Sustainable AI: What is it?

[Mattab]

Sustainable AI: What is it?

Creating Sustainable AI: What is it?

Decisions have implications. The better educated and informed we are, the better decisions we can make. Building AI systems is no different – the more we know, the more sophisticated systems we can build. This is the second in my series of articles exploring the implications of ‘the more you know’ from a sustainability perspective.

AI can be used in myriad ways to address a range of environmental and sustainability issues, from huge data centre efficiency to using earth imagery to spot plastic waste – and a lot in between. However, creating sustainable AI is an area that organisations need to look at as AI use cases grow. This means that the first question to answer is:

What do we mean by sustainable AI?

For me, it means a variety of things.

Environmental impact – A priority across most industries now, this is where many people’s thoughts go to when ‘sustainability’ is mentioned, and sadly, CO₂ is a by-product of building any AI system. In fact, it’s estimated that ChatGPT ‘drinks’ 500ml of water – used for cooling – for every 20-50 questions it’s asked. Carbon emissions need to be considered, monitored and offset, when planning an AI project. It’s important to embed any project into a wider sustainability approach, ensuring the desired outcomes are aligned with Net Zero commitments, such as these set by HPE.

Longevity – The aim is to build AI systems that offer sustained benefits – to be useful for as long as possible. It’s vital to create it with long-term aims in mind that align to your AI strategy. To quote from the ancient Chinese military book, ‘The Art of War’*, “Tactics without strategy is the noise before defeat”.

Change management – Life moves forward at incredible pace, and this should be factored in at the very beginning of the project, focusing on efficiency and avoiding waste of valuable resources. This is beneficial for the business and the environment. If data is ‘the new oil’, it is because of its value, the time it takes to find new reserves, and the amount of effort and planning to harness it. Data is like liquid, it moves, shifts and changes constantly, however unlike oil, once found, can be replicated, transformed, reused and developed.

Reusability – This is a complex one when talking about models – making something reusable increases development costs in the short term. To complete ‘The Art of War’ quote I referenced earlier, “Strategy without tactics is the slowest route to victory”. Building interdependent AI systems can create an unmanageable mess, but building with a firm plan and purpose avoids this.

Lastly, gathering metrics ensures that informed decisions can be made about model construction. Collecting them after is almost impossible!

The next article in this series will delve into the detail of building AI with software engineering, and using the right skills and techniques to create efficient, sustainable, maintainable, and supportable applications.

Matt Armstrong-Barnes
Hewlett Packard Enterprise

twitter.com/HPE_UKI
linkedin.com/company/hewlett-packard-enterprise
hpe.com/uk

*The Art of War, Sun Tzu, 2010, Capstone Publishing.