[Environmental burden of generative artificial intelligence from a planetary health perspective: A narrative review]

Generative artificial intelligence (GenAI) is widely used in healthcare to support evidence-based practice, policymaking, and research activities. Hundreds of millions of people worldwide use GenAI for a wide range of purposes, and this number continues to grow. In recent years, the discipline of "planetary health" has become mainstream as an urgent global issue, as excess mortality and morbidity attributable to extreme climate events have been reported worldwide. Although GenAI is increasingly used as a tool to promote planetary health, the rapid expansion of its use can add a substantial burden to the global environment. In Japan, however, the environmental impact of GenAI is rarely discussed. In this narrative review, we summarized quantitative data on the environmental impacts of GenAI on the basis of academic literature and reports from various industries. Data centers supporting the day-to-day operation of GenAI and the manufacturers of the required hardware consume vast amounts of electricity and water and emit large quantities of greenhouse gases. Furthermore, the mining of rare metals and the disposal of hardware release harmful chemicals that may contaminate soil and water. Unlike several other countries and international organizations, Japan currently lacks specific legislation and guidelines addressing the environmental impact of GenAI. The environmental burden of GenAI is underestimated because the reporting is largely voluntary and limited data are disclosed. Legislation, guidelines, and reporting standards promoting the transparent disclosure of environmental impacts will facilitate more reliable assessments. It is also important for individual users to make informed choices, such as selecting energy-efficient models or AI systems optimized for specific tasks.