There is a correlation between the training time and energy consumption, but that doesn’t mean there is a correlation between training time and carbon emissions.
An example of poor quality research with flawed assumptions designed as click-bait to get news coverage timed to land during COP26. Deleting a few photos will have zero impact on your carbon footprint.
There are currently only two credible estimates for global data center energy consumption: 196 TWh to 400 TWh for 2020. Why do we see vastly higher reports?
It is currently impossible to properly compare how sustainable one product is vs another. Pictures of wind farms look nice, but how do you choose which cloud region to deploy (or move) your resources if there is no way to compare them?