Study: AI can outperform doctors at diagnosing cases

A Science study found OpenAI’s o1 reasoning model often diagnosed unseen medical cases more accurately than GPT-4, physicians, and residents in text-based tests.

The model was also evaluated on electronic health record scenarios and showed strong clinical reasoning performance.

Experts said the results are promising for clinician support, but the study was limited to text-only cases.

They warned that real-world medical settings are far more complex, so prospective trials are still needed before clinical use.