GenAI and Sustainability: Being Aware of the Environmental Impact of AI

AI is a powerful tool that we can use to create a better tomorrow, but what are the environmental costs that we are seeing today?

By Gwenna Moss Centre for Teaching and Learning

There are some things in life that we might prefer to keep out of sight and out of mind, but our environmental impact is one thing that often needs more light shone upon it. One area that is especially out of sight, out of mind when it comes to environmental issues is our digital carbon footprint. Using a computer, being online, and especially our use of Generative Artificial Intelligence (GenAI) has a huge environmental impact. GenAI consumes more resources than other computing processes such as streaming video and cloud storage. Using more resources such as carbon (coal, natural gas, etc.) and water has both an ecological and economic impact. While the corporations and industries behind these resource-intensive tools are most responsible for reducing their consumption, we can each adopt wise practices for our personal use and be mindful of the tools we are using as we work towards a more sustainable future.

Out of Sight, Out of Mind – What's Actually Creating the Environmental Impact?

  • Data Centres: AI (and the internet in general) requires massive data centres for data storage and to operate our favourite tools such as Microsoft Copilot or ChatGPT.  
  • Cooling Needs: Data centres need to keep all of their hardware cool (air and liquid cooling) in order to perform properly. Cooling pumps and fans are electrically powered. 
  • Training Hours: In addition to current operations, GenAI tools have had thousands of hours of training before becoming publicly available. All this training required large amounts of resources. 

Did you know?

The global AI industry uses as much energy as a small country and has greenhouse gas emissions comparable to the aviation industry. This data can be seen repeated in many of the links in this article, however; Business Energy UK has made a nice visualization of this data. To see some comparisons of the resources consumed in different pieces of technology, and to learn more about what you can do, The David Suzuki Foundation (linked earlier) also has an easy-to-read article on the topic. 

How Can I Be More Mindful of Sustainable AI Use?

  • Use the Right Tool: When it comes to GenAI tools there are so many it can be difficult to choose the right one. Versatile tools that can do a variety of tasks are appealing because they are like the “Swiss Army Knives” of the GenAI world. However, these types of tools use more resources than tools designed for a specific job. If you want to generate an image, a general chat tool can do it, but you may prefer a tool designed purely for image generation such as Microsoft Image Creator. At USask, start with the approved and protected AI tools, which are reviewed and supported. And if you are looking for an AI assistant to help with course and assessment design, try USask's AI Learning Design Assistant (ALDA), a focused tool for tasks such as course structure and learning-activity design.   
  • Efficient AI Use: Asking a GenAI tool to do a large task will naturally require a larger amount of resources than a smaller one. Generating a full video with audio using a tool like Invideo AI is a huge task, whereas using Microsoft Copilot to draft an email uses much less energy. Before turning to AI, ask yourself if this is a task you can do without it.
  • Batch Requests: Combine multiple ideas into one prompt. For example, if you are leveraging Microsoft Copilot to help you write your course syllabus, ask it to edit your text for clarity, conciseness, grammar, and spelling in one prompt instead of four separate prompts. 
  • Necessity Check: Ask yourself: Do I really need to use an AI tool for this task? Many tasks can be done without AI. That AI-generated image might look cool, but there are plenty of free Creative Commons images (on sites such as Pexels or Unsplash) that already exist. Encourage your students (and colleagues) to use the “3 Before Me” strategy before turning to AI for assistance (try to find the solution without AI 3 different ways before turning to a chatbot).   

AI tools are very powerful and can be excellent assistants. But we need to be mindful when we use them. Remember, our digital actions have a bigger impact than we might realize. The negative environmental impact caused by our digital habits might be out of sight, but let's not leave it out of mind. 

For more, see Chapter 1: General AI Tips  in the USask AI Prompting resource — prompting efficiently gets more from AI while using fewer resources. 

Learn More:

Title image credit: Anastasia Shuraeva on Pexels

This article was created with the assistance of AI tools, as described in the GMCTL AI Disclosure Statement.

This resource is shared by the Gwenna Moss Centre for Teaching and Learning (GMCTL), University of Saskatchewan, under a CC BY-NC-SA license.