While ChatGPT helps students’ study, its environmental footprint is going unnoticed.
Hundreds of students have turned to using services such as ChatGPT for studying purposes thanks to its efficacy in summarizing notes, generating practice tests, and several other abilities. While students especially tend to focus on its positive offerings, there is a hugely destructive side to the tech industry that often goes overlooked, and the recent rise of artificial intelligence (AI) is no exception.
There are several instrumental moving parts for the optimal functioning of AI, many of which are actively contributing to irreparable environmental destruction. At their core, AI programs are designed to retrieve and generate information from the web, with moderate variation between specific services.
In an interview with The Journal, Warren Mabee, director of Queen’s Institute for Energy and Environmental Policy, claims the obscene amount of energy consumed by AI isn’t because of the amount of data being stored, but rather the number of times the data is accessed. Mabee is the director of the School of Policy Studies and has a cross-appointment with the School of Environmental Studies.
AI programs access information from data servers—computers that hold physical connections to a database for storing, processing, and managing purposes. For large scale programs like ChatGPT, these data servers are found in data centres that use up valuable resources at rapid rates. Recent research at Goldman Sachs suggests by 2030, the power demand of data centres is likely to rise 160 per cent and drive a more than 100 per cent increase in carbon dioxide emissions.
“To train an AI, like Chat GPT, [these servers are] constantly running and every time you put a query into almost any search engine, it fires up those models, so they keep running,” Mabee said.
A prominent component to data server maintenance is cooling systems that come in a variety of forms but all of which work to minimize heat generated by these servers. Naturally, these consume copious amounts of water—as a valuable natural resource—to do so.
To put this absurd quantity of water consumption into perspective, a research report by The Washington Post and the University California, Riverside, found ChatGPT generating a 100-word e-mail is equivalent to consuming approximately 519 milliliters or just over one bottle of water.
Given the newness and energy source of AI, its net impact isn’t as environmentally damaging as more established industries such as forestry and mining, however, concern arises due to its exponentially high growth rate that sures anything else in recent history. AI becoming increasingly advanced and accurate is representative of data centre expansion and subsequent energy usage.
“The development of AI is like adding another country to the world or another couple of countries to the world in of total energy debt,” Mabee said.
AI data centres are currently responsible for about 3.5 per cent of global greenhouse gas emissions, a figure that’s continually rising. Yet the impacts of AI aren’t isolated to the amount of power required to run their data centers. According to Mabee, there’s high variability in the type of energy uses, be it clean energy or dirty energy sources that must be considered when assessing the impact of AI on the environment and its future sustainability.
Queen’s students have taken to using AI for study purposes, including Liza Rowatt, ArtSci ’27, who uses AI program ChatGPT on occasion.
“I feel like it’s such a helpful tool […] I usually copy my lecture slides and then [ChatGPT] explains it to me [more simply],” Rowatt said in an interview with The Journal.
However, since becoming aware of the detrimental effects of AI on the environment, Rowatt says she will absolutely act to change the way she utilizes the service by turning to alternative resources, both online and physical, when appropriate.
As for the future of AI, Mabee encourages those who have high hopes for the future of AI while being consciously aware of the precarious balance between harm and good.
“Maybe [AI] will help us solve big problems, which could easily offset the footprint that’s associated because we’re still only talking about maybe a couple percent of global electricity. And that’s a small price to pay for a valuable tool,” Mabee said.
While its accuracy and environmental impacts have been called into question, many students are still excessively relying on ChatGPT for basic tasks that may not be worth the toll these seemingly innocent query’s take on the environment. AI services can be used for good—yes, that may even be as a study tool—but, in accordance with Mabee’s view, alternatives and the overall global threat must be carefully considered.
Tags
All final editorial decisions are made by the Editor(s) in Chief and/or the Managing Editor. Authors should not be ed, targeted, or harassed under any circumstances. If you have any grievances with this article, please direct your comments to [email protected].
Zedrick Serson
$8/year/student to print blatant nonsense about genAI. Datacenters use closed-loop cooling so no water is “consumed”, a fact the Journal would have recognized if they did any research whatsoever instead of regurgitating liars.