Further Reading: AI and the Environment — Climate, Resources, and Sustainability

These sources are organized by topic and annotated to help you decide what to read next. All sources are Tier 1 (published, peer-reviewed, or from established institutional outlets) or Tier 2 (reputable institutional reports and expert analyses).

AI's Environmental Footprint

Strubell, E., Ganesh, A., & McCallum, A. (2019). "Energy and Policy Considerations for Deep Learning in NLP." Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645–3650. The paper that brought AI's energy consumption into public consciousness. Strubell and colleagues estimated the carbon footprint of training large NLP models and proposed policy interventions. Clear writing, important findings, and still widely cited. Start here if you want the foundational argument. [Tier 1 — Peer-reviewed conference paper]

Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L., Rothchild, D., So, D., Texier, M., & Dean, J. (2021). "Carbon Emissions and Large Neural Network Training." arXiv preprint arXiv:2104.10350. A detailed analysis by Google researchers of the carbon emissions from training several large models, including T5, GPT-3, and others. Includes practical recommendations for reducing emissions. Offers a more nuanced picture than the headline-grabbing numbers alone. [Tier 1 — Published by researchers at Google with access to internal data]

Li, P., Yang, J., Islam, M. A., & Ren, S. (2023). "Making AI Less Thirsty: Uncovering and Addressing the Secret Water Footprint of AI Models." arXiv preprint arXiv:2304.03271. The UC Riverside study that estimated GPT-3's water consumption at approximately 700,000 liters. One of the first papers to systematically examine the water dimension of AI's environmental footprint. Accessible and important. [Tier 1 — University research, subsequently published]

International Energy Agency. (2024). "Electricity 2024: Analysis and Forecast to 2026." The IEA's authoritative analysis of global electricity consumption trends, including projections for data center energy demand. Essential context for understanding AI's place in the broader energy landscape. Available free on the IEA website. [Tier 1 — Intergovernmental organization]

Green AI and Efficiency

Schwartz, R., Dodge, J., Smith, N. A., & Etzioni, O. (2019). "Green AI." arXiv preprint arXiv:1907.10597. The manifesto for the Green AI movement, arguing for a shift from "Red AI" (pursuing accuracy at any computational cost) to "Green AI" (pursuing efficiency alongside performance). Proposes practical measures including reporting computational costs in publications. Accessible and persuasive. [Tier 1 — From the Allen Institute for AI]

Hinton, G., Vinyals, O., & Dean, J. (2015). "Distilling the Knowledge in a Neural Network." arXiv preprint arXiv:1503.02531. The foundational paper on knowledge distillation — training small models to approximate large ones. Technical but influential, and directly relevant to reducing the computational cost of AI deployment. [Tier 1 — Widely cited, foundational]

Henderson, P., Hu, J., Romoff, J., Brunskill, E., Jurafsky, D., & Pineau, J. (2020). "Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning." Journal of Machine Learning Research, 21(248), 1–43. A detailed proposal for standardizing the reporting of energy and carbon costs in machine learning research, including a framework and tools. Practical and actionable. [Tier 1 — Peer-reviewed journal article]

AI for Climate and Environment

Rolnick, D., Donti, P. L., Kaack, L. H., Kochanski, K., Lacoste, A., Sankaran, K., Ross, A. S., Milojevic-Dupont, N., Jaques, N., Waldman-Brown, A., Luccioni, A., Maharaj, T., Sherwin, E. D., Muber, S. K., Dubus, L., Zhong, Y., Lorber, R., Krawczuk, I., Balber, L., Greenhill, S., Gupta, A., Chen, G., & Bengio, Y. (2022). "Tackling Climate Change with Machine Learning." ACM Computing Surveys, 55(2), 1–96. A comprehensive survey of how machine learning can contribute to climate change mitigation and adaptation, covering energy, transportation, buildings, industry, agriculture, forestry, and climate science. This 96-page paper is the most thorough overview of the field. Read the introduction and the sections relevant to your interests. [Tier 1 — Peer-reviewed survey article]

