BTW, no one I know gives a shit about the energy consumption or water usage. They absolutely want to know if these datacenters will bring jobs to their area. So far Altman, Ellison, O'Leary, Amodei, Pichai, and Zuckerberg have refused to answer that question.
[0] All except Jensen who has been really trying to explain the benefits of AI and has said these massive layoffs are a huge mistake.
A plot of land that’s already zoned for the heaviest of industrial activities, is across the street from a dump, 3 miles from an airport, 16 miles from a nuclear generating station, and in a region with good climate, and no water crunch is a pretty good place for a data center.
Facts don’t matter, it’s a religious fight. Even if you provide numbers specific to the local area there’s no way to pierce the rhetoric.
Too much land? I added up lol the land used by the 10+ golf courses in the area. Dwarfs the proposal.
Too much water? I called the head of the parks department and asked them how much water the golf courses that they operate use each year. Massive.
Regional electricity costs going up? Our nuclear generating station already sells 80-85% of all power generated wholesale to other markets.
Data centers are loud? I measured the noise outside of my house. I live on a busy street. It was much louder than the viral videos going around Facebook with titles like “Data center noise from my porch SCARY MUST WATCH”.
I don’t know about all proposed data centers everywhere, but the one they’re eyeing to build in my backyard is fine by me.
I lived in Northern Virginia for years. Data centers are everywhere.
It’s really hard to explain that centers aren’t bad and are actually far more efficient than the alternatives. Just don’t run them on coal, natural gas, or the souls of orphans. And don’t rely exclusively on evaporative cooling if it’s in the desert.
They’re having fun treating tech people like villains. It was the same or worse with bankers back in 2008-2010. Anything I have to say, any data provided, any comparisons made, are biassed because I “use data centers”. When I explain that they use data centers as well, I get the finger.
When I talked to an anti data center family member who runs a local Facebook news group (5,000+ subscribers) they just kept sending me Google AI summaries as counter points… My god. I don’t even use gen AI.
People want to enjoy the benefits of progress and data centers while still being loudly “moral”. All of this on TikTok, Facebook, and YouTube. How many data centers are involved to get a post from poster to viewer? 2? 3? 8?
A bad data center project proposal somewhere should not mean opposition to data center projects everywhere.
> At the local level, the fallout was just as direct. “Do I think that the data center vote cost me the election? Yes I do,” former Box Elder County Commissioner Lee Perry said after conceding his primary race, after voting to advance the same project.
It's really no surprise at all voters hate data centers, no matter how useful they think AI might be.
But I don't think the rhetoric will end any time soon. The people saying it seem to really believe it.
It's complete speculation! It's the new gold rush and everyone wants a data center. Most of these data centers might not even be built! And the ones that are, might never make any profit.
Allegedly.
Another groups claims false flag operation. Ain't it great?
As a side note, I wish we could muster this kind of vigor for just about any other type of public infrastructure project… nuclear/wind/solar power, fiberoptic internet, public high speed rail, new cities built around human-centric principles… you know, the things that the better part of the population stands to benefit from so at least the initial unrest is somewhat justified.
There no reason to give them tax breaks. They don’t do anything of substance to the local economy.
Imagine if every AI company was a small local business run by middle class folks and there were thousands of these little companies. The total amount of data centers, water, and energy consumption is the same.
I don't think people would be anywhere near as mad then. There are still other societal externalities around AI to get mad about, sure.
But I think one of the biggest drivers of rage around AI is inequality. It's not about what is being consumed to produce AI, it's about the tiny fraction of soulless billionaire elites that benefit from it. It's about a small number of fantastically rich assholes who keep taking more and more and more while there is less and less left for everyone else.
The rage that Luigi Mangione felt is the same rage these voters feel and I believe has the same root cause. That rage won't go away if AI gets more energy efficient or stops using water.
Jump today most countries stable enough to build infrastructure are democracies and the white collar people you are demonizing do vote and that immense investment in infrastructure is not really easy to relocate.
this is one of the core flaws in democracy, while the popularity contest generally curbs blatant abuse (also note how even that fails miserably when the electorate groups start to hate each other), the vast majority of people have no real way to judge the impacts of non-trivial decisions and judgement doesn't even need to come with certainty, just knowing which risks are worth it. voters will never get it right.
and in the information age, democratic sabotage is many times easier than informing a public that in most cases has no interest in being informed when the group think/herd instincts are triggered.
i suspect democracy only worked well thus far because it was never truly real. media was always concentrated and there were no non-democratic peers. this is no longer the case. when the media was concentrated democracy was just an emergent properly of media dynamics. now it's chaotic and subject to external targeted perturbation.
It doesn’t make any sense to me as the externalities are future not current and at no other point in time has the public cared about the future without first seeing concrete examples of harm. That hasn’t happened yet for data centers nor AI. It’s all “if, maybe, sometime in the future”
People will claim real harms but the connections are spotty at best. So it feels like people are stirring the pot. like if not Russia or China then just influencers doing it for rage bait for likes and subscribes
I’m not saying they are wrong, I can’t predict the future, I’m only saying it feels unusual for the reasons mentioned above
Regardless of what they are used for, we do not need more "data centers." This is true even outside of AI.
Putting so much of us into "the cloud" is generally harmful; encouraging people to learn about, and to do more "computing" at home -- on local machines they, or someone who cares about them, control, is better.
The question is do we want to be a Petrostate or an Electrostate
https://youtu.be/gLnxzkiB-GI?is=CHj3J-ARp0iBq_NB (Adam Tooze prezi)
https://en.wikipedia.org/wiki/List_of_countries_by_electrici...
It did not feel organic at all. It's to the point now where that initial seeding of ideas has gained legitimate traction, but the initial burst of anti-datacenter content was wild to see in real time.
Neo-luddites are usually the educated elite and genealogy is from old green or left politics but includes nationalists and social conservatives.
I think media is broadly failing to recognise this new clan.
No other technology gets as much hatred as AI is getting as the public see that as a threat to their own jobs.
Of course techies here are having trouble understanding this backlash. Maybe they should read up a bit on the Unabomber Manifesto to draw parallels on the motivations of the awful attacks against CEOs recently.
Just like crypto, you cannot use it to solve social problems with technical solutions. The same goes for AI as it still requires humans and trust to use it effectively.
The more AI data centers get built, the more it is hated and the worse society gets with this loss of trust as more people read about more mass layoffs.
Once you realize:
1) LLM = CPU 2) Session Context = L1, L2, L3, CPU Cache
the entire AI industry is operating within the CPU cache of the LLM provider this is why cost moves quadratically, increases noise - signal ration, regenerative feedback loop, it dilutes the user narration, and it actually creates an architectural induce hallucination.
We've literally solved this problem in the 1960's with a memory architecture:
the OS has a memory controller, tasked with taking data from persistant structure storage (HD) loading into CPU Cache and the CPU computers, the output is stored in RAM and then moved into HD.
this is required on all AI applications, what the industry has done, is supplement a RAG which is summarizing context, however the entire context summarized chain is still being processed by the LLM.
if you employ a well sustain memory architecture you can retrieve the context you only need to feed to the llm. reduce token cost and then reduce energy therefore less demand of datacenters.
checkout my article and what i built metaop.ai it's a humble promotion but no one cares about ai memory. https://x.com/metaopai/status/2070187664192524528