Choose the right tool for the right job" - Devan
Gaming for an hour, or generating 10 images using an AI tool - do you know what uses more energy?
If you care about the environment, and have questions about AI's impact on it, watch Caelyn and Devan as they explore how much energy AI tools use compared to other everyday habits. Some results will surprise you!
EMMA-LOUISE: You could be wary about using AI tools because of their perceived energy use, but how much do they actually use? First up, which uses more energy?
A - Generating 10 high-resolution AI images with a detailed prompt… or
B - Gaming on a console for an hour?
DEVAN: Gaming for an hour. My console gets hot, the fan's screaming. That's pure power being used for a long time. Creating an image feels like it's an intensive burst, but shorter.
CAELYN: See, I'm not so sure. The game is running code that already exists. The generative AI takes my prompt, processes it through an AI model that's been trained on millions of images, and then generates ten unique images from nothing. That feels like a lot more work.
That's really surprising. So, gaming for an hour actually uses way more energy than generating ten AI images!
EMMA-LOUISE: The console gaming used more energy simply because your console's hardware drew power for that entire hour.
The AI task was intense, but took less time. It's a great reminder that how long you do something for is just as important as what you're doing when it comes to your energy footprint.
Training an AI model means using massive amounts of data, which takes a lot of energy. Then running it for all your prompts and image generations keeps those data centres humming twenty-four seven. Training a popular large language model AI consumes millions of litres of water, and they need even more water to stop them from overheating. But once trained, using these models for individual tasks requires much less energy per query.
EMMA-LOUISE: Which uses more energy?
A - Having a fifteen minute back-and-forth conversation with an AI chatbot… or
B - Scrolling through a video-rich social media feed for fifteen minutes?
DEVAN: Every time you send a prompt, it has to generate a brand new answer from scratch. Does that use a huge amount of processing power in some data centre?
CAELYN: See, I'm not so sure. Scrolling through videos is constant. For fifteen minutes straight, it's streaming data. That's got to add up.
EMMA-LOUISE: Well, surprisingly, the video scroll used more energy. That's because it's a constant data stream, while the AI chat works in short, intense bursts of server processing. It just shows that a passive habit can easily have a bigger footprint than an active one.
CAELYN: Everything uses some energy. It's made me realise that some things I do already that aren't AI, use quite a lot.
DEVAN: Choose the right tool for the right job. A simple search is more efficient for simple facts. Save the heavy duty AI for when you need new ideas or image generation.
AI and environment essentials
- There are different factors to consider when thinking about AI tools’ environmental impact, like energy, water use and carbon footprint. AI tools use lots of energy, but it’s for a very short amount of time, and used in a data centre owned by an AI company.
- It matters where the data centre is located. Data centres located in areas that mostly use fossil fuels for energy will have a much higher carbon footprint than areas with cleaner energy sources.
- Training generative AI tools uses millions of litres of water, this is needed to keep the computer systems cool while they process huge quantities of training data. But, once trained, using these AI models for individual tasks doesn’t use anywhere near as much water per query.
- Different AI models use different amounts of energy. Scientists and engineers are still studying how to make them more energy efficient, and minimise any harm to the environment.
- We can all enjoy using AI tools, just like other apps and electrical devices, if we are mindful of their impact. But it’s also important that the companies behind them make commitments to sustainability.
Energy saving tips when using AI tools
| Do | Don't |
|---|---|
| Remember something you do in the background for hours can use more energy than something intense that’s over in seconds. How long you do something for matters as much as what you're doing. | Forget that there are many factors to consider when building a full picture of an AI tool’s environmental impact, including energy and water use. |
| Choose the right tool for the right job. A simple search is more efficient for simple facts. Save the AI for when you need new ideas or image generation. | Rely on AI. Can you solve the problem yourself, or are other tools better for the job? |
