Study on the Full Life-Cycle Environmental Impact and Systematic Optimization of Large Models
While large-scale artificial intelligence models drive technological innovation, their entire lifecycle is accompanied by significant energy consumption and carbon emissions, posing a real risk of sliding from the vision of “Green AI” toward the reality of a “high-carbon technology”. This paper constructs a three-tier analytical framework encompassing the training, inference, and hardware infrastructure stages to systematically assess the environmental impact of large models. Furthermore, it analyzes the underlying causes of the high-carbon dilemma from three dimensions: technological dependency, economic incentives, and governance gaps. The research indicates that a single technological optimization is insufficient for achieving sustainable transformation. It is necessary to collaboratively advance algorithm lightweighting, system scheduling optimization, industry standard establishment, and an ethical paradigm shift to construct a multi-level, systematic development pathway for Green AI. Only through the co-evolution of technology, policy, and culture can artificial intelligence be genuinely guided to become an enabler for addressing environmental challenges.
References
- Strubell, E., Ganesh, A., & McCallum, A. (2019, July). Energy and policy considerations for deep learning in NLP. In Proceedings of the 57th annual meeting of the association for computational linguistics (pp. 3645-3650).
- Lacoste, A., Luccioni, A., Schmidt, V., & Dandres, T. (2019). Quantifying the carbon emissions of machine learning. arXiv preprint arXiv:1910.09700.
- Osondu Joshua. (2025). Red AI vs. Green AI in Education: How Educational Institutions and Students Can Lead Environmentally Sustainable Artificial Intelligence Practices. Doi: 10.13140/RG.2.2.27929.12644.
- Fedus, W., Zoph, B., & Shazeer, N. (2022). Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. Journal of Machine Learning Research, 23(120), 1-39.
- Katharopoulos, A., Vyas, A., Pappas, N., & Fleuret, F. (2020, November). Transformers are rnns: Fast autoregressive transformers with linear attention. In International conference on machine learning (pp. 5156-5165). PMLR.
- Mesarčík, Matúš & Solarova, Sara & Podroužek, Juraj & Bielikova, Maria. (2022). Stance on The Proposal for a Regulation Laying Down Harmonised Rules on Artificial Intelligence – Artificial Intelligence Act. Doi: 10.31235/osf.io/yzfg8.
- Yigitcanlar, T. (2021). Greening the artificial intelligence for a sustainable planet: An editorial commentary. Sustainability, 13(24), 13508.