I hate A.I.
I don’t use “hate” flippantly — semantics matter, but I think the word is appropriate here.
As I frequently discuss in most of my articles, I care passionately about creative thinking, unique writing, and creating a distinct voice for yourself within academic and recreational time. As you can imagine, this viewpoint vehemently opposes the use of artificial intelligence, or more specifically, Large Language Models (LLMs) like ChatGPT or Google Gemini.
This, however, isn’t about my potent beliefs that using A.I. as a crutch is rotting our brains, teaching computers how to speak for us, and damaging our abilities to think critically. If you want to pay your college tuition in today’s economy just to have a robot do all the work for you, who am I to judge?
But, if you can’t be persuaded to stop using A.I. by an educational argument, I raise you an environmental one. Bear with me, as I appeal to your environmentalist side.
Who doesn’t love the beach? I was raised in San Diego and I go to school in Santa Barbara, so I know I do. Who doesn’t love a nice cold glass of water, or a hot shower after a long day? While we’re at it, why not throw in a heater in the wintertime, light switches in your apartment, all your favorite frozen Trader Joe’s meals, a freshly charged computer, or the public bus you rely on to get around campus?
All these things — water supplies, electrical outputs, and various minerals — are threatened by the use of A.I.
When I say “A.I.”, I’m referring to the massive data centers and microchips used to store the material needed to process requests made when you ask Chat to “write me a thesis statement.” To run these immense amounts of calculations quickly, Graphical Processing Units are employed, which originally were used to render computer graphics. These chips are more efficient when run in large “cloud data centers.” As written in an article by Rasheed Ahmad, Ph.D., P.E., M.ASCE for a Civil Engineering publication, these facilities “contain the servers, routers, switches, storage systems, and other equipment that run continuously to meet what seems to be an ever-expanding demand for cloud-based computing, digital storage, artificial intelligence, digital gaming, streaming music and movies, data analytics, and other services.”
Already, one search query to an A.I. model uses 10x more energy than a standard search. Picture that singular prompt in the context of a massive data center, some larger than Central Park (843 acres, for reference). Today, according to a New York Times video examining the environmental impact of A.I., Virginia data centers consume over a quarter of the state’s electricity, with regular homeowner’s electricity bills expected to nearly double in the next seven to ten years to power said centers.
As A.I. usage becomes more integrated into mainstream use, the International Energy Agency projects that data centers’ electricity consumption in 2026 will be double that of 2022 — 1,000 terawatt-hours. By 2030, data centers alone are projected to match the entirety of India’s electricity usage, and by 2040, according to a journal article, “Artificial intelligence and climate change: ethical issues” by Anders Nordgren, emissions from the Information and Communications Technology industry as a whole will reach 14% of global emissions, with the majority coming from infrastructure like data centres.
Let’s take a look at some water stats next.
Water is used in data centers to both generate the electricity needed to run the technology, and as coolant to dissipate heat produced by the servers. This water isn’t from our almost indefinite supply of ocean water, nor even recycled runoff; instead, Shaolei Ren, an associate professor of electrical and computer engineering at UC Riverside, describes how this water must be free from bacteria and impurities that could damage the delicate electronics. In other words, data centers fight “for the same water people drink, cook, and wash with.”
On a small scale, this means that a typical session of Q&A with GPT-3 consumes a half-liter of fresh water. For a large Meta data center, upwards of 500,000 gallons of water are drained in a day. According to a Georgia city water request, one data center company demanded nine million gallons of water a day — that’s the equivalent of 30,000 households. In 2022, Google’s data centers alone consumed 5 billion gallons of fresh water for cooling, a jump in 20% from the year before. These issues are felt domestically and internationally, from Texas and Arizona and Louisiana and Georgia and Colorado, to Chile and Uruguay and the United Arab Emirates.
Not only are data centers abusing that vital 1% of fresh water available for the processing of data, but this water is being directly taken from people in need. These data centers tap into reservoirs supplying drinking water, well-water designated for specific communities as seen in one example from Newton County, and “directly or indirectly draws water from 90% of U.S. watersheds,” as seen in one U.S. environmental footprint study. In the words of one Newton County water authority Mike Hopkins, “What the data centers don’t understand is that they’re taking up the community wealth. We just don’t have the water.”
Moreover, pollution is a vital concern of the A.I. industry.
A recent study conducted by University of Massachusetts researchers revealed that simply training an A.I. model to efficacy can generate 626,000 pounds of carbon dioxide. This equals 300 round-trip flights between New York and San Francisco, and is five times the lifetime-emissions of the average car. To the concern of residents living around these data centers, Elon Musk built dozens of methane gas generators to support the center of the world’s largest supercomputer, the Colossus, releasing poisonous nitrous oxide and formaldehyde into the air, much to the outrage of surrounding communities.
No matter how you look at it — whether from an electrical, hydro, or pollution perspective — A.I. is the common denominator of physically scarring our Earth.
So, what can we do? While most of these emissions are as a result of a larger-scaled capitalistic system, there’s still some individual actions you can take to mitigate your individual environmental footprint.
For starters, call your representatives and implore them to codify visibility and regulations around the expenditures of natural resources for data centers.
On a day to day basis, I encourage you to be more mindful and intentional with your use of A.I. If you can’t quit altogether, consider stopping asking repetitive queries, and making Chat, Gemini, and Snap A.I. flip a coin for you, or write a script on how to text that boy from your section about going to your date party. And for the love of god, stop saying “please” and “thank you” every time you type something — the OpenAI CEO Sam Altman admitted himself that people’s manners are costing “tens of millions of dollars” from electrical expenses.
As they say, “it’s cool to care.” It may not seem like much, but your help contributes to a happier, healthier, and greener Earth. Love your Mother!