Managing the carbon footprint of AI

What is it about?

The Accuracy-Energy Paradox describes the challenge of needing large amounts of energy to achieve disproportionately smaller gains in model accuracy. Scholars often trade-off one conflicting demand for the other; ignoring the energy footprint in favour of increased accuracy or else sacrificing accuracy in favour of improved energy efficiency. Paradox Theory is an organisational conflict management strategy which suggests embracing both conflicting demands and progressing them in unison. This paper describes how paradox theory can inform the sustainability of AI, provides exemplars to illustrate how paradoxical thinking is already being applied to mitigate the Accuracy-Energy Paradox and suggests further research is needed to explore the Accuracy-Energy Paradox and develop proactive, sustainable techniques to manage its tensions.

Featured Image

Photo by Nejc Soklič on Unsplash

Photo by Nejc Soklič on Unsplash

Why is it important?

This is the first paper to define the Accuracy-Energy Paradox. It suggests that implementing a Paradox Theory mindset could help address the carbon footprint associated with the drive for increasing algorithmic accuracy.