As a bilingual content designer in civic tech and volunteer translator for groups like Stewardship Partners and the Ocean Blue Project, I’m often asked to share actionable knowledge in clear, accessible ways with the goal of improving life for others and protecting the environment. It’s been interesting (and even hopeful) to explore how these skills can intersect with our use of artificial intelligence (AI).
The rise of generative AI has changed how we create and access content. But beneath every summary and AI-assisted line of code, there are powerful data centers endlessly churning and processing information.
As a technologist in this moment, using AI feels inevitable, but I can still be intentional about it. For that, I’ve been looking for ways to use these tools in less resource-intensive ways to be a more environmentally-conscious AI consumer.
AI Energy Consumption
AI technology runs on data centers. In turn, data centers need energy to power their systems and a way (oftentimes water) to cool down their servers. A single large data center can consume up to 5 million gallons of water a day—enough to supply a small town—straining water resources in already-stressed communities.
America’s largest power grid is already struggling to meet energy demands driven by AI expansion, and big companies like Google have rolled back their carbon-neutral promises to keep up with increasing AI demand.
We can’t clearly see the resources data centers might use for our interactions with AI because there is no set standard for how they are built, powered, or cooled. But what we can do is consider the processing power we’re asking of AI in each of our prompts.
The more “thinking” AI has to do, the more resources the data center needs to deliver. And here’s the key: individually, our use of AI might seem insignificant, but there are billions of us prompting AI on a daily basis. This is where it all adds up, including the small inefficiencies.
Google Data Center, Council Bluffs Iowa By Chad Davis Photography, is licensed under Creative Commons Attribution 2.0 Generic license
There are about 6,000 data centers worldwide today. Even if only half of them use water to cool their servers, that’s still millions of gallons per day. An additional problem is that not all water used for this process can be reused. Some is lost to evaporation, and some cannot be treated for human use.
Still, there are things we as AI consumers can do to lessen the impact.
Plain Language Can Make AI Prompts More Sustainable
Plain language makes communication among humans easier to understand, and it also makes prompts lighter for machines to process. That’s because clear, straightforward prompts are usually shorter and more focused. These lighter, quicker interactions between humans and AI require less processing power and drive less energy and water consumption from data centers.
Responsible prompt engineering practices can elicit better answers from AI while also using less energy. It’s truly a win-win.
Tips for Building More Sustainable AI Prompts
1. Be clear and specific
The prompt on the left is the kind of input we might use in a search engine. But since AI can handle more complex requests, we can ask for more clarity upfront. From a resource point of view, the less back and forth you need to do to clarify a request, the fewer resources AI will use to process it.
2. Provide appropriate context
If we start a conversation from the halfway point of a story, others have to ask questions to catch up. Generative AI tools operate in the same way.
Providing context matters, and providing the right context matters even more. The prompt on the left sets the stage very broadly. AI will provide a very broad answer that will need refining until we get the information we need.
The prompt on the right will yield a more focused, precise answer that requires less follow-up and consumes less energy. To get there, work backwards. Establishing a framework for the information you’re seeking will help inform your questions and the context you need to provide.
3. Add creative boundaries when brainstorming
Kicking ideas around with an AI chatbot is a great creative exercise that can still be sustainable if we approach it intentionally.
Adding creative boundaries to a prompt makes for a lighter request. Endless prompts like “give me an idea for a bakery logo” are considered resource-intensive because they are vague and lack a clear endpoint. Starting with a more focused idea that you can then expand in a specific direction demands less processing power.
4. Ask precise follow-up questions
It’s important to follow up with precision. The example on the left is vague and directionless, whereas the one on the right takes a smaller piece of the answer and asks AI to expand on it. You’re saving time and resources by omitting the information you don’t need.
5. Use search engines for traditional search queries
A search engine is more resource-efficient for traditional search requests because, while AI “thinks” to give you an answer, a search engine just pulls up a list of hits for you to explore.
If you notice you can’t escape AI even if you’re using a search engine, you can turn off the auto-AI response. Some browsers have an option for this, sometimes you have to go into the settings to find it. If you don't want to mess with any of that, you can try adding “-AI” at the end of your search entry. It should skip the automatic AI response.
A Prompt for Scoring the Sustainability of Your AI Request
Here’s a quick prompt to score the sustainability of an AI prompt or interaction, where 1 is equivalent to about a half a cup of water, and 10 is a scorched-earth type of scenario. It’s a work in progress. Feel free to try it, tweak it, share it, and let me know how it works for you.
Assign a sustainability score to my request, where 1 = very resource-efficient and 10 = very resource-intensive.
Consider these scoring factors:
Tokens inTokens outTask complexityLikelihood of multiple iterations
Explain the reason for the score in 1-2 sentences, and suggest 1 way to make the prompt more efficient.
About the Author
Ce Ramirez is a Content Designer at Coforma focused on serving communities by bridging the gap between the abundance of information and actionable knowledge. With over 15 years of experience as a translator in public and nonprofit spaces, she leverages her experience with multiculturalism to understand and thoughtfully advocate for solutions that can create tangible, positive impact and better serve people.
