AI may have the computational ability to create a blueprint on how to mitigate major environmental threats such as global warming, sea level rise, pollution and droughts, but there may be a problem in the solution: AI’s massive requirement for energy. By some estimates, it could account for up to 4% of global power demand by 2030.
Congress has been scrambling over the past year to address the moral and ethical issues surrounding the use of AI, and now it is also focusing on the environmental impact. On February 1, 2024, two U.S. senators and two U.S. Representatives introduced legislation to study AI’s environmental impacts and create a framework for developers to report those impacts. The bill is known as the Artificial Intelligence Environmental Impacts Act of 2024.
The co-authors of the bill are U.S. Senators Edward J. Markey (D-Mass.), the chair of Senate Environment and Public Works Subcommittee on Clean Air, Climate, and Nuclear Safety, and Martin Heinrich (D-N.M.), founder and co-chair of the senate AI Caucus and a member of the Senate Energy and Natural Resources Committee, and U.S. Representatives Anna Eshoo (D-CA), co-chair of the House AI Caucus, and Don Beyer (D-VA), vice-chair of the House AI Caucus.
“There is a Dickensian quality to the use of AI when it comes to our environment: It can make our planet better, and it can make our planet worse,” said Sen. Markey. “Our AI Environmental Impacts Act would set clear standards and voluntary reporting guidelines to measure AI’s impact on our environment. The development of the next generation of AI tools cannot come at the expense of the health of our planet.”
The act would “require the administrator of the Environmental Protection Agency to carry out a study on the environmental impacts of artificial intelligence, require the director of the National Institute of Standards and Technology to convene a consortium on such environmental impacts, and require the director to develop a voluntary reporting system for the reporting of the environmental impacts of artificial intelligence, and for other purposes.”
The bill details the possible harm caused by the rapid growth of computational power, which is doubling every ten months. “Rapid growth in data center infrastructure, including cooling systems and backup power equipment, supporting artificial intelligence and other computing-intensive technologies contributes to pollution, water consumption, and land use changes.”
The co-authors assert there will be annual increases in what is known as “e-waste” which is electronic waste such as outdated hardware. Power grid stress, water stress, created by the demand for water to cool down systems, and even local noise pollution would increase, they said.
Training some AI models, such as OpenAI’s GPT, required energy capable of powering an average American home for hundreds of years, according to The Stanford Institute for Human-Centered Artificial Intelligence. And another study carried out by the University of Massachusetts analyzed the “computational requirement for neural architecture search for machine translation and language modeling,” and concluded the energy requirement for that function would be equivalent to the carbon dioxide output of 125 round trip flights from New York to Beijing.
In a 25-page report titled “The AI Gambit: Leveraging artificial intelligence to combat climate change — opportunities, challenges, and recommendations,” (licensed under a Creative Commons Attribution 4.0 International License), four Oxford University-based AI systems researchers detailed the immense potential of AI to deliver solutions in the fight against climate change, but they also assessed the carbon footprint of AI research.
“The ability of AI to process enormous amounts of non-structured, multi-dimensional data using sophisticated optimization techniques is already facilitating the understanding of high-dimensional climate datasets and forecasting of future trends,” the authors stated.
AI has already been used to forecast climate trends such as global mean temperature changes and when the El Niño phenomenon might occur. It can predict how rainfall trends and water demands in a region can affect animal and human migration, according to the report.
“We find that the carbon footprint of AI research may be significant and highlight the need for more evidence concerning the trade-off between the GHG (greenhouse gas) emissions generated by AI research and the energy and resource efficiency gains that AI can offer.”
Contained in the Artificial Intelligence Environmental Impacts Act of 2024 are statements by environmental groups, tech agencies and university researchers, and they carry a theme that more research needs to be done on the potential environmental impact of AI to guide what action or guidelines should be implemented.
“The public deserves to know the very basics about how AI tools affect our material world,” said Caitriona Fitzgerald, Deputy Director at the Electronic Privacy Center. “By creating transparency around AI’s impact on our environment, the Artificial Intelligence Environmental Impacts Act of 2024 will give the public and policymakers a tool for pressuring technology companies to develop their products in line with our collective health and safety.”
The act would need to be approved by the Senate and the House before it can be signed into law by the President.