Data Centre Dilemmas.

Yesterday I attended a Green Tech South West meetup on data centres. There were loads of great talking points brought up by the three speakers, Kierra Box (Friends of the Earth), May Rodriguez (Exeter University) and Rory Murphy (Exeter University), too many to go into here.

Kierra talked about the environmental impacts of data centres: how this goes beyond only carbon emissions, how companies are obfuscating and even rolling back on their climate commitments, and how communities and activists are organising against them.

May talked about her work building machine learning models for climate and health, and how the resources and impacts of AI conflict with the desire to use it for public good.

Rory talked about the mind-boggling numbers behind the data centres, his work reviewing academic research and public engagement, and how public AI policy is at odds with the social and environmental impact.

Sadly, the talks weren’t recorded, but hopefully the speakers’ slides will be available in due course. The Q&A session at the end was also incredibly insightful. One question (and Kierra’s response) really stuck out to me:

Do we really need to worry about the environmental impact of data centres when AI could solve the problem of producing unlimited clean energy?

Kierra pointed out that solving the clean energy problem, even if it were possible, would not negate the consequences for biodiversity, e-waste, mineral and resource extraction (which occurs deeply problematic human rights costs) and the social impacts. She also added this (paraphrased):

There are lots of technologies that have the potential to do good. Unfortunately, as humans, we often do not use them for that purpose. We have a history of deploying them in the worst possible ways. So no, I don’t think AI is going to save us.