An AI model outperformed physicians at diagnosing in emergency room conditions, a study showed.

ER diagnoses are challenging, characterized by limited information and little time, but high-stakes outcomes: A fever could be a seasonal flu, or it could be life-threatening sepsis. The researchers tested an OpenAI model, giving it medical profiles and patient history; it achieved a correct or close-to-correct diagnosis in 67% of cases, compared to 55% for physicians.

In 10 years, AI agents will likely be commonplace in emergency medicine, Science reported, leading to faster and more informed decisions. The CEO of one medical AI startup thinks that timeline should be faster, and that it borders on “malpractice” if doctors do not use frontier models for second opinions already.