Radiologists who trained for over a decade to read X-rays are being outperformed by AI models that spot tumors they miss.

A Microsoft study found that when their AI system was paired with an OpenAI model, it solved more than 80 percent of complex health cases, compared to a 20 percent success rate for human doctors.

In an emergency-department simulation, GPT-4 outperformed resident physicians in diagnostic accuracy based solely on patient histories.

A mother who took her son with a rare neurological condition to 17 doctors came back with no diagnosis, but when she uploaded all of his medical documentation to ChatGPT, it spit out a correct diagnosis in minutes. Six weeks later, her son had lifesaving surgery.