Carbon Footprint of Machine Learning for AI

The global digital environment, due to its electricity consumption and massive number of users, is now a significant generator of greenhouse gasses. Evaluating the carbon footprint of computers and digital devices is a hot topic among researchers, as they remind us that fossil fuels are burned to generate electricity on the grid where we charge our batteries and plug in our machines, and the network and internet infrastructure, most notably data centers, which draw vast amounts of power worldwide. The French Institute for Research in Computer Science and Automation launched a national Digital and Environment Program in 2022 to address this issue, as concerns grow about the staggering scale of digital emissions that we are predicted to generate in the future. A recent study simulated the potential emissions of onboard data processing in electricity-powered autonomous vehicles, which interact with sensors and artificial intelligence, and concluded that computing required by a global fleet of one billion autonomous vehicles would have a carbon footprint at least as big as the one currently generated by all of the world’s data centers. The potential emissions of machine-learning models were also analyzed, suggesting that it is challenging to estimate a detailed carbon footprint.