AI in medicine talk at UCLA: doctor+GPT4 teams can underperform AI alone, only 25% of US adults comfortable
jonc101x · x · 2026-10-17
Researcher jonc101x announced an Oct 26 noon talk at UCLA's Frontiers in Computational Biosciences seminar, "AI in Medicine: Real Magic or Technological Illusions?", featuring live demos that mix real capability with illusion.
Key points from the thread: AI models now match or exceed physicians on simulated clinical tasks, yet human-AI teams can underperform the AI working alone; capability and adoption are outrunning evidence, governance, and public acceptance — only 25% of US adults say they're comfortable with their doctor using AI. The slides show a randomized study of 50 licensed physicians across 6 clinical vignettes, scoring diagnostic reasoning for combos like Doctor+Internet and Doctor+GPT4.
More from Companies & People
- "If you use AI to replace yourself, you will get replaced" — jbarbier — jbarbier · 2026-10-17
- DeepSeek fanatic explains: it's about talent and strategy fundamentals, not benchmarks — teortaxesTex · 2026-10-17
- "Startups LARPing 996 without any karoshi cases": tech culture jab goes viral — HanchungLee · 2026-10-17
- 10 years of ML meetups: 5 organizing rules from ParisMLgroup veteran — IgorCarron · 2026-10-17
- Musk shows new Tesla Roadster as "flying cars" moment, echoing Thiel's stagnation thesis — esrtweet · 2026-10-17
- OpenAI London staff earn average of £2.3m per person annually — IndependentCoast3974 · 2026-10-17