Is this a side effect of ketamine or something? I can’t even ask them followup questions as to how they could possibly believe that, they just get 
Is it just “data centers use less water than farming”? Because, like, we actually need farming. (Some kinds of it.)
If anyone’s encountered this amount of denial in real life, help me out, as I’d like to be better prepared in my next conversation with one of these cultists. It always happens when I least expect it.
They’re greedy liars.
Man I have no clue. Our CEO is one of these guys now and it’s baffling. He’ll defend it to the death (to the point of callouts during stand up) as well and the strange part is that it mostly started this year. I don’t know if he went to some convention where they served the koolaid he’s sipping, but he insists AI isn’t a bubble and there are no negative environmental effects or “negligible environment effects” caused by big data centers gobbling up entire rivers. Send help
It’s in their class interest to deny reality. I don’t think it’s any more complicated than that
scoffs don’t you stupid luddites know you can just run the mini puppy kicker 9000 locally for ethical puppy kicking? proceeds to not even do that and just pays premium for the corporate puppy kicker 9000
it’s entirely likely running queries on dedicated hardware is more energy efficient per flop than using your home pc at full load. haven’t seen anyone work that out.
If models where local they would be dramatically smaller and require less power. People would use them less often because it wouldn’t be made to keep you on as long as possible. Most likely they would be specialized models like a spell checker, simple tools.
People wouldn’t even be using them all the time, let alone at full boar.
Like the raspberry pi ai camera can run models with 8mb of memory and uses practically none of the pi’s resources. There’s a medium between consuming all the power in your county, and small local models.
There’s a world where a model doesn’t have to be this all encompassing thing that only exists in a data center. There’s no logical reason to do it that way unless your goal is to hoard and restrict access to the world’s entire library.
Let me present two arguments on the side of the tech bros:
-
Whatever small town gives the AI company the best sweetheart deal gets it, and that’s sort of inherently selecting for whoever made the biggest blunder in considering the external costs, before we get into the bribes, which there are of course tons of. But this isn’t really an AI problem, this happens throughout capitalist society from factories to stadiums.
-
The overall environmental effects of AI are pretty small compared to the amount of spending being poured into it. If we had a trillion dollar gold rush into any other industry, the environmental impacts would be far more devastating.
And one counter-argument that applies to both:
- Okay sure, but just because these are the predictable consequences of any massive industrial investment, that doesn’t mean that we need to be making this massive industrial investment. Building out AI at this scale is a choice, which should be weighed against not doing it, not against some hypothetical other thing that draws trillions of dollars of investment.
(And one final point that makes a mess of any thesis this post might’ve had: I think the “datacenters use all your water” talking point is wildly overstated and I hate that it gets so much emphasis. We should be focusing on the many, many other problems caused by AI, including its actual environmental impacts like increased CO2 emissions.)
The overall environmental effects of AI are pretty small compared to the amount of spending being poured into it. If we had a trillion dollar gold rush into any other industry, the environmental impacts would be far more devastating.
The direct impacts of the data centers might be comparatively small in aggregate, but consider all the impacts of all the chip manufacturing capacity and the eventual effects of all those GPUs (etc) becoming e-waste.
I did consider! That’s still small compared to the same level of investment in, like, concrete or steel or whatever. You can buy so much more concrete per dollar than you can buy GPUs.
(And reiterating out of caution because I really don’t want to get drawn into a squabble with some debate bro waiting in the wings - I don’t agree with the pro AI position here, and I think the counter argument of “we don’t have to be doing this” is sufficient.)
-
I’ve legitimately seen people on the Python forums arguing that it’s a net positive because AGI will solve global warming…
(By people I do mean one guy and it was really only Tim Peters that was kinda on his side because he intentionally takes every post in good faith)
The year is 2047. After several complete dead ends built on the back of LLMs, a company has finally achieved AGI. It explains in perfect detail exactly the steps that will need to be taken to minimize the effects of global warming and save the maximum number of human lives.
The AGI is promptly nuked.
Hey Grok, lower the sea levels. Make no mistakes.
The guy I heard this from also thinks AI will cure cancer, and soon, so let’s hope he’s right, and also that the other health problems that come from living near pollution are in the prompt, too.
It’s corpo accelerationism. Like a bastardized form of the already bastardized Effective Altruism bullshit.
