Back to news
Large Language Models
May 3, 2026

Study finds AI models may outperform doctors in emergency room diagnoses

May 3, 2026
AI Summary

A study from Harvard Medical School indicates that AI models can provide more accurate diagnoses than human physicians in emergency room settings. The findings highlight the potential of AI in medical diagnostics but also emphasize the need for further research and accountability in its application.

  • A study published in Science examined the performance of AI models in medical diagnostics, specifically in emergency room cases.
  • Researchers compared diagnoses from OpenAI's models with those from two attending physicians at Beth Israel Deaconess Medical Center, using data from 76 patients.
  • The AI model o1 achieved an accurate or close diagnosis in 67% of triage cases, outperforming one physician at 55% and another at 50%.
  • The study emphasized that AI was tested with unprocessed data from electronic medical records and highlighted the need for prospective trials to assess AI in real-world patient care.
  • The researchers noted limitations in AI's reasoning capabilities with non-text inputs and the lack of a formal accountability framework for AI diagnoses in medical settings.
aihealthcarediagnosisemergency roommachine learning

Related Stories

Large Language Models

Thinking Machines Lab develops AI model for simultaneous conversation

1h ago

Large Language Models

ChatGPT Sees Increased Adoption Among Older Users in Early 2026

15h ago

Large Language Models

Optimizing Matrix Multiplication for Swift in LLM Training

1d ago

Large Language Models

arXivLabs Encourages Collaboration on New Features with a Focus on Privacy

2d ago