Beyond the glass slipper: environmental impact is the Cinderella factor in Responsible AI

When Sue Turner talks to leaders and consults with organisations about using Artificial intelligence (AI) with wisdom and integrity – known as Responsible AI – there’s one element that is always like Cinderella in the scullery: the environmental impacts.

Training large AI models requires massive amounts of computing power, translating to a hefty carbon footprint and high water consumption to cool the data centres. Companies producing large language models are not revealing the CO2 or water consumption of their models, so many guestimates are being produced – none of which point to trivial figures. Added to this, the hardware required contains rare earth elements, often mined under ethically questionable conditions. But these environmental costs are rarely factored into discussions of AI’s benefits.

New report from Microsoft

Now Microsoft have published their 2024 Environmental Sustainability Report showing AI driving up its carbon emissions, despite the company’s pledge to be carbon-negative by 2030. Microsoft President Brad Smith explained the change to Bloomberg: “In 2020, we unveiled what we called our carbon moonshot. That was before the explosion in artificial intelligence. So in many ways the moon is five times as far away as it was in 2020, if you just think of our own forecast for the expansion of AI and its electrical needs.”

What can you do about the hidden cost of AI’s power?

The excitement over AI’s capabilities is understandable, but for any organisation with pledges to do better on their environmental impact, the CO2, water and other issues that arise from using AI can’t be ignored. Whether you are building your own AI models or using those supplied by others, here’s how you can incorporate environmental impacts into your Responsible AI approach:

  • Transparency and Foresight: Conduct thorough environmental impact assessments before deploying AI systems – which means asking third-party suppliers to be upfront about their environmental impacts too
  • Keep it simple: Do you really need to use an energy-intensive AI model? If a simpler algorithm, that uses significantly less compute, will do the task well enough, then that may be the right option
  • Green AI: Consider your options to develop energy-efficient AI models and even invest in renewable energy sources to power them if you can.

By taking a holistic approach to Responsible AI we can ensure that this powerful technology creates a positive future.

The clock is ticking. Let’s ensure AI doesn’t leave us with a pile of ash after the ball.

You may also like...