Happy weekend friends :) I've got a GREAT BIG new post for you. I've been meaning to write about the per-query based claims that generative AI is environmentally benign for quite some time, so the release of Google's new paper was a nice chance to bring it all together.
- Google's paper only provides a statistical summary, doesn't consider image and video generation, doesn't provide absolute values, and doesn't consider the fact that a growing proportion of generative systems are being forced on users rather than being intentionally queried.
- A bunch of key Substack posts presenting these systems as essentially environmentally harmless were obviously key in three major AI companies adopting the intensity tactic in rapid succession
- When you zoom out, the industrial damage being wrought by this novel form of digital bloat is measurable and tactile, with immediate and acute impacts occurring from data centre overdevelopment at the regional, state and country level
As Mar Hicks says: none of this inevitable. But the dismissive approach of downplaying the broader industrial impact of these technologies using intensity metrics and shifting the focus on consumers mirrors those of the fossil fuel industry.
Please get a cup of coffee, sit down on your couch, and have a read (and tell me what you think!) ->
https://ketanjoshi.co/2025/08/23/big-techs-selective-disclosure-masks-ais-real-climate-impact/