Deep Learning Techniques for Automation and Industrial Applications

11Green AI: Carbon-Footprint Decoupling System

Bindiya Jain^* and Shikha Sharma

Department of Computer Science, Poornima University, Jaipur, India

Abstract

To advance the objectives of Green AI, artificial intelligence (AI) can be used in a variety of applications. Understanding the features, hazards, causes of environmental problems, and making proactive efforts are recommended practices for green artificial intelligence. The number of calculations needed for deep learning (DL) research has been doubling every few months, with an expected increase of 300,000 times in observations from 2012 to 2021. The carbon footprint of these calculations is rather high. Surprisingly, deep learning consumes significantly less energy than anticipated. Deep learning research can be challenging for academics, students, and researchers, especially those from emerging economies due to the high-cost of the calculations needed. In addition to accuracy and linked metrics, productivity is included as an evaluation criterion in this paper’s proposed practical solution. Our objective is to create Green AI that significantly outperforms Red AI in terms of receiver performance and to increase Green AI by reducing its environmental impact.

Keywords: Artificial intelligence (AI), auditability, ethics of AI, interdisciplinary science between AI, interpretability

11.1 Introduction

Green artificial intelligence (AI) is part of a larger environment of open systematic research that uses expensive models, has ...