Have you wondered how you may be affecting the environment when you make use of AI? While it’s true that AI is incredibly ‘smart’ and offers us access to an enormous array of knowledge and information, AI is not without environmental costs.
Just what are the impacts of our AI uses, as individuals, on the environment?
Many of us here at Riderwood often consult ChatGPT or one of the other AI services, with questions ranging from the trivial to the weighty. Even those among us who have decided not to make use of AI are, in fact, using it anyway. That’s because AI is rapidly being incorporated into every industry and into many processes and activities that affect our daily lives.
For example, AI has been integrated into search engines, language, writing, editing, and video apps. AI is also part of background programs like the autocomplete feature in texting, notes Karen Panetta, a fellow at the Institute for Electrical and Electronics Engineers and a professor at Tufts University. “Unless someone is going to disconnect from all electronic devices and the internet, they will not be able to escape AI,” she says.
So most of us are using AI frequently, whether we intend to or not. What impact is our use having on the planet.
I decided to do a little digging. The big tech companies don’t talk much about the environmental impacts of AI, but some experts outside these companies have been studying AI’s impacts.
According to the International Energy Agency, an individual’s question or prompt to AI typically uses about 2.9 watt hours of electricity. To put this figure in perspective, compare it to drying one load of laundry in a dryer, which uses 2-5 KILOWATT hours. Keep in mind that 1 kilowatt hour = 1,000 watt hours. Or consider that charging an electric car uses 50-80 KILOWATT hours.
But when the 2.9 watt hours required for that single AI prompt are multiplied by the more than 1 billion prompts fed to GPT every day by 122 million individual users, the impact is huge, according to a University of California, Berkeley, study. Individual queries around the world add up to 2.9 million KILOWATT hours daily, about the same amount of electricity that powers 100,000 average U.S. homes in a day.
Then there is the impact of ‘training’ AI to respond to our prompts and to keep the data centers that power AI going. These sprawling data centers guzzle huge amounts of energy. The U.S. Department of Energy estimates that data centers consumed 4.4% of the U.S.’s electricity in 2023 and could increase to as much as 12% by 2028.
SO WHAT’S A PERSON TO DO?
Even though we individuals at Riderwood account for only an infinitesimal portion of the impact created by millions of AI-using individuals around the world, it still makes sense to educate ourselves in ways to use AI that are less wasteful and more energy-efficient.
Here are a few suggestions for lessening our environmental impact when we ‘consult’ AI:
- ASK FOCUSED QUESTIONS.
The clearer and more specific your question, the less computation the system needs to generate an answer. Vague prompts often lead to longer, more exploratory responses. - REQUEST CONCISE ANSWERS.
Directing AI to give a short answer or keep it brief reduces the number of words generated, which reduces the energy used. - AVOID REPEATED REWRITES OF LONG TEXT.
Each time the AI regenerates a long passage, it runs the full computation again. If you are using AI for editing, have it edit small sections, rather than asking it to redo the whole piece. You’ll save energy. - WORK IN SMALLER STEPS.
Instead of asking for a very long piece at once, ask for one section at a time. You’ll get shorter responses and avoid AI generating material that you’ll end up discarding.
THE BIG PICTURE
While we want to act responsibly as individuals, we should not forget that it is the tech companies and governments who are the most responsible for lessening the environmental impact of AI. Big tech companies have invested a great deal in lobbying to ward off any federal regulation of their operations. But we need federal policies and standards that hold companies to account for energy efficiency and other environmental impacts.
The Trump administration has issued a plan to fast-track AI infrastructure, which would roll back environmental protections. More than 90 environmental organizations have publicly opposed Trump’s plan, which would allow data centers to be built without review called for by the long-established federal Clean Water Act and the National Environmental Policy Act.