Today is a solstice: the longest day of the year in the northern hemisphere and the shortest one south of the equator. The trend of lengthening days, which has felt so reliable for months, reverses today — on a schedule so precise that astronomers can calculate it millennia in advance. That is what deep understanding of a system buys you: not just knowledge of where things stand, but foresight about when and how they will change. Artificial intelligence is changing our world on a similarly predictable trajectory, if we take the time to understand it properly. So let’s dig in, as we do every year, to consider the consequences of AI technology carefully as it becomes a more powerful tool for acquiring knowledge.

Many of us feel we had little say in how quickly AI entered our lives. But now that it is out of the labs, it’s transforming how you can practice sustainability. This week’s article helps you understand what AI is and how to use it sustainably.

This step sits on both the Community and Energy pathways. AI can turn electricity into knowledge, or it can simply waste power without increasing knowledge. You have tremendous agency to choose whether to use AI technology for good or ill.

You’re Already Using AI

If you search the Internet or use social media, you’re using AI now. In 2026, Google made “AI Mode” the default way for more than two billion people each month to search. Even more people see AI-selected feeds across every social app.

You can be more deliberative in your use of AI with a chatbot from one of the “big three”: Anthropic’s Claude (claude.ai), OpenAI’s ChatGPT (chatgpt.com), and Google’s Gemini (gemini.google.com/app). Type a question in plain English, and you’ve got free access to more knowledge than any person ever had in the history of the world. At great expense (and before obtaining permissions from authors and publishers), each of these AI models has been trained on every book, article, or piece of art ever published.

For acquiring wisdom, it has always been essential to know how to ask the right question, but this is becoming even more important now that answering questions has been automated. You can ask AI to explain your electric bill, plan a week of plant-based meals, generate images or videos, use a website, or write computer code on your behalf. At the same time, a good education and critical thinking skills are more valuable than ever because AI systems are unpredictable: they occasionally “hallucinate” and are not always fully “aligned” with the goals of the person interacting with them. For better results, “harnesses” are being developed that constrain an AI’s output. For example, if an AI is asked to conduct research and cite its sources, a harness might include code that automatically verifies each citation and prompts the AI to remove any hallucinated sources, forcing it into a loop until no hallucinations remain in its output.

AI is also being embedded directly into “smart appliances” and “smart cars,” with many more types of robots coming soon. AI applied to driving is on a scale from 0 to 5. A Level 0 car has no artificial intelligence. A Level 1 car adds lane-keeping or adaptive cruise control; Level 2, common in new cars today, does both at once while you keep your hands ready on the wheel. Level 3 lets the car handle driving in limited conditions. Level 4 requires no driver at all within a mapped zone—this is the Waymo robotaxi, now carrying passengers in several U.S. cities. Level 5, a car that drives itself anywhere, is not available yet (due to regulatory concerns, not necessarily technological limitations).

Everyone Hates Data Centers

Data centers make AI possible. They are buildings packed with computers capable of training models on large datasets, such as all the information available on the Internet. They’re an environmental flashpoint because they demand large amounts of power and cooling around the clock. While they could run on batteries charged by solar power and cooled quietly and passively without consuming water, it’s easier to build data centers powered by burning gas and cooled by evaporating water.

Some data centers have been built offshore, using wind power for renewable energy and free cooling from seawater. Many companies are planning to build data centers in space to use the constant power from our sun and the constant cooling of deep space, beaming the results down to Earth once they have completed the calculations necessary to train AI models. Until that happens, data centers are being built on land. In 2024, they used about 1.5 percent of the world’s electricity. The International Energy Agency expects that to roughly double by 2030.

Tokens, Electricity, and Where the Work Happens

AI works by sending input “tokens”—a “prompt” consisting of text, sounds, images, or movements converted to a series of numbers—through a model on a computer, which responds with output tokens that can be interpreted as text, images, sounds, or motor movements. Unlike most computer interactions, this prompt and response is intentionally designed to be somewhat random rather than strictly deterministic. In other words, when you type “hi” as a prompt, an AI model could respond in many different ways, such as “hi, to you, too,” “hello,” or “what’s up?” This is unlike a word processor, which should always respond by displaying “hi” if you type “h” and then “i” when it is expecting to receive input from you.

Sending tokens through an AI model requires electricity. Plain text prompts and responses are cheap, meaning they require very little electricity. Google estimated that responding to an average Gemini text prompt uses about as much electricity as showing 9 seconds of television, plus roughly 5 drops of water for cooling. OpenAI has put a typical ChatGPT query in the same ballpark for energy and water consumption. But generating an image or video, running a “reasoning” model that thinks in many steps, or turning an agent loose for an hour to write an app for you, can use hundreds to thousands of times more energy than a simple text question.

Where the work happens matters too. Using an AI model in the “cloud” means sending a message to a distant data center that does the computing. An “edge,” or on-device, AI model means your own phone does the calculation. Google’s Gemini Nano and Apple Intelligence now run on phones—requiring a tiny sip of energy from your battery and no internet connection, since your data never leaves your hand.

The Other Side of the Ledger

Using AI consumes resources, but to what end? Some uses benefit society and our planet. Google DeepMind’s GraphCast correctly called Hurricane Lee’s landfall nine days out. It produces a 10-day global weather forecast in under a minute, beats the gold-standard system on about 90 percent of measures, and runs roughly 1,000 times more energy-efficiently than conventional forecasting. Google’s Flood Hub gives free AI flood forecasts up to seven days ahead across more than 150 countries. The same kind of AI that strains data centers has been turned on them: a DeepMind system’s analysis cut the energy used to cool Google’s data centers by 40 percent. So, deploying AI to study and solve problems could reduce the electricity required for weather forecasting and YouTube video streaming.

Your Step This Week

Given this background knowledge about AI, how can you begin using this new technology in ways that have a positive impact on Earth?

Remember that the energy impact of asking an AI model a question that it can answer with text is equivalent to watching a few seconds of a YouTube or Instagram video streamed from a data center. Consider what knowledge you might gain that would be worth that environmental impact.

Open one of the three free tools (Claude, ChatGPT, or Gemini) and ask it a real question—something you actually want help with, ideally something that nudges you toward living a little more sustainably. Compare different ways to cool your home, or plan a week of meals around what’s already in your fridge. Check the answers AI gives you, remembering you need to check anything they generate because of their innate randomness. An effective strategy is to ask all three leading AI systems the same question, have them compare their answers, and then add your own insights and thinking to the mix.

Decide your own AI policy for the coming year. Where does AI genuinely help you act with more knowledge and less waste—and where would you rather not bother? Match the tool to the task: reach for a quick text answer freely, but treat image and video generation and long agent runs with more care. Explore whether AI models that run right on your phone are good enough. Explore whether AI software that runs on your phone or laptop is good enough for your needs. Many software packages now have built-in AI that works without sending anything to a data center. Don’t assume you always need to use a cloud-based AI subscription. Used with clarity of purpose, AI can be an effective lever for wiser living. Used absent-mindedly, it’s just a waste of electricity.

To see how this week’s step connects to understanding sustainability and using clean energy efficiently, explore the Community and Energy pathways in Sustainable Practices: Your Handbook for Effective Action and visit www.suspra.com. Two earlier pieces make good next reads: Insight: How We’ll Get More AI With Less Energy and AI as Your Sustainability Study Buddy.

References and Resources

Related One Step Articles