Our thinking

Let’s Chat(GPT)

12 July, 2024

We’ve written before about the environmental footprint of AI. The servers required to compute the complex requests asked of it require a huge amount of water for cooling and energy for processing. It’s in the news again now with reports that Google’s carbon footprint has increased by nearly 50% since 2019 because of AI. A significant spanner in the works when it comes to achieving its 2030 Net Zero target. 

To demonstrate the scale of the demand, one third of the nuclear power plants in the US are apparently discussing deals with tech companies to power AI. In fact, the Electric Power Research Institute estimated that by 2030, 9% of electricity in the US could be consumed by data centres, more than double what’s being used now. While these tech giants may be buying into nuclear to reduce their carbon footprint, where does the displaced demand go and is AI worth the strain on resources? 

It’s a huge question, and one the world might not yet be able to fully answer, but we thought we’d start by seeing what ChatGPT thought – what’s Friday 5 if not balanced?  “Is AI good?” Answer? Perhaps unsurprisingly, it was inconclusive. There are many benefits from efficiency savings, automation and optimisation, all of which come with the caveat that it is “essential to address its challenges and ethical concerns to ensure it benefits society as a whole.” It’s hard to disagree with this. And while there needs to be a grace period in innovation to allow its potential to be realised, it’s also important to ensure that we’re pragmatic and continue to weigh the benefits against the undoubted impact this innovation has on the climate.  

Sitting behind this is  another question which we hope that Google can answer for us sooner rather than later too – can they (and other tech businesses?) reach their 2030 Net Zero target while supporting AI, and allowing the benefits (and risks) of the innovation to unfold? Time will tell.  

By Anna Heis