Effective accelerationism (e/acc) is a 21st-century ideological movement that advocates for an explicitly pro-technology stance.[2] Its proponents believe that unrestricted technological progress, especially driven by artificial intelligence, is a solution to universal human problems, such as poverty, war, and climate change.[3] They perceive themselves as a counterweight to more cautious views on technological innovation and often label their opponents derogatorily as “doomers” or “decels” (short for decelerationists).[3][4]
Decels? Unhinged slur to make it sound like incel.
Leave it to capitalists to come up with a worse version of posadism. At least the posadists had something tangible. AI bros haven’t even invented dynamite and they’re talking about blowing up entire countries with comically large amounts of black powder.
I think I met someone like this IRL kind of recently. It was just a sort of social mixer thing, but they were hitting a lot of these points in casual conversation. Very neolib + pro-AI was basically how I had summarized it, but this explains it more specifically.
Oh shit I thought I made that up just now…
I’d hoped you had.
I hate that it’s actually big too… Of course it’s also from the OpenAI psyop skunkworks.
perhaps they’re on some idealism where the datacenter is not built in a desert and not powered by methane generators upwind of where people live.
I too like trying to lie when I do a bad thing so no one gets mad at me.
Individual inference (querying the model for output) may take less water but the eternal need to train and retrain the model has a very large power draw. That, and the fact that we have competing private models from Facebook, Google, Anthropic and OpenAI means that the costs still stack up.
Also the techbro denialism is about figures that OpenAI and these other companies give. “Elon Musk says his cybertruck is bullet proof” level of argumentation.
Inference also takes a lot or water ever since chain of thought was invented. And although closed loop cooling systems exist and so do solar PV and batteries and wind power, these technologies are by and large not being used to power and cool data centers in practice so the environmental impact is non negligible
Every reputable explainer article I’ve found does make painstakingly clear that the worst environmental impact comes from the full lifecycle of the data center, so maybe they’re implicitly countering a counterclaim that never goes beyond the individual inference part of the game.
These people who’ve flabbergasted me on this are likely also folks who have long casually assumed that “tech will save us from climate change.” It may be a hopeless argument.
We saw this same line of argumentation with NFTs and crypto where people made pointed, empirical criticisms of them to get whataboutist responses that usually weren’t even true. They don’t care. They have capital and the market on their side (or so they think) so they have no reason to develop critical thought. They just talk for something to do and run out the clock (or move onto the next trend).
We don’t need most of the land being dedicated to agriculture. Growing animals is massively inefficient and they eat tons more grains and silage than humans would eat if we just ate the food directly. That said;
If you are talking to pedantic tech bros it isn’t just denial. Their world view rejects what you are saying. Building the data center is a net positive because it creates economic activity. They believe the act of negotiating will always lead to the best case outcome. Interference with economic activity is the worst case because that is where everything breaks down and causes problems. Here you are you proposing to them their worst case.
Most are data centers are general purpose for large public clouds like Amazon, Microsoft, and Google. They run everything from streaming video, to websites like Hexbear, to telecom like mobile devices, discord, zoom. Decoupled from training, streaming video uses far more data center resources per use than AI to answer questions and write code. So what else is it that you are coming out against that you want them to stop that they are doing that is less important than food?
To me, besides being bad for the environment, the data center build outs are a problem because they are servicing organizations that use machine learning and visual processing specifically to enable surveillance. The content generators enable corporate propaganda that is designed to prevent class solidarity. The data centers enable the military-industrial complex to use a variety of automated systems to commit acts that humans (in uniform or otherwise) might actually stop and or at least report on and now they can happen without any or recourse.
This is why to me utilitarian ethics are so suspect, if measuring utility is arbitrary, you can come to any decision if you just choose the measurements to confirm a bias. Their measurement for success, wealth, power, influence is GDP. So you will never get a moral or ethical imperative to make sense to them like “you can’t fuck up the environment because the environment is good for us.” Can you show to them how investing in nature, wildlife, conservation, regulations on pollution would increase the GDP and by extension their portfolios? That is their armor.
I think to actually get through to one of them, if it is even possible, would be to start with something bad like surveillance or never ending automated military domination and figure out why it causes people to buy fewer labubu’s.
