The application of large language models in orthopedic postgraduate education: potentials, challenges, and future prospects

With the widespread integration of artificial intelligence (AI), orthopedics postgraduate education is transitioning into the intelligent era. Large language models (LLMs), which leverage deep learning and natural language processing (NLP), have profoundly influenced orthopedic postgraduate education through their sophisticated capabilities in coherent comprehension, contextual response and content generation. These models encompass both general-purpose tools (e.g., ChatGPT) and specialized orthopedic applications (e.g., DocOA, BioinspiredLLM, MechGPT, DrSR, and AmbossGPT). They can offer interdisciplinary research training, targeted academic guidance, real-time resource access, interactive case exercise and immersive simulation practice in orthopedics. Most models exhibited promising performance: for instance, ChatGPT-4 achieved a 61.2% accuracy on the Orthopedic In-Training Examination (OITE) comparable to orthopedic residents, while DocOA significantly outperformed ChatGPT-4 with more than 142% improvement in orthopedic benchmark evaluations. The LLMs' capabilities could promote a personalized, interactive, and adaptive transformation of orthopedic postgraduate education. However, the application of LLMs faces significant challenges such as over-reliance, delayed updates and inconsistent outputs, sparking ongoing controversy in the medical education. Moving forward, establishing a comprehensive human-AI collaborative framework is imperative to optimize the application of LLMs in orthopedic postgraduate education. This holistic framework integrates learner-centered perspectives, multidimensional governance, phased implementation strategies, and geographic diversity. Together, adopting this innovative human-AI approach will strengthen the cultivation of high-level orthopedic talents for orthopedic postgraduate education.