Canadian researchers say we really shouldn’t trust ChatGPT for answers to our medical questions. They asked the AI to answer 150 Medscape case challenges – a series of diagnostic problems for doctors – and found it was only able to answer just shy of half of them correctly. Additionally, ChatGPT was not very sensitive in its answers, and was overly complicated in half of its responses, as well as only providing complete and relevant answers just over half of the time.
