To me, the way this is being handled seems like a typical growth first, profitability after thing. I'd definitely love for it to be reversed, but there don't seem to be enough incentives for that. Optimization tends to happen when it's needed and not a moment sooner. Most likely there's a humongous amount of software that could run more efficiently if it was optimized better, but with improvements in hardware, optimization has become less and less important.

I know, more or less, that there's some worrying things going on and probably mostly becoming worse too, but that has little to do with AI. I guess for us in the West, Snowden was the eye-opener, and at this point, that's already a long time ago. What I was referring to as a conspiracy theory is data centers/AI being driven by surveillance needs.

Isn't quantum computing supposed to be better than AI data centres anyway?

Someone else here is probably more qualified to talk about quantum computing than I am, but based on my understanding, quantum computing isn't (well) applicable to everything current computation methods are. What it's good for, it's really good, but it's not a good fit for everything. I'm guessing one thing it can be great is a large number of similar computations that can be done in paralle, which I'm guessing, based on my limited understanding, also largely applies to AI. Hence it might be great for AI as well, but I'm not at all sure. I also asked AI, and it suggests that quantum computing currently has limited applicability to AI and requires further research breakthroughs to be applicable to more things.

Of course there's also the major issue of quantum computing still being mostly under research than being technology that's ready for wide use. Regardless of everything else, I'm guessing it's going to take years at best until quantum computing can be a big thing.

You expected AI to not give you a pro-AI answer, really?

What can we do so that AI uses less energy, less water, less RAM, and take up way less space?

Well, Moore's Law continues, and there are continued improvements in power efficiency and cost efficiency of chips, albeit at rates somewhat slower than historically.

The thing is, Jevon's paradox and all. If a fairy suddenly made all GPUs 1,000x more efficient, you wouldn't get 100x less power consumption, you'd just get plans for >1,000x more powerful data centers.

A reminder, though, that electricity consumption isn't an inherently bad thing (renewables account for almost 80% of electricity generation increases nowadays), and it's the reason we're able to live comfortable lives in the first place:

I imagine the R² would be even higher without the offshoring of low-end manufacturing to developing economies in the past few decades, which contributed to flatline electricity consumption for a few decades in developed economies. 

You know data centres are driving up people's electric bills, right?

And now AI bros are resorting to thuggery.

There's no correlation between the US states where electricity prices increased the most (which, nationwide, has averaged to no more than the overall inflation rate, by the way) and data center buildup, though.

(I'm not saying that it will never happen or that there were no one-off crunches, but so far it's a typical example of correlation not implying causation.) 

If anything, historically speaking, heavy industrial activity helps drive electricity prices down due to being a reliable consumer for power companies, allowing them to invest more in the grid.

Well, yes. I currently have no reason to suspect AI has been tweaked, at least in any major way, to give answers that favour AI. Notice how I also formed my own thoughts before asking AI after realizing how limited my knowledge was - I already knew where I was and wanted to see if there were any sensible arguments for/against my current understanding.

Anyway, I asked AI what the downsides of AI are to see how any bias might show up. It's not a great test, admittedly, but it doesn't really seem to favour AI in particular, as far as I can see:

Well I don't see any investment being made in a stronger grid, do you?

AI bros are too rich to care about the rest of us, so they're basically stealing drinking water and driving up noise pollution, just so they can fire more people.

Unless data centers are moved into space so they don't hog resources on Earth you won't change my mind.

I guess you haven't heard what SpaceX, of all companies, is doing? Yup, exactly that.