Juliet Scallon

Most people use AI daily and are unaware of its devastating effect on the environment.

“Write an email to my English teacher,” “Generate an image of a cat in a tree,” “Write 100 words about the importance of community.” Since its creation in 2022, ChatGPT has become a standard practice in the lives of many. Why Google something when you can just use ChatGPT? Why write an entire paper and develop your own ideas when you can just use ChatGPT? It might seem like AI is completely harmless other than its effect on academic integrity, but who needs math or history in the real world anyway? Even if you believe this childish idea, the effects of generative AI go far beyond the classroom or workplace. According to Shaolei Ren from the University of California Riverside, the average AI interaction, such as generating an image or writing an email, consumes almost 16 ounces of water. This is around the same amount as the average water bottle. These “harmless queries” are literally sucking the world dry.

When you ask AI a question, it isn’t an actual brain that immediately knows the answer to your question. The AI database searches the entire internet for a comprehensive answer, making thousands of searches within seconds. These databases are held on physical servers, which need cooling agents to work because, like regular electronics, they can overheat too. Cooling agents require a lot of water, and in a full study that will be published later this year, Ren and her team estimate that between 5-50 questions with ChatGPT consume 16 ounces of water, which is the average size of a bottle of water. Now, multiply this one interaction by four billion (roughly how many queries ChatGPT is processing within one day). This equals around 500 million gallons of water that ChatGPT uses per day.

There is more to AI’s environmental impact than water scarcity. As MIT News explains, the databases used to train generative AI models like ChatGPT demand an absurd amount of energy. Although databases are crucial for all computing, generative AI is different because, according to Noman Bashir, Computing and Climate Impact Fellow at MIT, generative AI training consumes seven to eight times more energy than your average computing workload. When you ask AI a question, you aren’t just making one single search; the generative AI model is making dozens of searches and summaries to provide you with an answer. The convenience of AI causes many users to never even consider the possibility of long-lasting impacts.

AI’s environmental effect is a relatively new concept, as most of the studies are rather recent. Most people are aware of the effect AI has on people’s jobs, artistic integrity, and the future of the education system, but these facts are often ignored because of how convenient it is to use AI. We’ve all been living in willful ignorance surrounding the social consequences of AI and blissful ignorance surrounding the environmental consequences of AI. It’s time we open our eyes.

I’m not here to pretend like I haven’t fallen victim to the draw of generative AI, because I have. It is one thing to look at a statistic and understand the gravity of the situation, and it is something else entirely to do something about it. I know that even if I stop using ChatGPT, these databases are still going to run, and there will not be an immediate change in the world. That doesn’t mean that there isn’t a point in stopping because any difference possible is a difference worth making.

AI is terrible for the environment, and it’s time we all woke up and made the necessary change. Making a simple Google search is slightly less convenient than going to ChatGPT, but it makes a difference. It might not save us from the impending consequences of climate change or water scarcity, but it is a step in the right direction.