TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Ethics & Provenance · AGE 11—15 · FACTUAL · DURABLE
AI costs energy and water
Every AI answer is made in a building that needs electricity and often water.
The idea
An AI tool feels like it lives in your phone. It does not. Your question usually travels to a large building full of computers, called a data centre. Those computers need electricity to think. Many of them also use water to stay cool — often, not always; it depends on the building and the climate. So every answer has a small cost somewhere in the world. The same tools can also do real good — forecasting a cyclone, or cutting waste in a factory. So you weigh the help against the cost.
Why it matters
Anything that looks free is easy to waste. Once you know the cost is real, you start using the tool more carefully.
See it in the studio
A student asks for fifty pictures of a house and keeps one. Forty-nine pictures were made, cooled, and thrown away. Nobody saw the bill, but there was one.
Watch for this
Thinking "the cloud" is a cloud. It is a building with a roof, a car park, a power line, and often a water pipe.
Try it
Find out where the nearest large data centre to your city is. Look at it on a map. Notice what sits next to it — a power station, a river, a highway.
Prove it
Explain to a friend where an AI answer is actually made. Name the one thing that building always uses up, and the one thing it often uses up.
How it works
Data centres use electricity to run thousands of computer chips, and those chips get hot. Keeping them cool takes fans, chilled water, or both. Some data centres are cooled by air alone and use almost no water on site; others use a great deal; and water also goes into making the electricity at the power station. That is why this card says often, not always. A small model running on your own computer sends nothing anywhere — it still uses electricity. How much energy and water AI uses in total is still being measured, and the estimates change every year. So this card teaches that there is a cost, not how big it is. Later cards on this map put rough scales on it, and one of them treats the data centre as a building an architect may be asked to design.
What this idea builds on
- A starting idea.
What this idea opens up
Sources
- mt_ZpCcTU8j_o (Marble Skill Taxonomy, CC BY-SA 4.0)
- social-impact-ethics
- MIT News, Explained: Generative AI's environmental impact
- IEA, Energy and AI
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.