Your argument seems to be flattening everything so it's just the same - "everything has an impact" - but different things have different impacts, and each needs to be measured against their utility. Me ordering a coffee at my local cafe has an impact, but it's a good deal less than me driving from London to Edinburgh. The author's argument, as I read it and generally agree, is that we don't really need LLMs to get things done, but their use comes with a large environmental cost. I don't think that will stop its use, unfortunately. There's just too much capital behind them at the moment.
We don't really need LLMs like we don't really need the internet. We don't really need that blog, nor do we really need Netflix, porn or Facebook.
Individuals value things differently, so attempting to do society-wide prioritization is always going to be a reductive exercise.
For example: Your local cafe doesn't need to exist at all. You could still drink coffee, you'd just have to make it yourself. That cafe is taking up space, running expensive commercial equipment, keeping things warm even when there aren't customers ordering, keeping food items cool that aren't going to be eaten, using harsh commercial chemicals for regular sanitization, possibly inefficient cooling or heating due to heavy traffic going in and out the door, so on and so forth.
Imagine the environmental impact of turning all cafes into housing and nobody driving to go get a coffee.
Again, this just feels like throwing your hands up in the air and saying "it's too hard to decide!" But we have to take decisions somehow if we're going to do anything.
Yes, that's the challenge of centralizing economies. You aren't going to be able to do so efficiently because you don't have every person's preferences.
If by "we're going to do anything" you mean presumably fiat power to ban LLMs, then you're better off using that fiat power to just put a sin tax on carbon emissions and letting people decide where they want to cut back.
Take one fewer hot shower a week and you've saved enough energy to power a lot of ChatGPT queries. Play one fewer hour of Minecraft. Turn off raytracing. Eat one fewer burger per month. All of those things would save more energy than forgoing a few ChatGPT conversations.
Their use does not come with a large environmental cost. The average American lifestyle has a “water footprint” of 1200 “bottles of water” per day. 10-50 ChatGPT queries == 1 bottle of water. If you decide to use ChatGPT but shorten your daily shower by a second or two you will more than offset your total water usage increase.
Thus LLMs don’t have to be that useful to be worth it. And if used in certain ways they can be very useful.
Crossing your fingers and hoping for a sci-fi technology solution to be invented for climate change seems far more realistic at this point than expecting Taylor Swift to not dump 10 tons of CO2 into the atmosphere because she wanted to have lunch in France, so I'm putting all my eggs into that basket.
