The Elephant in the Server Room: Unpacking AI's Energy Consumption - and how you can help!

Whenever we recieve a task to complete in school, I see every day, almost all of my peers and friends using ChatGPT to complete it - most of us, without hesitation, turn to AI models as our go-to solution for just about everything.

AI is extremely helfpul and makes life so much easier, and I do believe we must use it, but as a tool rather than a medium of work, to truly let our creativity, problem-solving, and the brilliance of the human mind flourish. Although, this article is not about when and what for we can use AI, but rather the environmental impact of even just a single ChatGPT prompt, and how we can modify our usage to mitigate that.

In our modern lives, we often find AI at the forefront of everything we do: School work, asking queries, writing articles (I promise you this one is entirely human-made ;D ). Although, the numbers of how much energy, water and Earth's resources large tech companies and AI models actually consume are truly unreal!

According to RW Digital (and Google itself), a single search requires 0.0003 kWh of energy, which doesn't seem like much.

However, depending on the amount of searches Google processes daily, estimated to be around 8.5 billion, means that the total energy produced would be around 2.55 gWh (Gigawatt-hours), that could power more than 250,000 Indian homes in a day.

Microsoft, another leading tech company, skyrocketed in water consumption in 2022. Their water consumption spike from 2021 to 2022 itself was nearly 1.7 billion gallons, which, to put in context, is almost 2500 full Olympic size swimming pools. That is an absolutely enourmous amount of water, used to provide power and cooling to data centers and AI servers.

Now, we come to Open AI's ChatGPT, the world's most popular interactive artificial intelligence platform, whose resource consumptions dwarf even those of Google and Microsoft. Here are a few numbers related to common daily activities and locations showcasing the energy usage of ChatGPT. According to RW Digital, the amount of energy ChatGPT is projected to use annually is enough to:

• Fully charge 3.13 million electric vehicles

• Power approximately 22,000 U.S. homes for an entire year.

• Run the entire country of Finland or Belgium for a day.

• Charge almost 48 million iPhone 15s every day for a whole year!

Since ChatGPT prompts daily consume enough energy daily to power the Empire State Building for a year and a half, it puts a great strain our planet's resources, and energy consumption at this level significantly contributes to climate change.

So what can we do at an individual level to reduce this energy exhaustion from the prompts that we input in ChatGPT every day?

Firstly, we can try our best and avoid unecessary or repeated prompts to ChatGPT. We must send it a prompt only when it is absolutely necessary and can help us in our work, and not for unrelated things. For example, even a simple "hi" or "thank you" to ChatGPT consumes so much water and energy, we can refine our prompts to maximize the output ChatGPT gives us in one go, minimizing our inputs.

Secondly, we can use smaller models or talk to experts for certain solutions. Instead of always using GPT-4o, which is a highly advanced model requiring a lot of energy, we could use GPT-4o mini, GPT-1o mini, or even Google Search to tackle certain problems, as they do give advanced responses as well. Google Search takes up about 1/10th-1/15th of the energy ChatGPT does, so we could prioritize using other tools or AI models to gain our desired data. Furthermore, moving slightly away from AI, we can even talk to experts around us for certain topics, instead of searching up solutions on ChatGPT. For example, if we are working on a project related to education for all, we could have impactful discussions with educators around us, to gain multiple perspectives, build communication and interpersonal skills, and save a lot of energy.

Lastly, each of us must spread awareness about the environmental impacts and energy usage of AI, and encourage people around us in our communities to realize how they must change their usage to better conserve environmental resources. Looking at these numbers, it is truly shocking and a wake-up call to motivate people to change their habits for a greener collective future. We must also support tech companies who use renewable forms of energy and are transparent with their usage, and provide collective encouragement for them to continue their sustainability journey.

So, let's work on debugging our usage of AI, promoting energy conservation, reducing our dependency on technology, and contributing to greener, smarter and cleaner server rooms- across the world.