Editorial Commentary: Artificial Intelligence and Language Learning Models Can Be Improved by Curated Input of Medical Training Data but Still Face the Limitations of Available Literature and Require Continued Human Oversight

Abstract

Artificial intelligence and language learning models (LLMs) are rapidly evolving. Several popular and easily accessible platforms, like ChatGPT and Gemini, are increasingly being explored by clinicians and patients for their utility in clinical decision making. Although these tools provide rapid access to information, their inconsistent adherence to evidence-based guidelines raises concerns. A potential solution is to generate more specialized LLMs for orthopaedics. A curated database of validated orthopaedic literature can be used as input in order to address concerns about the quality of input data. However, a curated LLM may still have limitations of selection bias and limited high-quality literature. In addition, patients using these models may possess limited health literacy. LLMs represent an advancement and potentially powerful clinical tool but still require ongoing evaluation, refinement, and validation. Artificial intelligence should continue to be viewed as an evolving resource rather than a replacement for clinical judgment.