Dear Editor Thank you for this timely experiment dealing with ChatGPT’s passing a family medicine exam (Morreel et al. Citation2023). It is particularly useful that the authors have given details of their processes, so that replication, by others in other contexts, is possible for comparative purposes. In addition, I find it noteworthy that ChatGPT was not disadvantaged by the negative phrasing of questions. Early experiments such as these are necessary so that health professions educators can test various limits and determine best practices as the AI models evolve and improve. As the authors discovered, ChatGPT’s results can be unexpectedly variable, but, even at this stage, such work raises several points and questions that will need to be addressed, sooner, rather than later.
展开▼
