Advanced · Ethics and society · module 147 of 155
Sustainability and the footprint of AI
Understand why AI costs energy and raw materials, and make choices that limit waste.
After this lesson
You know the main sources of AI’s environmental impact and choose ways of working that avoid unnecessary use.
In view · How it connects
Being mindful of AI’s footprint
Every question costs electricity, and often water for cooling.
A small model where possible, good prompts and saved results.
Figures vary widely; ask what is measured and what is not.
Part 1 of 3 · Data centres
Your preference for less movement is on. Choose each part yourself.
AI runs in data centres
Every question to an AI model is processed in a data centre with powerful computers. They use electricity, and often water for cooling too. Training large models requires a lot of energy, but daily use also counts, because it involves enormous numbers of questions. Exactly how much one question costs depends on the model, the length of input and answer and the data centre; public figures vary widely. So be careful with precise comparisons.
Sensible choices
You can limit the impact without giving up AI. Use a smaller model for simple tasks (lesson 128). Build a good prompt, so that you need fewer rounds (lesson 102). Do not have twenty images made if, with a clear image brief, you need three (lesson 122). Keep reusable results instead of having them made again. And ask yourself whether AI adds anything here: an automatic summary that nobody reads is pure waste.
Look more broadly and stay critical
Large providers sometimes publish figures on energy, water and renewable electricity. Read them critically: what is measured, and what is not? AI can also help save energy, for example in planning or logistics, but such promises require evidence just as much. Include sustainability in your choice of tools and in your AI policy (lesson 118). A sustainable way of working is often cheaper too: fewer unnecessary questions also means lower costs (lesson 80).
New practice example
From explanation to a usable result.
AI costs energy, and sometimes water. Use it where it adds something, choose the smallest model that will do the job and avoid repetition.
Fictional practice situationFictional: every morning, a municipal department automatically has all news items about the municipality summarised by the largest available model. A quick poll shows that two of the twenty staff members read the summary.
View a possible solution
This is how it can be done
The daily summary is now only made for those who want to receive it, with a smaller model that, according to a test, makes equally usable summaries. A month later, people are asked again whether it is still useful.
Why this works
Work that nobody reads only costs energy and money. The approach keeps the value for those who use it.
Make it slightly more difficult
Which automatic AI task in your environment should you review for usefulness?
Extra practice, at your own pace. You do not need to submit anything. This does not count towards an automatic assessment.
Your workplace · learning by doing
Make something. See what changes.
Checking the AI connection… You can already view the written example.
Present your assignment to Ende.
Read the answer, change one thing, and compare.
Previously held
New version
You remain in control
What do you want to improve in your next attempt?
On this device only. AI answers remain only in this page memory. This is self-review, not a grade.
Why this approach works
Work that nobody reads only costs energy and money. The approach keeps the value for those who use it.
A step further
Which automatic AI task in your environment should you review for usefulness?
Try & discover · no obligation
Become the source checker yourself
Only mark statements that the source does not support. Then check your selection.
Fictional habits in a team that uses AI every day.
Select which sentences you would correct above. The answer remains hidden until you check your selection.
This trial works locally with fictitious data and does not count towards your lesson completion. It starts anew with a new lesson or reload.
A prompt to get you started
Help me write a short guideline for a fictional team on sustainable AI use. Give five concrete habits, each with an example from office work. Do not use precise figures on energy use per question without a source, and say honestly that those figures are uncertain.
Apply it
A marketing team has fifty image variants made for every social media post and picks one each time. Most variants barely differ.
Which approach best fits this situation?
- Work with a clear image brief, have a few variants made and save usable images for later.
- Carry on; more variants always give a better result.
- Switch to a more expensive model and make even more variants.
With JavaScript, you can check your approach and receive immediate feedback.
Test your insight
Which way of working limits unnecessary AI use the most?
- Have every email summarised automatically, even if nobody reads the summary.
- First build a good prompt and keep reusable results.
- Always choose the largest model, so that you need to ask again less often.
Extra practice
Look at your own AI use for a week. Note three times you used AI without it adding anything, and what you could have done instead.
Habit 1: choose a small, fast model for short bits of text. Habit 2: write a complete prompt first instead of five vague attempts. Habit 3: for a poster, have three images made based on an image brief, not twenty. Habit 4: save good answers in your prompt library. Habit 5: switch off automatic summaries that nobody reads.
