Exploring Sustainability Through Clean Energy, Artificial Intelligence, and Machine Learning

Abstract

As concerns over climate change intensify due to human-induced greenhouse gas emissions, attention is increasingly turning to the energy consumption of artificial intelligence (AI) applications. Recent studies have underscored the significant carbon footprint of machine learning (ML) driven applications, largely attributed to the remarkable expansion in the size of deep learning models, which have grown exponentially over the past six years. This issue is prompting a call to action among responsible AI researchers and practitioners to address energy consumption and adopt sustainable practices in AI development. This research offers an initial examination of the energy usage patterns in AI and ML training conducted in cloud environments, along with a demonstration of how transfer learning can mitigate this energy consumption.