Associate Professor Yuan Yao is part of a National Science Foundation-led research initiative aimed at reducing the carbon footprint of computing. She recently spoke with YSE News about some of the environmental costs and benefits of AI.
Artificial intelligence (AI) is powered by massive amounts of energy, and much of the energy is from the burning of fossil fuels, which is the biggest contributor to global warming. The International Energy Agency estimates that by 2026, electricity consumption by data centers, cryptocurrency, and artificial intelligence could reach 4% of annual global energy usage — roughly equal to the amount of electricity used by the entire country of Japan.
Yuan Yao, associate professor of industrial ecology and sustainable systems at the Yale School of the Environment, is part of a multi-institutional National Science Foundation (NSF)-led research initiative aimed at reducing the carbon footprint of computing by 45% within the next decade. She spoke with YSE News about the environmental opportunities and harms posed by AI.
Powering computing systems for AI requires energy, such as electricity. Electricity generation emits pollutants, especially in regions where fossil fuels dominate electricity generation. Energy use impacts the environment through fossil fuel combustion that generates greenhouse gas emissions, contributing to climate change. Fossil fuel combustion also releases pollutants that cause air and water pollution, respiratory issues, and acid rain. Extraction of fossil fuels and energy infrastructure can disrupt ecosystems and contribute to environmental degradation. Transitioning to renewable energy sources, such as solar and wind, and adopting energy-efficient practices can mitigate these negative impacts.
Besides its energy usage, AI needs hardware devices. The production, transport, maintenance, and disposal of these hardware components, such as servers and data centers, require additional energy use and substantial materials and natural resources, for example, cobalt, silicon, gold, and many other metals. The mining and production of these metals used in AI hardware can lead to soil erosion and pollution. Many electronics are not properly recycled, leading to electronic waste that can cause further pollution. The materials used in these devices can contaminate soil and water when not disposed of correctly.
We need transparent, robust methods to assess AI’s environmental impacts. Without accurate quantification, it is impossible to mitigate and address these challenges effectively.”
The applications of AI can bring environmental benefits. A few years ago, we published a paper that examined the benefits of AI applications in the chemical industry. AI can enhance energy efficiency and reduce energy usage, and it assists in environmental monitoring and management, such as tracking air emissions. Moreover, AI supports process and supply chain optimization to minimize environmental impacts.
Additionally, my research group has been utilizing AI to support life cycle assessment (LCA), a standardized method to evaluate the environmental impacts of a product’s entire life cycle. AI enables us to assess these impacts for products made from diverse biomass species, a task that is quite time-consuming using traditional methods.
The project, funded by a $12-million grant from the U.S. National Science Foundation, focuses on reducing the carbon footprint of computing by 45% within the next decade. It will pursue three main goals: create standardized protocols to measure and report carbon costs over the lifetime of computing devices; develop ways to reduce the carbon footprint of computing; and explore ways to reduce the carbon emissions of fast-growing applications, such as artificial intelligence and virtual reality systems. I will lead efforts on carbon modeling, accounting, and validation of semiconductors and computer systems, covering both embodied and operational emissions.
We need transparent, robust methods to assess AI’s environmental impacts. Without accurate quantification, it is impossible to mitigate and address these challenges effectively. Our NSF-sponsored Carbon Connect project aims to tackle this issue by developing transparent carbon accounting tools. Specifically, my lab focuses on developing robust life cycle assessment methods tailored for computing systems. Doing this will allow us to perform holistic environmental impact assessments of AI and identify potential solutions to mitigate these impacts.
