As I was using LLMs at work (I never use LLMs for personal projects), a question occurred to me:
What in the fuck are these new data centers supposed to actually do?
The enterprise users are all in on having LLMs write 300% of the code. And yet, both OpenAI and Anthropic are still giving these “$200 gets you $10 000 in API costs” subscriptions to every victim of AI psychosis.
Why do we have RAM, hard disk, SSD, GPU and CPU shortages and no “token shortage”?
I mean, we all know the answer, these are the offerings for the robot god, but like how in the fuck are they still obtaining loans for this shit?
You must log in or register to comment.
My personal conspiracy theory is the AI companies are being used as a front for a new Manhattan Project for the surveillance dragnet. All those flock camera feeds gotta go somewhere. Imagine being able to look up who, what, where, and when (and infer why) for any person or place in the US, linking online interaction to physical places in near-real time.
Here’s who it’s being built for.
ETA: What I see happening over the next decade, the AI bubble ‘pops,’ government quietly absorbs the (by then) mostly complete facilities and hardware as part of the bailout, then news outlets go silent about data centers and ICE activity while we fall deeper into a fascist hellpit.
Imagine Hitler’s reign, but this time there will be no Anne Frank’s.
tante thinks come the bubble pop, surveillance bureaucracy will buy the now cheap rigs, cos they have a use for that much ML compute
I think it is a combination of VCs really believing this is the next big gold rush and radical innovation which will change the world plus a few true believers who think we need all the … compute (I dislike the word) to run the acausalrobotgod who will fix all our problems (and ‘We’ need to prepare so China doesnt build it first).
To go back to the first group, when they say ‘dont get left behind’ they are bot addressing you or me, but other capitalists (as in people with money, not just people who believe the ideology is best)
Edit: wow this question attracted a lot of conspiracy theorists.
I’m purging the worst of the conspiracy theorists (which I’m sure they’ll love)
I considered posting a reminder that we do analysis here, not weirdo daydreaming, but fuck it anyone who doesn’t know where the line is deserves a ban
i had to take a lyft when my car broke down the other day and the driver told me that she thought the data centers were for powering the billionaires’ bunkers when the world ends. i thought about this a lot and at the end of the day i decided this is “directionally correct”
Im amazed how many people think the surveillance state would need AI as an excuse to build data centers.
(And not that this is the was AI is going as govs tend to have infinite money and very bad quality control (at least nowadays). Seems like people have their timelines wrong).
But I’m also getting annoyed in general (not here) with how many people are going ‘everyone who disagrees with me is actually a bot/everything I wish wasnt true is AI now’.
It’s a bubble in a specific subcategory of commercial real estate, and commercial real estate usually eats shit first when a recession hits. The tech essentially serves the same purpose as American consumer credit did in the runup to 2008, just one step removed.
Because the rich people that have invested into it want AI to be too big to fail. That way we’re forced to keep moving forward on AI, and the rich can achieve their dream of not paying workers.
My theory is that the data centers are just shell companies for farming tax credits. The various gov officials are all NDA’d and can’t won’t say what kind of deals they cut nor the kickbacks they’re harvesting.
They pre buy the memory stock years in advance with IOU money.
mass surveillance, and dynamic pricing/ad targeting.
I remember seeing an interview of Satya Nadella saying that he has thousands of GPUs he’s just sitting on, waiting to get enough power infrastructure to plug then in, so in short, they’re hoarding.
Big corporations are not in the business of selling things and making profits. Their whole line of business is gaming financial markets.
In that regard, datacentres don’t need to do anything. It just needs to demonstrate some material ‘investment’ into Making AI Happen, presenting a potential for super duper revenues in the glorious future, and thereby securing outlandish valuations today.
I could have gone with destroying excess capital to preserve the status quo, but this stage and with this administration in the US going for subtlety seems unnecessary .
For the rich to create a panopticon to keep us In check…and also to charge for knowledge they did not earn or work for
A few things:
- I think nation states anticipate this new wave of machine learning will enable unfathombly large surveillance and a new age of hurk cyber warfare.
- Some investors seem to believe that basically all fields will adopt “ai” with the same enthusiam and level as software. The film industry alone would need a hideous amount of computer hardware if that disturbing fantasy of everyone living in their own solipsistic culture bubble came to life.
- Humanoid robot wierdos are anticipating applying deep learning and computer vision to copy movements etc from humans with camera’s strapped to their faces (part of the reason glasses are being pushed again). Some of the anticipate enormous fleets of robots doing everything from manufacturing to warfare. This would need a lot of compute
- Nobody actually has any idea how to direct businesses or run nations, and a considerable amount of people at just following everyone else because everyone else can’t be wrong.
Oh also it’s entirely possible to make huge amounts of money in a bad idea as long as you get out first. I suspect a lot of investors and banks are anticipating being that person. Capitalism is hilarious.
Not to sound like a conspiracy theory, but after the Russian bot farm thing I think this is a race to have the means to Astro turf the Internet for propaganda.
I think we are going to end up with infranetworks, regionally and nationally to avoid this. And I fear that the only way to confirm that someone is not a bot soon will be ID or some sort of biometrics. I think this is a controversial take, but I’d support a non-anonymous internet/social media alternative if it meant everyone I talk to is a human.
As for a more optimistic take…. I’m sure there’s a lot of complex science shit that instead of actually mindfully researching and experimenting with, can instead just be brute forced. Kinda like those videos where they train “AI” to play racing games or Mario levels, but letting it randomly attempt it thousands of times — and eventually it even beats humans playing it in time and efficiency. Maybe we can use AI to generate more efficient machines?….
I think we are at the point where AI would be okay enough to do simple shit like simplified farm and factory work, low stakes stuff where it fucking up can’t actually have that severe of consequences. That’d require sweeping economical and social reforms to tackle unemployment, and that’s too optimistic for me. This probably won’t get us to a post scarcity society, but it could make a lot of things very trivial and cheap… if introduced ethically. There might be a point where there’s a lot of people that won’t/can’t get a job, many that shouldn’t need to get a job, and what to do with them. Not everyone can be the guy that fixes and builds the AI. Maybe being unemployed will become not a big deal
that’s great thanks
There is also a token shortage, Anthropic at least started noticeably throttling users like 6ish months ago.
Not saying there is demand at API price points since the majority of use still comes from sub’s, but still.
Maybe it’s to keep the existing assets from cratering in value. Storage and RAM prices have gone up in value, if all those assets returned to their sane values then many more loans would be upside down.
