The persistent popularity of discussing how AI is supposedly bad for the environment actually gives me the opposite impression.

Ever since reading Ryan Holiday's book, Trust Me, I'm Lying: Confessions of a Media Manipulator, popular controversies like this always make me wonder who benefits from their continued discussion.

Every time someone talks about how AI is using energy, that's free, viral publicity advertising AI, keeping us obsessed with talking about it for one reason or another. The controversy makes it stickier in our brains. It makes it feel relevant even if you aren't using the product yet.

And yes, when done correctly, a leading figure in AI acknowledging the allegation and responding to it helps to further our obsession with the idea. The reassurance might not be to assuage, but rather to keep us engaged.

The reasoning is highly motivated in both directions. I would just look at other sources. I haven't seen any source with alarming numbers, though.

Why do you think the anti-AI side has highly motivated reasoning?

AI is polarizing. People who don't like it really don't like it. This wouldn't be the first anti-AI argument that was more about sentiment than actual analysis. Again, I think the climate impact argument here is basically imported from the argument against crypto (where I enthusiastically join the skeptics) and doesn't apply here. Either way, my point is, just go look at different sources. Simon Willison has linked to some on his blog.

True.

I don't think he'd outright lie, but possibly the calculation is ... stretched ... to fit the narrative.

Let's call it a small pizza, then :)