Eco2AI: carbon emissions tracking of machine learning models as the first step towards sustainable AI

While the global ESG agenda (Environment, Social, and Corporate Governance) is guided by agreements established between countries[1]), the development of ESG principles is happening through corporate, research, and academic standards. Many companies have started to develop their ESG strategies, allocating full-fledged functions and departments dedicated to the agenda, publishing annual reports on sustainable development, providing additional funds for research, including digital technologies and AI. Despite growing influence of ESG agenda, it remains the problem of transparent and objective quantitative evaluation of ESG progress in particular in environmental protection. This is of great importance for IT industry, as about one percent of the world's electricity is consumed by cloud computing, and its share continues to grow.[2] Artificial Intelligence (AI) and machine learning (ML) being a big part of today's IT industry are rapidly evolving technologies with massive potential for disruption. There are number of ways in which AI and ML could mitigate environmental problems and human-induced impact.