The CodeCarbon project is offering an open-source tool to help developers estimate the CO2 emissions of their algorithms, in the aim of reducing the carbon footprint of computer code. The program takes into account the complexity of computing required to execute the code and the geographical location of the servers used.
CodeCarbon, which estimates the CO2 footprint of executing computer code, was jointly developed by Mila, a world leader in AI research founded by the Canadian researcher and AI expert, Yoshua Bengio, winner of the prestigious Turing Award in 2018.
The tool is aimed primarily at professionals and developers. As well as estimating the carbon footprint of algorithms, CodeCarbon offers advice on how to reduce the carbon emissions of that code. In fact, the carbon footprint depends on the efficiency of the code to be executed and the location of the infrastructure used to do that.
Training a powerful machine-learning algorithm can, for example, mean running multiple computing machines for days or weeks, which contributes, on a certain scale, to global warming. That's why it is important to look for "clean" and renewable energy sources, even if that means moving infrastructure.
The CodeCarbon project's long-term mission is to show developers how to reduce emissions by optimizing their code. Companies and organizations can also be advised on less energy intensive solutions if necessary.
More about CodeCarbon: codecarbon.io
