This semester, the syllabi for every single one of my classes mention generative artificial intelligence and sites such as ChatGPT or Copilot. It makes sense. Professors must craft some sort of policy on AI, whether in favor or opposed. According to a survey from the Digital Education Council, 86 percent of students use generative AI for their studies. But for the sake of our environment, we cannot afford to use AI in its current state.
The energy consumption associated with ChatGPT is colossal. For example, the energy consumption for training one large Language model such as ChatGPT results in a carbon footprint of 300,000 kg of CO2 emissions, comparable to 125 round-trip flights between Beijing and New York.
Still, the larger environmental problem associated with AI is the depletion of an essential natural resource: freshwater. The latest United Nations Environmental Report found that nearly two-thirds of the world’s population experiences water scarcity for at least one month a year; this number is expected to grow. Simultaneously, the high-energy data centers that power AI sites use enormous amounts of water for cooling processes. AI’s projected water usage is on track to hit 6.6 billion meters cubed by 2027, a concerning figure. For comparison, that is the same amount of water used by Cuba, a country of 11 million people, each year.
Due to both carbon emissions and water consumption, the role of generative AI has been overwhelmingly negative for our environment.
Companies such as OpenAI and Microsoft have been vague and evasive when pressed on efforts to improve AI’s sustainability. Brad Smith, president of Microsoft, has proudly reported that “By 2030 we will be water positive, meaning we will replenish more water than we use.”
However, Microsoft has failed to provide extensive details about how this “water-positive” plan would look and how it will work specifically to mitigate the horrific consequences of generative AI. Their 2023 environmental report found that from 2021 to 2022, Microsoft’s water consumption increased by 34 percent, following the introduction of generative AI into the company. Companies, especially companies that turn a massive profit from AI (Microsoft gains roughly $10 billion in profit annually from AI), should not be trusted to correct their behaviors for the sake of the well-being of both our planet and the people living in it.
It is our responsibility as consumers to demand better. Those who have a vested interest in the planet and the climate’s continued health have a moral imperative to oppose generative AI, at least in its current unsustainable form. We can, and must, advocate for increased sustainability practices when it comes to AI and increased transparency regarding energy sources as well as overall resource consumption.
There are options to make AI a more sustainable practice. Shifting AI operations to energy-efficient data centers that are more transparent about energy usage and utilize green energy sources is one such option. We should also advocate for prioritizing targeted, field-specific AI models rather than expansive, generalist AI platforms such as ChatGPT, which would increase overall efficiency and minimize environmental impact.
In the meantime though, we must not use AI. The large-scale usage of generative AI is what makes it so perilous for the environment. An effective method of opposition is a large-scale boycott of sites such as ChatGPT and Gemini.
This is not to say that it is the individual’s responsibility to end climate change (just 100 companies are responsible for 71 percent of emissions related to fossil fuel production) but it is to say that consumers have the power to fight anthropogenic climate change and we must utilize that power. ChatGPT is unique in that its environmental destruction is driven by consumers.
ChatGPT is primarily a consumer product. The bulk of usage comes directly from individuals, which makes it uniquely vulnerable to consumer boycotts. Since its carbon and freshwater consumption stems largely from single individuals, they have the power to enact change through a targeted boycott of generative AI.
As the AP reports, a ChatGPT session involving 5-50 prompts consumes as much 500 milliliters, equivalent to a bottle of water. Students have completed school work before and can continue to without using generative AI.
The level of convenience associated with generative AI simply is not worth it. Signing out of ChatGPT is a small but effective step one can take to reduce one’s carbon footprint and put financial pressure on these companies to operate more sustainably.
Alex Benach PO ’28 is from Washington D.C. He remembers the days when Northern Virginia was not considered the data center capital of the world. He would like to go back to that time.
Facebook Comments