/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

days · browse · Enter similar · o open

Study: OpenAI's o1 correctly diagnosed 67% of emergency room patients using electronic records and a few sentences from nurses, vs. to 50-55% for triage doctors

Researchers say results mark a ‘profound change in technology that will reshape medicine’  —  From George Clooney in ER …

The Guardian Robert Booth

Context & Ripple Effects

The related coverage traces a long shift from AI as a diagnostic aid to hospital use in risk prioritization, targeted care, and point-of-care tools such as AI stethoscopes. This study puts a general-purpose OpenAI model into the more consequential emergency-department triage context.

Its result also sits beside recent coverage finding ChatGPT Health frequently misclassified the urgency of emergencies and nonurgent cases. That contrast makes validation conditions—not just a headline accuracy comparison—central to whether such systems can be trusted in clinical workflows.

First-order effects

  • OpenAI gains a high-profile piece of evidence for o1 in clinical reasoning from electronic records and nurse notes, while triage teams get another benchmark against which AI-assisted intake can be evaluated.
  • The reported result will increase scrutiny of how the study defined correct diagnoses, selected cases, and handled urgency, because emergency-room triage requires safe prioritization as well as diagnostic accuracy.

Second-order effects

  • Hospitals and clinical-AI vendors will face pressure to distinguish narrowly validated diagnostic-support uses from autonomous triage claims, especially given the conflicting severity-assessment result in related coverage.
  • Health systems already using predictive prioritization tools may seek comparable evaluations of their models and workflows rather than treating general-model performance as directly transferable to patient care.

Third-order effects

  • If replicated across settings and paired with reliable safety performance, clinical AI is likely to move from specialist point tools toward workflow-level support built around medical records, clinician notes, and escalation decisions.
  • The durable constraint will be evidence and accountability: inconsistent results across medical-AI evaluations could push the market toward more rigorous, use-specific validation before broad deployment.

The trend: This is one data point in the transition from AI diagnostic augmentation to AI embedded in frontline clinical decision workflows, with validation quality determining adoption.

Discussion

  • Vox Dylan Scott on x
    A major new study found AI outperformed doctors in ER diagnosis — but there's a catch
  • @emollick Ethan Mollick on x
    New paper (on an old AI) tests o1 against doctors on medical benchmarks & real ER cases: “across a variety of scenarios and applications, the large language model outperformed both human physicians and older models” The potential suggests an “urgent need for prospective trials.” …
  • Gilles Frydman Gilles Frydman on linkedin
    First of several posts on the new Science paper claiming LLMs have eclipsed the benchmarks of clinical reasoning. …
  • Bill Faruki Bill Faruki on linkedin
    Harvard, in Science today: OpenAI's o1 reasoning model out-diagnosed expert ER physicians on real triage cases — 67% vs. 50-55% — and beat them on management planning 89% to 34%. …
  • @dropthet Mr. Herbert Garrison on bluesky
    FYI, this model is almost 2 years old now.  Studies take at least this long to catch up.  [embedded post]