AI’s Harm on the Environment

- Jamil Jon

According to a recent global market report, the AI market is currently worth 233.46 billion dollars in 2024. That’s a ton of money being spent on AI. This number will supposedly increase to 1,771.62 billion by 2032. In fact, 40% of companies now use generative AI & will continue to invest in it. Generative AI produces text, audio, and images and provides personalized user experience. Some of AI’s more specific aspects include machine learning (ML) and natural language processing (NLP), both of which are used in popular AI tools like chatbots. However, these prey upon one specific resource: the environment. 

A single computer needed to run AI requires 800 kilograms of raw materials. Additionally, the microchips of these computers are said to have been mined through environmentally-destructive means. But that’s only the beginning. 

AI data centers gain energy by burning fossil fuels, which releases large amounts of carbon dioxide (CO2) into the atmosphere, which then traps heat within the atmosphere and contributes to global warming, ultimately resulting in a higher carbon footprint. But with how big it’s becoming, AI’s footprint won’t be diminishing anytime soon. With bigger demand come greater costs. Ireland, for example, reported that 32% of its energy would be dedicated to fueling data centers by 2026. 

In a similar vein, a bigger demand for AI also urges further innovation. With that, more complex datasets and AI models require more energy. To put into perspective, a 176-billion parameter language model would require 50.5 tonnes of carbon dioxide. 

Other applications of AI, such as drones, negatively impact animals as well. In a 2014 study, researchers Mulero-Pázmány et al., reported that, while drones are useful for conservation, animals may perceive them as threats, causing stress or displacement. 

However, animals are not the only ones affected. When it’s not producing waste such as mercury and lead, AI is depleting water. Like fossil fuel, AI uses water as its lifeline. AI uses water for data center construction, and once operational, to cool its electrical components. According to one estimate, AI-related infrastructure might require six times more water than Denmark, which has six million people. 

A solution to maintain these data centers might be to use alternatives like air-cooled chillers. But this only overloads the power grid in certain areas, making electricity more expensive for locals. It poses yet another conundrum, but it also suggests that others will bear a greater share of AI's environmental burdens since different areas have different levels of wealth and resources.

For example, it would be cumbersome to put such facilities in arid climates like Arizona. Similarly, these facilities also possess a large regional disparity for carbon efficiency. For context, only 4% of Singapore’s Google data centers are carbon free, in contrast with Finland’s 94%. Taking these differences into account makes the task of regulating AI even trickier.

It becomes more confounding when we consider how secretive the AI industry is. Alokya Kanungo writes, “The shadowy realm of AI development and utilisation breeds a lack of transparency and accountability regarding its environmental impact.” Moreover,  certain companies prioritize money & competition over AI’s negative environmental effects. The methods used to train these AI models lack the transparency that our laws and procedures should have. “A responsible approach,” Kanugo says, “will be made possible by working toward greater accountability.” 

Kanungo isn’t the first to address this. Over 190 UNESCO countries have already  published non-binding principles on the ethical use of AI, including AI’s  environmental effects. These principles state that the deployment and usage of AI systems should align with the UN’s Sustainability Goals. It also addresses the reduction of AI’s carbon footprint. The question about whether these measures are actually being implemented is still skewed. On one hand, governments still lack the accountability these measures require. “The lack of environmental guardrails is no less dangerous than the lack of other AI-related safeguards,” said UNEP Chief Digital Officer Golestan (Sally) Radwan “Sally” Radwan. 

This makes it difficult for users to grapple with the convenience AI offers, given its environmental footprint & shady machinations. TikTok users, namely @nikitadumptruck, also known as Bimbo University, are now sharing knowledge on the impact that chatbots like ChatGPT have on the environment. Public consensus remains divided though, with some users justifying their dependence on the chatbot while others have completely renounced it.

This issue however is not as simple as one side conceding defeat. AI’s momentum is inevitable. Until a compromise is set, we should continue looking for answers, while regulating the control AI has on our own lives.