The same AI question emits more carbon dioxide in Texas than in Sweden (www.cio.com)

A team from Swedish AI firm Berget AI and the Stockholm Environment Institute has introduced a new methodology for calculating the carbon dioxide emissions associated with AI requests. This approach highlights significant variances in emissions depending on geographical location, revealing that a typical request to the AI model Kimi K3 produces notably higher emissions in Texas (3.2 grams) compared to Sweden (0.4 grams) and other regions like Ireland (1.6 grams) and Germany (2.5 grams). The findings underscore the environmental impact of AI operations, even when using the same hardware. This development is significant for the AI/ML community as it provides a transparent framework for understanding the climate footprint of AI technologies. By enabling companies to benchmark emissions across different suppliers, it encourages more sustainable practices within the industry. The method can influence corporate decision-making and drive a shift towards greener AI technologies, addressing the mounting concerns over the energy consumption and ecological impact of AI systems during their operation.