AIQ
AIQ
Part of the AI & the Environment lesson guide. Teaching a different grade? 🌈 Explorer (5–7) · 💻 Hacker (11–14) · ⚡ Architect (15–18)
Open by asking the class to guess: "How much electricity do you think it takes to train one big AI model — like the ones behind chatbots or image generators?" Take two or three guesses and write them on the board (kids at this age often guess wildly high or wildly low — both are useful to have visible). Then reveal the real hook:
"AI helps fight climate change — but it also uses enormous amounts of energy. Training one large AI model can emit as much CO₂ as five cars!"
Let that sink in for a second, then ask: "Five cars — driven for how long, do you think? A day? A year? Their whole lifetime?" (It's their whole lifetime of driving — meaning years of normal use, compressed into one training run.) This is a good moment to introduce the idea that AI has a real, physical cost, even though it feels like it lives entirely inside a screen. Tell the class today's mission is to look at both the cost and the benefit side of AI and the environment, and to figure out where the actual balance point might be — not just decide "AI is bad" or "AI is good" for the planet.
If your class enjoys quick math, have them estimate: if one car produces roughly 4-5 tons of CO₂ per year on average, "five cars" over a lifetime is a genuinely large number — several dozen tons. Writing that number on the board next to their earlier guesses about AI's electricity use sets up the numbers-driven activity that follows.
A local comparison also lands well here: ask if anyone has ridden in a jeepney or tricycle and noticed the exhaust smoke — that's the same kind of CO₂ this lesson is talking about, just from burning gasoline instead of from a power plant burning coal or gas to run a data center. The chemical byproduct is the same gas; the source is different. This helps ground "training an AI model emits CO₂" in something the class has probably actually smelled or seen, rather than leaving it as an abstract number about a faraway server.
Work through the lesson's three scenes as a guided tour, pausing at each card to add context and check understanding — Builder-mode students can handle real numbers and should be encouraged to react to them, not just receive them.
Scene 1 — AI's Carbon Cost. On "Training Energy": "Training GPT-4 used enough electricity to power 1,000 homes for a year, and released hundreds of tons of CO₂." Ask: "Is that training happening every time someone uses the chatbot, or just once?" (Guide them to the answer: mostly just once, or occasionally when the model gets updated — but every single time someone actually asks it a question afterward, that also uses a small amount of electricity, and that happens constantly, worldwide.) On "Data Centers": "AI data centers use 1-2% of all the electricity used on Earth — and that share is doubling every few years." Ask students to predict: if it keeps doubling, what does that mean for the future? Write "1%" and "2%" on the board and remind them that "all the electricity used on Earth" includes every factory, home, hospital, and streetlight on the planet — so even a small-sounding percentage represents an enormous absolute amount of power, and a percentage that keeps doubling can become a much bigger deal within a decade even if it looks modest today. On "Candle" (flag clearly as NOT AI): "A candle turns burning wax into about 80 watts of heat but only 0.3 watts of usable light — incredibly inefficient, but on a tiny scale compared to a data center." This is a good moment to point out that "inefficient" isn't unique to AI — plenty of old technology wastes energy too, just at a much smaller scale. On "Water Usage": "Google's AI systems used 5.6 billion gallons of water in 2023 for cooling data centers." Ask: "Why would computers need water instead of just fans?" (Fans move air, but water absorbs and carries away heat far more effectively for the dense racks of chips inside a data center.) If a student asks where that water comes from and where it goes afterward, it's fine to say honestly that it depends on the facility — some water evaporates during cooling and some is returned to local supply, and this is actually one of the more contested parts of AI's environmental footprint precisely because water availability varies so much by region.
Scene 2 — AI Helping the Planet. On "Energy Grids": "AI optimizes power grids, cutting energy waste by up to 15% — which researchers say can save more energy overall than AI itself consumes." That "more than it consumes" line is worth repeating and having students react to, since it directly sets up the quiz question about weighing costs and benefits. On "Smart Farming": "AI drones that monitor crop health let farmers cut pesticide use by up to 90%." Ask: "Why would using less pesticide be good for the environment, beyond just saving the farmer money?" (Less pesticide running off into rivers and soil, less harm to insects like bees.) On "Composting" (flag as NOT AI): "Composting turns food waste into soil with zero technology involved — it's worth noticing that some of the most effective environmental actions need no AI, no electricity, and no data center at all." On "Ocean Cleanup": "AI can scan satellite images to identify ocean plastic and plan the most efficient cleanup routes for cleanup boats" — ask students how they think a computer "recognizes" plastic in a satellite photo versus, say, a wave or a boat (pattern recognition from lots of labeled examples — same underlying idea as image recognition covered elsewhere in AIQ).
Scene 3 — The Balance. On "Net Impact" (flag as NOT AI — it's the debate itself, not a technology): "The big question: does AI's environmental benefit outweigh its energy cost? Researchers actively debate this." Make clear there isn't a settled answer, and that's the honest, correct thing to say — not a dodge. On "Green AI": "Researchers are building smaller, more efficient AI models that use 100x less energy for similar results." Connect this back to the earlier "doubling every few years" fact: efficient models are one of the main tools being used to try to slow that growth down. On "Solar Power": "Running data centers on renewable energy is the fastest way to cut AI's carbon footprint" — note that this doesn't reduce the electricity used, just the emissions caused by generating it. On "Carbon Labels": "Some AI research papers now disclose how much energy was used to train the model — like a nutrition label, but for carbon." Ask the class: "Why might it help other researchers, or the public, if every AI model came with a label like this?" (It lets people compare models, hold companies accountable, and reward genuinely efficient design instead of just raw performance.)
Close with: "AI has a real environmental cost — electricity, water, and CO₂ — but it also has real environmental benefits, from smarter power grids to farms that use less pesticide to satellites that spot ocean plastic. The honest answer to 'is AI good or bad for the planet' is: it depends, and scientists are still measuring both sides."
Extension activity (15-20 minutes): Split the class into two groups — "Cost" and "Benefit." Give each group 5 minutes to list every environmental cost or benefit of AI they can recall from the lesson (or invent a plausible new one, marked separately as a guess). Then have each group present their list, and as a class, vote on which single item from either list feels like the biggest deal, and why.
Close by asking: "If you were in charge of a company building a new AI data center, name one decision you could make that would shrink its environmental cost." (Good answers: build it where electricity comes from renewable sources, make the AI models more efficient, reuse the heat the data center produces for something else nearby.) For a class that finishes early, add a stretch question: "The lesson said AI's grid optimization might save more energy than AI itself uses. If that's true company-wide, why do you think people still worry about AI's environmental cost?" (Because that "savings" claim is specific to certain use cases like grid management, it isn't automatically true for every AI application, and the overall picture — training costs, water use, e-waste, and the rebound effect of AI simply being used more and more — is still being measured and debated.)