ChatGPT excels at differential diagnostics in hard cases

A recent study conducted by Harvard researchers at Beth Israel Deaconess Medical Center has shown that ChatGPT-4, a generative AI model, is effective in providing probable diagnoses for difficult-to-diagnose patient cases. The researchers challenged the ChatGPT bot to give differential diagnoses for 70 complex cases previously selected for medical conferences. ChatGPT-4 provided the same final diagnosis as expert physicians in 39% of the cases and included the final diagnosis in its differential lists for 64% of the cases. The researchers concluded that generative AI is a promising tool to assist in medical diagnosis, although further research is needed to investigate biases and blind spots in such models.

In other news, a survey conducted by Capgemini Research Institute found that over half of 10,000 adults across 13 countries are aware of and have tried generative AI technology. The adoption of ChatGPT and similar AI technologies was found to be consistent across different age groups and geographies, with 67% of consumers worldwide trusting generative AI for medical advice. Additionally, there are discussions on the need for retraining the healthcare workforce to ensure safe and effective use of AI in healthcare, and the potential impact of AI on the U.S. vs. China “AI race” is being debated.