Tu quoque, Brute?
22 Aug 2026
Here’s another post where I wrestle with my ambivalence towards the technology I’m currently working with...
OK, so one response to people pointing out generative AI’s environmental impact goes something like this:
“Yeah but industry X is way worse! Why all the focus on generative AI?”
There’s an important point being made here, but also (I think) some deliberate misdirection.
This kind of rhetorical argument is sometimes called Tu quoque - "You too" in Latin.
It’s true that industries like agriculture and fast fashion currently do cause more harm to the environment than AI data centres.
Small example; it takes up to 2,000 gallons of water to make a pair of jeans. That’s equivalent to about 7.5 million queries of an LLM (according to OpenAI). Do I consider this when I’m shopping for clothes? I do not.
I know these kind of simplistic comparisons infuriate some people. Still, I do think we need rough heuristics like these, so that we can calibrate our response, and also renew our commitment to reforming those other industries as well.
It’s not easy - these things are really hard to compare and there are vested interests who don’t want us to have this information.
But that’s not the full picture anyway, and I think the people who make this argument are missing an important point, deliberately or otherwise. Which is this:
Generative AI is a NEW industry. Right at a point when the environment has never been under more strain, we’re proposing to add a whole new industries worth of environmental damage? And this time around we can’t say we didn’t know the extent of the impact it would have.
Pointing at other mature industries and saying “They did it, why can’t we?” is an appeal to the status quo. No progress is made by doing that - imagine if that had been our response to slavery or child labour? (actually, for a long time, it WAS).
Second point; the gen AI industry is exploding. It might not compare to agriculture right now in terms of environmental impact, but in 10 years time?
A huge concern I have about this technology is that there’s no natural upper bound on the potential demand for it. There will never be ‘enough’ gen AI in the world - we may need to force ourselves to stop building it at some point.
We should probably start thinking about that now.
Further reading #
Clara Greo has done great work to map out the various kinds of harms that applications of AI might produce. She's turned them a series of 12 prompts that teams can use if they are considering using AI in their product or service.