challenges. The conversation covers hallucination risks, clinician-AI workflows, equity of access, and how doctors can shape AI’s role in healthcare.
AI Snips
Chapters
Transcript
Episode notes
ChatGPT Solved Yearslong GI Mystery
- Matthew Williams was misdiagnosed in an ER, later had life‑saving surgery, then suffered years of diarrhea until ChatGPT suggested oxalate sensitivity.
- He changed diet with a nutritionist after ChatGPT identified oxalates in leafy greens and regained his life.
ChatGPT Recommendation Caused Bromide Poisoning
- A man asked ChatGPT for low‑salt alternatives and the bot suggested sodium bromide, which he ordered and then developed severe bromide toxicity.
- He hallucinated, required weeks of treatment, and highlights AI risk when users act on suggestions without oversight.
Specialized AI Matched Expert Diagnostician
- Cabot, an AI trained on clinical pathological conferences, solved a Harvard CPC and reached the correct diagnosis (Lofgren syndrome) within minutes.
- The room went silent when Cabot matched an expert diagnostician, showing AI can crack very complex cases when fed curated data.