Danielle Bitterman, an Assistant Professor of Radiation Oncology at Harvard Medical School, recently shared on LinkedIn:
“Our study evaluating the ability of AI chatbot ChatGPT to provide cancer treatment recommendations is hot off the press! ChatGPT includes recommendations that disagree with national guidelines 1/3 of the time. Incorrect responses are mixed in with correct responses, making them especially difficult to detect. These results highlight the promise, but also some key safety concerns, of Large Language Models AI for clinics. Take a look at our findings here:”
Source: Danielle Bitterman/LinkedIn