Lam, R., Sanchez-Gonzalez, A., Willson, M., et al. (2023). "Learning skillful medium-range global weather forecasting." Science, 382(6677), 1416–1421. The paper describing Google DeepMind's GraphCast weather forecasting model, which outperformed the leading traditional model on most metrics. A concrete example of AI advancing climate-relevant science. Technical but the introduction and results sections are accessible. [Tier 1 — Peer-reviewed, published in Science]

Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina, A., Thau, D., Stehman, S. V., Goetz, S. J., Loveland, T. R., Kommareddy, A., Egorov, A., Chini, L., Justice, C. O., & Townshend, J. R. G. (2013). "High-Resolution Global Maps of 21st-Century Forest Cover Change." Science, 342(6160), 850–853. The foundational scientific paper underlying Global Forest Watch's deforestation monitoring. Hansen and colleagues used Landsat satellite data and machine learning to map global forest cover change at 30-meter resolution. A landmark achievement in environmental remote sensing. [Tier 1 — Peer-reviewed, published in Science]

Environmental Justice and AI

Crawford, K. (2021). Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence. Yale University Press. A wide-ranging examination of AI's material infrastructure — from lithium mines in Nevada to Amazon warehouses to data centers. Crawford argues that AI cannot be understood without understanding the physical systems that support it. The chapter on environmental costs is particularly relevant. Accessible and vividly written. [Tier 1 — Academic monograph from a leading AI researcher]

Hao, K. (2019). "Training a single AI model can emit as much carbon as five cars in their lifetimes." MIT Technology Review. A widely shared article summarizing the Strubell et al. findings for a general audience. A good entry point for someone who wants the key takeaways without reading the academic paper. [Tier 2 — Reputable technology journalism]

The Rebound Effect

Sorrell, S. (2009). "Jevons' Paradox revisited: The evidence for backfire." Energy Policy, 37(4), 1456–1469. A thorough academic review of the evidence for the rebound effect (Jevons paradox) across multiple sectors. Not AI-specific, but provides the theoretical and empirical foundation for understanding why efficiency alone does not solve environmental problems. [Tier 1 — Peer-reviewed]

Masanet, E., Shehabi, A., Lei, N., Smith, S., & Koomey, J. (2020). "Recalibrating global data center energy-use estimates." Science, 367(6481), 984–986. An important corrective to exaggerated claims about data center energy growth, arguing that efficiency improvements have so far kept total data center energy consumption lower than some projections suggested. An important nuance to the rebound effect discussion — efficiency gains have been significant, even if they have not prevented absolute growth. [Tier 1 — Peer-reviewed, published in Science]

Tools and Resources

CodeCarbon. https://codecarbon.io/ An open-source Python package that tracks the carbon emissions of computing. Useful for researchers and practitioners who want to measure and report their own AI workloads' environmental impact. [Tier 2 — Open-source tool]

ML CO2 Impact Calculator. https://mlco2.github.io/impact/ A web-based tool for estimating the carbon footprint of machine learning training runs, developed by the Mila AI institute. Simple to use and a good starting point for the exercises in this chapter. [Tier 2 — Academic tool]

Global Forest Watch. https://www.globalforestwatch.org/ The World Resources Institute's free platform for monitoring forest change worldwide. Explore it to see AI-powered environmental monitoring in action. [Tier 2 — Nonprofit platform]

Where to Go Next

• If you want the big picture on AI and climate, start with Rolnick et al. (2022) — it is comprehensive and well-organized.
• If you want to understand AI's material infrastructure, read Crawford (2021).
• If you want the foundational argument for green AI, read Schwartz et al. (2019).
• If you want to measure your own AI footprint, try CodeCarbon or the ML CO2 Impact Calculator.
• If you want to see AI environmental monitoring in action, explore Global Forest Watch.