Artificial intelligence (AI) is a powerful and rapidly growing technology, but it comes at a significant cost. A recent study published in arXiv examined the impact of AI not only on human health but also on the environment. The development of AI requires high energy consumption by data centres supporting the technology. This consumption results in a surge of harmful air pollutants.
According to the research, training a big AI model like Llama-3.1 produces as much pollution as driving a car over 10,000 times back and forth between Los Angeles and New York. These calculations are based on the energy consumption of data centres, many of which in the US still rely on fossil fuels. These fuels release harmful air pollutants, including gases that worsen air quality and fine particles that can penetrate deep into the lungs, causing issues like asthma and heart disease.
The health effects of these energy-intensive data centres are already felt by local communities. For example, in Data Centre Alley, located in Northern Virginia, the generators located there emit substantial amounts of pollutants and particulate matter contributing to respiratory issues. The study estimates that even if these generators operate at just 10% of their permitted emission levels, they could cause 14,000 cases of asthma annually, along with other health impacts.
The public cost of this is between $220 million and $300 million per year. Similarly, the total health costs could increase drastically to between $2 billion and $3 billion annually. The cost estimates are derived using a risk assessment tool created by the Environmental Protection Agency, which calculates the monetary value required to prevent adverse health outcomes such as premature deaths, asthma symptoms, heart attacks, and lost school or workdays.
Additionally, the impact of polluting emissions is not limited to local residents. Emissions can travel far, impacting states like Florida. And low-income areas such as parts of Ohio and West Virginia bear a disproportionate burden, facing health costs up to two hundred times higher per household than wealthier regions.
