Written by Alex Fisher

Thumbnail & Banner Photo by v2osk on Unsplash

Artificial Intelligence, commonly referred to as AI, is something we’ve covered on several occasions. From an article on the relationship between AI and Academic Integrity to discussions on the usefulness of AI in research and business management, and a critique of how generative AI could potentially harm your long-term academic performance, we’ve talked about AI in several contexts. One component of AI that is not often discussed, however, is its environmental footprint. With climate change and the climate crisis being some of the most pressing issues of our time, understanding the current and potential environmental impacts of the tools we use is exceedingly important.

The adoption of Artificial Intelligence—and especially generative AI, which are models that can create entirely new content—has increased dramatically over the past few years, both for business use and in people’s personal lives. The market share of AI has also grown significantly in recent years and is projected to reach over $400 billion by 2027—only three years from now. In fact, Forbes predicts an annual growth rate of 37.3% for industrial adoption of Artificial Intelligence. It’s clear that despite the struggle for businesses to make a profit off of AI, as well as growing ethical concerns surrounding the technology, companies are still investing huge amounts of money into its growth and the AI train is not slowing down.

Processing information requires energy, and a single AI tool such as ChatGPT can process tens of billions of petaFLOPS just while training. For reference, a rate of one petaFLOPS essentially refers to a computer’s ability to process one quadrillion simple mathematical operations every second—one quadrillion floating-point operations per second. As AI grows, the amount of information it processes is only going to continue to grow alongside it. After all, that data is necessary for training all forms of Artificial Intelligence, not just generative AI. Researchers at OpenAI, the company behind ChatGPT and other AI services, have estimated that the processing needs of AI have been doubling every three to four months since 2012. This leads to an enormous amount of power, solely used for training and operating Artificial Intelligence programs. While estimates vary, training just one AI model can produce over 300 tons of carbon dioxide waste. For comparison, a person’s car will produce, on average, around 63 tons of carbon dioxide waste over the entirety of its lifespan. Furthermore, training an Artificial Intelligence model is not a “one and done” situation. New data is being generated every day, and for AI to be up to date it will have to be retrained from time to time, massively increasing the amount of energy that even a single AI model would need to keep operating.

Artificial Intelligence can even outpace a small country when it comes to power usage. Data centres around the world already account for between 1% and 1.5% of worldwide energy usage, according to the International Energy Agency, and this number would only skyrocket if these data centres were suddenly powered by AI. If current predictions are accurate, then by 2027, an additional small country’s worth of electricity consumption could be added to the world power grid each year. This doesn’t even account for the cooling systems, either, which can add an additional ten to fifty percent power usage per data centre. It’s already estimated that by 2026, the infrastructure supporting AI models could use as much energy annually as the entire country of Japan, which is reportedly the fifth-highest electricity-consuming country on the planet as of 2023.

Of course, it also matters exactly where all this energy comes from. If it all came from renewable sources then there would be little to no concern beyond stressed power grids. However, Sarah Luccioni, a lead climate researcher at AI company Hugging Face, stated in an interview with Vox that the big data centres—and therefore the ones using the most energy—are almost always built in places that rely heavily on fossil fuels. The future of AI’s energy usage isn’t entirely bleak, however. By choosing to build new data centres in locations with sustainable grids, or where the ambient air temperature is much lower, this could drastically reduce the amount of carbon dioxide emissions these centres produce. Furthermore, by using more efficient algorithms and technology, or even being smart about scheduling when data centres are online to be more compatible with renewable energy sources, emissions produced by the infrastructure surrounding Artificial Intelligence could be massively reduced.

Energy consumption is not the only impact that Artificial Intelligence has on the environment, however. While little research has been done on the topic, the rate of e-waste production will likely increase as a result of AI becoming more widespread, especially as technology improves and servers are swapped out for faster, more efficient ones. E-waste, or electronic waste, is the material waste specifically produced by technology. It includes things such as old devices and computer parts, and is incredibly harmful to both people and the environment, releasing chemicals such as mercury and lead into the air and water. To make matters worse, e-waste is readily recyclable. However, recent studies have shown that less than 18% of global e-waste has been recycled, leaving millions of tons to continue polluting the planet.

There are yet more impacts that AI has on the environment: it has the possibility of evaporating already-limited fresh water, the potential to increase waste produced by industries which heavily utilize AI, could cause biodiversity loss through the overuse of pesticides and fertilisers, and there are ethical concerns over whether or not different AI models will be trained to value the protection of the natural environment. This all combines to worsen an already existing problem: environmental inequity. Essentially, by using valuable resources in regions where they are already sparse—such as fresh water in regions which face regular droughts—the people living in these regions will face harsher and harsher living conditions as time goes on, negatively impacting both them and the surrounding environment. Then, as AI infrastructure continues to grow, these inequities become even more widespread and serious.

There is a brilliant, shining light at the end of this seemingly endless abyss, however. When used properly, Artificial Intelligence has the potential to help us solve the climate crisis and maybe even end it completely. According to the United Nations, AI is already able to help with this most important of goals. In this article, which was released in November of 2023, the author explains how AI models such as IBM’s Deep Thunder could be used to help communities prepare in advance for significant weather events such as droughts or heavy rain seasons, while other models could help with the prevention of or recovery from weather-related or climate-related disasters. Other AI models could help to combat the issues that have been raised by critics of the technology. The emissions and waste produced by the fast fashion and food industries could be significantly reduced through supply-chain optimization and agricultural improvements, and the sustainability of these industries could be greatly improved. Furthermore, pollution from all sources could be greatly reduced by AI-optimising algorithms that would improve traffic and waste management systems in cities. Carbon dioxide emissions can be reduced across the board, too, by using AI models that can improve the efficiency of electrical grids—especially when it comes to sustainable or renewable sources.

While Artificial Intelligence is already helping in some areas, the potential of what it can do is far greater. From managing valuable resources to creating advanced models of the world climate, or even tracking wildlife populations and monitoring mining and deforestation, the possibilities of AI in helping to mitigate and fight climate change are nearly endless. Even reforestation and other environmental repairs can be done using fleets of AI-controlled drones. This goes to show that the only limitations of AI’s role in helping to keep our climate safe are the models that are produced and whether or not people follow their recommendations.

There’s no disputing that Artificial Intelligence has great potential. Whether this potential is used to worsen an already terrible situation—the climate crisis—or it is used to help in the fight against climate change is entirely dependent on how it is used. The use of AI up until now has already produced thousands of tons of carbon dioxide emissions and increased e-waste, and it is predicted to only make these problems worse in addition to placing additional stress on power grids and already-stressed environments. That does not have to be the case, however. If used correctly, AI could be a great boon in solving these very problems and more, helping to heal the Earth and prevent the climate crisis from getting much worse. It all comes down to the models that people create and the ways that these models are used, and the next generation of scientists, researchers, business owners, and engineers will be pivotal in deciding this.

There are several ways that a person could help the future of AI be more sustainable. By advocating for environmentally-friendly policies related to AI or reducing their digital carbon footprint by powering off devices and using less cloud storage, among other methods, any one person is able to contribute just a little bit more to the struggle against climate change. Taking it further, there is also the Department of Environmental Science at Saint Mary’s, which offers majors, honours, and minors for any students interested in learning more or even doing research in the field. Beyond that, it is important for a person to be mindful of their actions, and the environmental impact of their choices and habits.

What do you think? Let us know in the comments on our social media pages: do you believe Artificial Intelligence is going to make the climate crisis worse, or will it pave the path towards a greener future?

The Environmental Cost of Artificial Intelligence

FOLLOW US