Mapping the Knowledge Base on Sustainable AI: A Systematic Review

John W. Kasubi
Local Government Training Institute, Dodoma, Tanzania
Email: johnkasubi7@gmail.com

Abstract: The rapid progress of Artificial Intelligence has sparked substantial concerns about its environmental impact, especially regarding the significant energy usage and carbon emissions involved in training and deploying large models. The field of sustainable AI has emerged as a vital area of research, focused on creating AI technologies that strike a balance between performance and environmental and social responsibility. The PRISMA approach was applied to select documents of the study. The study used the following research algorithm phrases: ((“sustainable AI” OR “green AI” OR “eco-friendly AI” OR “energy-efficient AI” OR “low-carbon AI” OR “environmentally conscious AI” OR “sustainable artificial intelligence” OR “climate-conscious AI”) AND (“sustainability applications” OR “green technology applications” OR “environmental solutions” OR “climate applications” OR “ecological applications”)) to narrow down the search in the Scopus database and identify relevant records. The study employed VOSviewer software, and analysis was performed on 853 articles, enabling the identification of network expansion, significant contributors, the intellectual framework, the most extensively studied subject, and areas necessitating further examination. The result highlighted some new key areas such as “energy efficiency,” “internet of things,” “carbon footprint,” “ethical behavior,” “generative AI,” “large language models,” “cloud environment,” “energy-efficient software,” “edge-cloud computing,” “computing-in-memory,” and “blockchain technology,” suggesting the existence of research gaps that warrant attention. These findings provide researchers with valuable insights, delivering a comprehensive grasp of past, current, and potential implications in the field while pinpointing subjects requiring additional examination. The study demonstrates a remarkable contribution of AI to environmental effect by addressing massive energy consumption and carbon emissions.