Kleigh Carroll
As AI’s energy demands soar, it is increasingly necessary to find ways to green the technology before it's too late.
Data center. Wim Klerkx. CC BY-NC-ND 3.0.
Ever since generative AI took off, there have been concerns over its environmental consequences. Every time you enter a query into ChatGPT or DALL-E, both popular AI platforms, carbon dioxide is emitted and water is consumed. Researchers have found that a ChatGPT prompt consumes around five times more electricity than a typical web search. Over time, and as more and more people use the technology, the impact of this is substantial. There are ways, however, to mitigate the environmental impact of AI. But this necessitates the cooperation of large tech companies with greener goals and compliance with federal regulations. Whether AI will ultimately be a force for good in the face of climate change remains to be seen, and for now, the outlook remains uncertain at best.
Generative AI requires massive amounts of power to train and operate, which in turn greatly contributes to carbon dioxide emissions. Kenza Bryan, a climate reporter for the Financial Times, puts it this way: “The AI boom has taken off at an awkward time for the fight against climate change.” With global temperatures rising at alarming rates, faster than any scientist had previously predicted, AI is simultaneously contributing to a massive increase in power demand. This increase puts more pressure on the electric grids, systems that we have yet to decarbonize. “They’re still hugely reliant on fossil fuels,” Bryan says. This is why the power industry is one of the top sources of greenhouse gas emissions in the United States.
Beyond electricity demands, vast quantities of water are needed to cool the hardware used for training, deploying and fine-tuning generative AI models. Data centers, like mini factories, are temperature-controlled buildings that house computing infrastructure. The facilities consume as much as 50 times more energy per square foot than an average commercial office building. They are also consuming an increasing share of the energy in the states and communities in which they operate. According to Visual Capitalist, data centers account for 7.4% of power usage in Arizona and 11.4%in Oregon. The chilled water required to cool AI operating systems in these buildings puts strain on municipal water supplies and disrupts local ecosystems. It has been estimated that, for each kilowatt hour of energy a data center consumes, it would need two liters of water for cooling.
Furthermore, while every human on the planet will be affected by rising global temperatures and extreme weather events, the environmental impact of these data centers is unevenly distributed among certain communities, disproportionately affecting those who live close to them or in places already vulnerable to the effects of climate change. The strain on local freshwater resources can prolong or worsen drought in places like Arizona and Chile, and fossil fuel combustion to generate electricity results in local air pollution. And the environmental impact extends beyond just the energy needs of data centers. Indirectly, increased use of AI technology has created a greater demand for advanced computing hardware, which has negative effects associated with its manufacture and transport.
So, the question is, what can we do about it? The growth of AI appears inevitable. It continues to show up in more and more places in our lives, from helping us with our taxes to driving our cars. Does the responsibility to limit emissions and cut consumption costs rest primarily on the shoulders of the individual consumer? The government? How can we hold AI companies accountable for the environmental strain they create?
Leaders in the tech industry, like Sam Altman and Elon Musk, have claimed that any negative impacts we are witnessing are a temporary price to pay for the rewards we will surely reap in the future. Because AI has the power to solve climate change, they claim, the ends will justify the means. Musk’s company, xAI, came under fire in April for using methane gas turbines to help power a supercomputer in Memphis without the proper permits, contributing to serious air pollution in the area.
It is true that AI can contribute positively to the environment. According to Yuan Yao, Associate Professor of industrial ecology and sustainable systems at Yale, ironically, “AI can enhance energy efficiency and reduce energy usage,” while at the same time, its very use consumes exorbitant amounts of energy. AI can also assist in environmental monitoring, such as tracking air emissions. Yao says further, “We need transparent, robust methods to assess AI’s environmental impacts.” Otherwise, it is impossible to effectively address these issues. She, along with a team of researchers, is leading a project that focuses on reducing the carbon footprint of computing by 45% within the next decade.
Many companies developing AI technology, like Alphabet and OpenAI, are investing in renewable energy sources to power their computing. Given the inevitability of AI, there are also a number of things that companies can focus on to reduce their carbon footprint as they develop new technology. Larger models, for instance, require more energy to train and run. Companies can “prune and distill” their models, reducing the size and complexity while optimizing efficiency.
GET INVOLVED
According to Forbes, there are numerous strategies that tech and AI companies can adopt to mitigate environmental impact, such as switching to metadata-based training for models, tracking and reporting emissions and switching to renewable energy sources. You can read more about these approaches here. On the individual level, it is important to use these technologies judiciously and to support sustainable practices whenever possible. This looks like being mindful of your usage, opting for energy-efficient hardware and sharing your awareness with others.
Kleigh Carroll
Kleigh is a student at UC Berkeley studying Geography and Journalism. She hopes to integrate her skills in these fields in pursuit of a career in journalism. She is passionate about being outside, exploring, and writing in all of its forms.