This is a bit personal. As a technologist, there’s probably never been a better time to be alive. As an environmentalist, it’s probably just the opposite.
As usual, we largely ignore the environmental impact and sustainability of large language models compared to the use cases and value they create. This whitepaper uses some descriptive data to provide a contrarian yet realistic view. TL;DR – It’s not a crisis per se yet, but it could be soon.
The comparisons need to be refined though. For example, the trend is more important than the snapshot (there is no kettle boom). We also probably need to use the kettle and the oven more than we need language models to “write a biblical verse in the style of the King James Bible explaining how to remove a peanut butter sandwich from a VCR” (from the article).
The article goes on to offer another positive: Responsible AI can spur efforts toward environmental sustainability, “from optimizing model-training efficiency to sourcing cleaner energy and beyond.” We will see about that.