JAMA Network Open features CODEX Director Sumant Ranji’s invited commentary on the popular results of the Goh et al. study, “Influence of a Large Language Model on Diagnostic Reasoning: A Randomized Clinical Vignette Study.”
The findings from the recently published “Influence of a Large Language Model on Diagnostic Reasoning: A Randomized Clinical Vignette Study” by Goh et al. have been gaining widespread attention* in the diagnostic excellence, patient safety, and artificial intelligence communities – and for good reason.
It raises urgent questions, such as what do demonstrations like this of the diagnostic capabilities of large language models (LLM) mean for clinicians? And will LLMs ultimately be capable of preventing harm from diagnostic errors?
In his invited commentary for JAMA Network Open, Sumant Ranji, MD, director of UCSF CODEX, explores the nuances of the questions that followed the study and the potential path ahead for integrating LLMs to achieve diagnostic excellence.
“…the result that prompted consternation was the performance of the LLM alone, which scored significantly higher than either group of physicians. On hearing these results, more than one audience member wondered aloud, ‘Are we going to be out of a job?’”
-- From “Large Language Models—Misdiagnosing Diagnostic Excellence?”, published on October 28, 2024.