Curbing The Carbon Footprint Of Machine Learning

The European SustainML project is aiming to create an innovative development framework that will help AI designers to reduce the power consumption of their applications. A study published in the journal Nature revealed that a training model used for natural language processing in 2019 emitted 300,000 kg of CO2. The project is coordinated by the Spanish middleware company eProsima and includes IBM, UPMEN, Kaiserslautern-Landau University, l’Université de Copenhague, the German Research Center for Artificial Intelligence and Inria. The project will focus on the quantification of the environmental impact of algorithms and the consequences of each decision taken throughout the ML life-cycle.