Ars Technica
1 month agoMedicine
Passing part of a medical licensing exam doesn't make ChatGPT a good doctor
ChatGPT struggles significantly in real-world medical diagnostics despite performing well on multiple-choice tests.
The study highlights limitations in AI when applied to complex medical cases with complications.
Google's AMIE demonstrates advanced diagnostic capabilities, suggesting that specialized AI may outperform general models like ChatGPT. [ more ]