AI is starting to beat doctors at making correct diagnoses
science.org
May 3, 2026
5 min read
🔥🔥🔥🔥🔥
30/100
Summary
AI systems are expected to assist in emergency rooms by providing faster diagnoses and supporting care teams in decision-making. Continuous monitoring of patients' vital signs and AI analysis of doctor-patient conversations will help identify potential errors in diagnoses.
Key Takeaways
- A large language model (LLM) outperformed physicians in diagnosing complex conditions in emergency care, achieving correct or close diagnoses in 67% of cases compared to 50-55% for doctors.
- The LLM, OpenAI's o1 model, received a perfect clinical reasoning score for 98% of cases examined, while attending physicians achieved this in only 35% of cases.
- The study demonstrated that AI can effectively analyze incomplete or biased real-world data during the early triage process in emergency rooms.
- The adoption of AI systems in healthcare will depend on understanding the contexts in which they are most reliable.
