This book establishes a shift towards sustainable AI by providing researchers with alternatives that optimize energy consumption and reduce the environmental footprint of artificial intelligence. Sustainable AI addresses these challenges through a novel approach focused on the inversely proportional relationship between computational efficiency and impactful outcomes.While generative AI has significantly transformed various domains, the challenge lies in the extensive computational resources required for its operation; its environmental impact associated with high energy consumption and carbon emissions has inevitably made the sustainability of AI a growing concern. This book comprehensively addresses the technical challenges inherent in AI's current research and adoption while surveying the fundamental capabilities of sustainable AI—specifically data-efficient learning and resource-efficient learning. These insights transform sustainable AI into a more viable and realistic solution for various scientific applications, such as semiconductor failure detection, remote sensing, autonomous navigation, and more. Readers will find useful roadmaps for balancing high-performance computing with environmental responsibility in real-world scenarios.
