CAMBRIDGE, Mass.—Much has been written by journalists, scholars, policy makers and even the UN, about how artificial intelligence could be used to help combat climate change. But a new article from the MIT Sloan Management Review, warns that AI’s potential contributions to solving the climate crisis could be overshadowed by its enormous energy use and carbon emissions, e-waste, and water use.
The article points out a number of areas where the expanded user of AI technologies could have serious environmental consequences. For example, a single ChatGPT query can generate 100 times more carbon than a regular Google search. In addition, training OpenAI’s GPT-3 model is estimated to have used the equivalent of 120 average U.S. households’ annual energy consumption; an average data center, critical infrastructure for AI, consumes the equivalent of heating 50,000 homes yearly; and Microsoft and Google’s use of water to cool data centers has grown by millions of gallons as they develop cutting-edge AI technologies, the research found.
“Several factors contribute to the carbon footprint of AI systems throughout their life cycles,” explains Niklas Sundberg, author of the article, “Tackling AI’s Climate Change Problem.” “The AI industry must adopt practices that emphasize sustainability, make sustainability central to its AI ethics guidelines, and actively seek opportunities to reduce the environmental footprint of AI technologies.”
Sundberg, a board member of SustainableIT.org and chief digital officer at Kuehne+Nagel, a global transport and logistics company, urges that transparency is critical to addressing the potential problems that could be created by the AI industry. Reliable measurements of new models’ energy use and carbon emissions must be published to raise awareness and encourage AI developers to compete on model sustainability.
In addition, end users must also be aware of the factors that contribute to the environmental impacts of these tools to guide their use of them and add sustainability to the list of criteria they use to evaluate vendors and products.
In the article, Sundberg also details the