Sustainable AI

Preface

Climate change is at our door step. It is causing heat waves, flash floods, droughts, and other erratic weather patterns. Addressing the challenges posed by climate change will be the defining project of our times. To do this, we should employ all the tools at our disposal. And one of the most powerful tools currently is artificial intelligence (AI), which has revolutionized tasks in many application domains. As such, AI can be indispensable in our efforts to combat climate change.

The recent class of AI methods, however, is growing to be extremely resource-intensive. Developing and using them requires powerful datacenters, which consume vast amounts of energy with correspondingly high carbon emissions. In addition, the datacenters used for AI require large volumes of fresh water in their cooling systems, rely on extractive mining to manufacture the electronics, and incur additional carbon emissions along their lifecycle. These factors, and other broader environmental impacts, pose a dilemma about using AI for sustainability.

The main argument in this book is that the material basis for any technology should not be discounted even in the light of their promised benefits. This is also true for AI. Even though AI has promised—and delivered on some—solutions to the sustainability challenges, the underlying resource cost of AI should not be ignored. If we don’t pay close attention to these massive costs, the supposed benefits offered might be eclipsed by the negative impacts ...