Artificial Intelligence, commonly known as AI, has had a huge impact on our society recently, specifically Generative AI (GenAI), which creates original text, images, videos, etc. Whether it is big name brands using it to make their commercials, or your mom sending you a video of bunnies jumping on a trampoline that she found on Instagram and thought was real, AI has fully integrated itself into our everyday lives. There has been discourse around AI actresses and singers and where the line should be drawn on usage of AI. Students are often culprits of using AI platforms like ChatGPT to complete assignments or explain topics they do not understand, and while this sounds like harmless dishonesty, AI has a large negative impact on the environment and is not as sustainable as you may think it is.
First, to understand how GenAI impacts the environment, we have to understand how it works. Companies like OpenAI (who own ChatGPT) have multiple buildings called “data centers” which are basically houses for servers, data storage drives, and other examples of computing infrastructure. These data centers are temperature controlled and require massive amounts of energy in order to operate. They are used to train the GenAI and run models in order to research and advance it (MIT News).
Data centers are not new— in fact, the first was built in 1945. Amazon has data centers across the world (MIT News). The difference in GenAI data centers though, is that it requires large and dense amounts of power and energy compared to other computing workloads. The rise of data centers is mostly due to the popularity and demand for GenAI, increasing North America’s energy consumption exponentially. Luckily, most of these centers are powered through renewable materials, such as natural gas, and nuclear energy is expected to play a large role in powering data centers in the coming decades. Additionally, data centers account for less than 1% of global CO2 emissions (International Energy Agency). Unfortunately, that does not take away from the massive amounts of energy they use in order to operate.Â
If you have heard arguments against GenAI, you have probably heard people say that it uses massive amounts of water, and if you are like me, you were probably extremely confused. However, these people are correct; GenAI uses large amounts of water at data centers. Water is used in order to keep the servers and computing equipment cool. Not every data center uses water to cool their equipment. The ones that do not use “Air Cooling”. The catch is that it is not as energy efficient compared to water based cooling systems. Studies have found that approximately 75-90% of data centers use water in some way (Florida Water and Pollution Control Operators Association). Additionally, most of the water being used is potable water— meaning it is safe to drink. A medium sized data center can consume up to 110 million gallons of water per year. For reference, the average household uses approximately 300 gallons per day, or 110,000 per year. So, the medium sized data center consumes the equivalent of 1000 households in one year (Environmental and Energy Study Institute).Â
As the United States reaches what researchers are calling a “tipping point” in water security, they conclude that the United States is in a “Peak Water Security Crisis”, and the demand for clean drinking water is higher than ever (Texas A&M). Potable water is becoming harder and harder to obtain for many Americans, and the immense water usage done by these data centers is by no means helping.
Now, I am not going to sit here and pretend I am the perfect writer. Prior to having the knowledge I now have, I have used GenAI brands like ChatGPT to help explain concepts that I had trouble with, specifically when it came to Statistics. ChatGPT is an easy way to get a quick understanding of how to complete a complicated problem. But, there are many other alternatives that not only go more in depth than GenAI, but also explain topics on a level that I can comprehend and apply to the test of the unit. I urge other students who are in the same boat to open youtube before going to ChatGPT. There are so many online resources and videos that can explain numerous topics in under five minutes. Think about your schooling before GenAI became a major component, and how you would have solved a problem then. I can guarantee that the results will not only serve you better, but will be 100 times more rewarding. (Also: GenAI has been proven to give false information time and time again, so if you’re using it to complete assignments, or complete tasks for you: there’s a high chance it’s giving you wrong or misleading information.)
AI does have great promise when it comes to revolutionizing scientific research, and maximizing possibilities across fields. But AI should be left to those disciplines. We should not be using AI to complete small tasks and waste precious resources when it could be used to strategize better global climate practice, or aid in breakthroughs with surgical technology. We as a society need to do better, and limit AI usage to research where it really counts. Making bunnies jump on a trampoline for a million likes, isn’t revolutionizing the world, like AI has the power to do.