JMIR Publications

ABSTRACT

Artificial intelligence (AI) has already been integrated into the healthcare field, much of which is in the realm of diagnostic imaging, with sensor technology and the Internet of Medical Things playing an increasing role. Software capitalizing on advances with Large Language Models (LLMs) has been introduced to replace or assist traditional physician roles, such as taking the medical history and documentation in the medical record. While such well-meaning efforts to improve workflow efficiency may, in the short term reduce physician burnout, the increasing expectations of physician productivity may have a negative impact in the long term, especially as healthcare systems face provider shortages and pressures to maximize profit. There is undeniable allure to incorporate using LLMs in clinical diagnostic support. Physicians must understand the limitations of these models, as well as the legal implications associated with their use, especially when there is discordance between the clinical judgment of the human physician and the AI model. Efforts are needed to empower physician decision support by using AI to enable more meaningful and effective access to existing medical knowledge. Those who control access to the repositories of research and clinical content are encouraged to revamp access to this information, such as by providing a turnkey chat-style user interface to their proprietary content. Furthermore, we encourage medical teaching institutions to include in the curriculum, the fundamental concepts of applied artificial intelligence.