医疗专业人士认可 ChatGPT Health 发布，尽管存在 AI 幻觉风险

The Hallucination Problem

　　The main issue with chatbots remains "hallucinations," which is particularly critical in healthcare. A Vectara study showed that OpenAI's GPT-5 "hallucinates" more frequently than competitors from Google and Anthropic.

　　However, Stanford University professor of medicine Nigam Shah considers these concerns secondary. According to the doctor, the real problem with the system is the difficulty in accessing doctors, not the risk of receiving incorrect advice from ChatGPT.

"Currently, if you approach any healthcare system and want to see a primary care physician, you'll have to wait three to six months. If you have a choice: wait six months to see a real specialist or immediately talk to someone who can do something for you, which would you choose?" he noted.

　　Administrative tasks can take up about half of a doctor's time, significantly reducing the number of appointments. Automating these processes would allow specialists to devote more attention to patients.

　　Dr. Shah leads a team at Stanford developing ChatEHR, software that enables doctors to work with patients' electronic medical records more rationally and efficiently.

"By making it more user-friendly, doctors will be able to spend less time searching for the necessary information," said one of the first testers of ChatEHR, Dr. Sneha Jain.

　　In January, Anthropic announced the release of Claude for Healthcare, a set of tools for healthcare providers and patients.