AI‑Doctor Test Experiment

A medical researcher did a new study. He wanted to learn how well a top chatbot (聊天机器人) could find out what was wrong with patients. He made three groups. Doctors alone made up the first group. The second group was a group of doctors using the chatbot. The third group was only made up of the chatbot.

Each group got six new patient cases. They needed to find out the patients' problems and the best treatments. Some medical experts checked the results. They didn't know which group gave which answers.

The researcher is a doctor in Boston. He was surprised by the results. The chatbot alone did the best. It scored 90 percent correct with its answers. Next came the doctors using the chatbot, with 76 percent. The doctors alone followed in third place, with 74 percent correct.

Experts said the doctors using the chatbot didn't listen to its advice often. This was usually a mistake. Some people noted the doctors didn't use the chatbot well. They asked simple questions, similar to an online search. Experts said a better way to use AI models is to put the full, detailed question into the chatbot and see what answer it comes up with.

Another point is that the AI's typing is much easier to read than a doctor's messy handwriting on prescriptions, even though the study didn't check this.