Generative AI is a common occurrence in most people’s everyday lives. It’s on social media, billboards, commercials, and even in music. Its usage is also increasing. Therefore, the impacts on the environment are also increasing.
According to a survey done by Malihe Alikhani, Ben Harris and Sanjay Patnaik, who are journalists for Brookings, roughly 40% of respondents said that their usage of AI increased over the course of one year.
When people think of AI, they are usually thinking of generative AI. Generative AI focuses on creating new content such as videos and images, but other traditional forms of AI usually have specific goals. Examples of traditional AI systems include NPCs in video games and virtual assistants such as Siri and Alexa.
On the environmental side, traditional AI systems make a smaller impact on the environment because, according to the University of Illinois, generative AI “requires larger datasets, while traditional AI can operate effectively with smaller datasets, depending on the complexity of the task and the model used.”
As the use of generative AI increases, so does the impact it makes on the environment. One of main environmental concerns involving generative AI is its alarming use of water. According to Miguel Yañez-Barnuevo, a project manager for the Environmental and Energy Study Institute, large data centers “can consume up to 5 million gallons per day, equivalent to the water use of a town populated by 10,000 to 50,000 people.”
Jennifer Calland, who formerly taught environmental science and currently teaches biology and the Project Lead the Way Classes here at Alder, attributes the overuse of water to a simple reason: computers get hot.
“Those huge data centers get really, really hot and need water for cooling,” says Calland. “…a one hundred word email generated by AI would take about a half a liter of water.”
Nathan Snedeker, who teaches environmental science and biology here at Alder, says that the issues these centers cause is especially concerning in areas that already have a lack of water, such as California or New Mexico.
“[The data centers] have to use a ton of water, which means they’re pulling from aquifers and natural resources that we could be using for our own water,” says Snedeker. “It’s putting stress on a system that is already stressed. It’s a problem [in Ohio], but it is not as significant as in California where there already is a water shortage.”
Snedeker also sees how that data centers negatively affect the land, specifically because of mining.
“The land damage that comes from it would be the mining for the rare metals that are needed for a lot of the hardware that goes into it,” says Snedeker.
Generative AI isn’t the only thing that causes negative environmental trends however. It is only one piece of the puzzle, as pointed out by Calland.
“Agriculture uses a ton more water than AI does,” says Calland. “And that’s been the case for as long as we’ve had mechanized agriculture. The amount of water it takes to raise a hundred pounds of beef is around five million liters of water. That’s a lot. And considering a cow is around a thousand pounds, that’s 50 million liters of water. …And all the agriculture we use for crops, some of it (around a third) is going to feed animals.. So we do use a lot of water for cooling the data centers, but we also use water for a lot of other things. So when it says that we use…two trillion cubic meters of water on data centers, it is a lot. But it’s like a drop in the bucket compared to some other things.”
According to Naomi Oreskes, a professor of the history of science at Harvard University, agriculture using so much water is one the reasons that some people choose a vegetarian or vegan diet.
However, since most people still eat meat, agriculture is still an essential part of both the economy and food production, and generative AI is completely optional to use for the average person.
In some cases, such as using generative AI for schoolwork, it is even discouraged. In a survey done on teachers by Lauraine Langreo from EducationWeek, only 27% of teachers think AI will have a positive impact on K-12 education over the next five years.
With this in mind, it isn’t too late to reverse the environmental impacts caused by AI. If the environment continues on this current track, Professor Bill McGuire from University College London (UCL) believes that the Earth will reach the point of no return by 2050. That means that both generative AI companies and the average person have 24 years to turn around and make a positive impact on the environment.
