In the rapidly evolving landscape of artificial intelligence, researchers Min-Kyu Kim, Tae-An Yoo, and Ji-Bum Chung from the Korea Advanced Institute of Science and Technology (KAIST) have turned their attention to a pressing issue: the environmental impact of generative AI. Their recent study, published in the journal “Environmental Science & Technology,” delves into the carbon footprint and environmental implications of AI, particularly focusing on the often-overlooked inference phase.
Generative AI, while creating substantial social and economic value, has raised concerns due to its high energy consumption and potential environmental sustainability issues. Previous studies have primarily focused on the energy-intensive training phase, but the cumulative environmental footprint during large-scale service operations, especially in the inference phase, has not been thoroughly examined. This study aims to bridge this gap by conducting a scoping review of methodologies and research trends in AI carbon footprint assessment.
The researchers analyzed the classification and standardization status of existing AI carbon measurement tools and methodologies. They compared the environmental impacts arising from both training and inference stages, identifying how factors such as model size, prompt complexity, serving environments, and system boundary definitions influence the carbon footprint. The study reveals critical limitations in current AI carbon accounting practices, including methodological inconsistencies, technology-specific biases, and insufficient attention to end-to-end system perspectives.
Building on these insights, the researchers propose several future research and governance directions. These include establishing standardized and transparent universal measurement protocols, designing dynamic evaluation frameworks that incorporate user behavior, developing life-cycle monitoring systems that encompass embodied emissions, and advancing multidimensional sustainability assessment frameworks that balance model performance with environmental efficiency.
For the energy sector, this research underscores the importance of considering the environmental impact of AI technologies, which are increasingly being integrated into energy management and optimization systems. By understanding and mitigating the carbon footprint of AI, energy companies can work towards more sustainable and efficient operations. The study provides a foundation for interdisciplinary dialogue aimed at building a sustainable AI ecosystem and offers a baseline guideline for researchers seeking to understand the environmental implications of AI across technical, social, and operational dimensions.
Source: Kim, M.-K., Yoo, T.-A., & Chung, J.-B. (2023). Toward Sustainable Generative AI: A Scoping Review of Carbon Footprint and Environmental Impacts Across Training and Inference Stages. Environmental Science & Technology.
This article is based on research available at arXiv.