Skip to content
Hacker News front page

OpenAI's o1 correctly diagnosed 67% of ER patients vs. 50–55% by triage doctors

OpenAI's o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors

OpenAI o1 correctly diagnosed 67% of ER triage patients, versus 50–55% for doctors. The title cites a Harvard trial, but the RSS post does not disclose sample size, case mix, or evaluation protocol. Practitioners should track the test setup, not only the accuracy gap.

Why it matters: HKR-H/K/R all pass: a high-risk ER comparison gives the hook, 67% vs 50–55% gives a testable number, and clinical trust/safety creates resonance. Missing sample size and protocol keep it in 78–84, not P1.

Read the original ↗Export Markdown