A Call to Address the Generative AI-Environment Paradox in Graduate Medical Education

Generative AI has been hailed as a positive catalyst for revolution in medical education and health care.1,2 At the same time, systems vital for human life and health are faltering due to compounding environmental crises.3,4 While AI has the potential to support sustainability initiatives, including in health care, rapid uptake of AI is accelerating drivers of climate change and climate-driven health risks.3,[5][6][7] Large language models (LLMs; eg, ChatGPT, Copilot, Gemini), multimodal LLMs, and other generative AI systems require data centers that consume massive amounts of water and energy.[8][9][10][11] Demand for this AI is fueling rapid expansion of data centers-and their associated resource consumption and greenhouse gas (GHG) emissions.6,12 Confronted with both realities, graduate medical education (GME) must simultaneously prepare trainees for an ongoing climate crisis and widespread use of AI.How can GME reconcile this innovative technology's potential to improve health with its deleterious effects on the natural systems on which human life depends?Currently, AI discussions in the medical literature center on privacy, confidentiality, efficiency, and accuracy.GME concerns also include mis-skilling, de-skilling, and never-skilling.13,14 As yet, environmental and patient health concerns do not appear to be key factors in selecting or employing these tools.In a 2024 American Medical Association survey, two-thirds of physicians-a 78% increase from the prior year-used AI. 15 The most common uses of AI among those surveyed were note documentation, other charting, billing and coding, and the creation of discharge instructions.15 Many medical students, GME trainees, and faculty query LLMs daily.Some uses are intentional, driven by individual curiosity or pursuit of efficiency.Other uses are more insidious, as AI tools are integrated into electronic medical record systems.16,17 Most trainees and faculty likely BOX Questions for GME on the Generative AI-Environment Paradox & How do residents, fellows, and faculty currently approach the AI-environment paradox?& What are GME educational gaps and curricular opportunities around the AI-environment paradox?& What tools are needed for trainees and programs to measure the environmental effects of AI deployed in GME settings?& How can the GME community assist in seeking transparency around the resource intensity of health care-related AI tools?& How can discussions about AI tool conservation be included in program or institutional decisions about AI use? & How is AI used currently to improve waste reduction and sustainability efforts in GME training settings?& How could AI be included in GME curricula on medical ethics, aside from privacy and confidentiality concerns?& How can systems provide meaningful feedback on health care delivery-associated energy use and environmental impacts at the individual, program, and organizational levels?