Assessing Carbon Footprint Estimations of ChatGPT

ChatGPT takes the world by storm, while its environmental impact remains a serious concern. In the absence of official data, several researchers have tried to estimate the carbon emissions linked to the service, with very deviating results. We reproduce three popular calculations using an open data model for carbon footprint quantification. This enables a transparent comparison of the approaches and helps to identify their main differences and possible potential for improvement. Our work demonstrates how open data models can be used to lead the way to more robust carbon footprint estimation.