Open AI's GPT-4 chatbot, created by Harvard's Beth Israel Deaconess Medical Center scientists, has shown promise in medical diagnostics.

In a JAMA Internal Medicine study, the chatbot's diagnostic justifications were on par or better than those of assistant doctors and residents.

GPT-4 achieved a higher median R-IDEA score of 10, surpassing attending physicians' score of 9 and residents' score of 8, showcasing its advanced reasoning skills.

Despite its successes, GPT-4 exhibited a higher error rate than residents, highlighting areas for improvement in AI-assisted clinical decision-making.

The findings suggest that AI like GPT-4 could enhance, but not replace, human clinical reasoning, necessitating ongoing research into its integration in healthcare.