Can AI really be frugal?

Summary

The authors of the article observe that digital technology is permeating all other sectors.

From their perspective, AI represents one of the latest links in this technological shift: it now supports all automated processing that leverages what they call the "deluge of digital data."

Thus, the two researchers ask: given the ecological challenges we face today, will it be possible to design AI that respects environmental constraints?

The editorial perspective

The article first sets the context of AI’s environmental impact using previously published statistics—though some figures have since become outdated due to the article’s publication date. However, the key issue remains relevant: the need for clearer definitions of frugal AI.

Published in May 2024, the article makes an important point:

"Everyone is talking about frugal AI, but the term lacks a clear definition."

This website aims to bring more clarity on the topic, and the numerous initiatives listed here contribute to that goal.

The two researchers behind the article make a strong appeal:

"When it comes to AI, frugality must go beyond simple efficiency—it must be compatible with planetary boundaries. It must also challenge upstream usage decisions, including renouncing certain services and practices, based on rigorous life cycle analyses."

Their vision of frugal AI could seem radical—to them, it is the only viable way to address ecological transition challenges.