Reducing the Carbon Footprint in Machine Learning With Eco-Friendly AI Training
Tanish Aggarwal, Sudhakar Kumar, Sarjana Singh, Chandra Kumari Subba, Biraj Upadhyaya, Varsha Arya, Abhay Ratnaparkhi, Sunil Kumar Sharma · Advances in computational intelligence and robotics book series · 2025
Artificial intelligence (AI) has revolutionized many industries, but its rapid development poses significant environmental challenges due to the high energy consumption required to train large machines. This chapter analyzes the significant carbon footprint of AI systems and explores ways to reduce their impact on the environment, presenting strategies that can be used to bring about AI contributions to the bottom of climate change. Key techniques include model optimization, pruning, energy-efficient hardware, green data centers, distributed training, and transfer learning. Algorithmic innovations such as sparse training and variable numbers of classes are also explored. Case studies from Google, OpenAI, and IBM highlight the successful implementation of sustainable AI practices. The chapter also discusses policy recommendations and the potential of emerging technologies such as neuromorphic quantum computing to support the development of environmentally friendly AI.