Id ask for an update on my work contract, so what we can have it written out, and negotiate my salary doubling every 6 months as well then
post
"Survival depends on sustained exponential growth" is a lot of words for "unsustainable"
Did anyone, like, hire an accountant to check on the feasibility? Or was accounting outsourced to an LLM?
Is it demand, or is it that no one gets a choice whether they use it or not?
But where is the demand coming from? Seriously.
I've stopped using Google except for when I try to find products near me, mainly because it's becoming an AI slopfest that makes it harder to find what I'm actually looking for.
That sounds sustainable
Christ on a cracker, if this doesn’t describe a bubble I don’t know what does. Not only is this just blatantly unsustainable, it’s fucking nuts.
How long until the entire planet is one giant ai facility?
Uh oh. By my napkin math, and some quick searches, there are currently 400k acres of ai/ cloud data centers, (625sq miles) and 57million sq miles of land on the planet. If we keep doubling square footage each time, the 18th 6monthly doubling will bring us to 81,920,000 sq miles, meaning we have a little under 9 years before AI has consumed the entire planet!
Ai demand or their desire to have Ai demand?
Eternal growth is one of the biggest myths. It can't be eternal and will eventually lead to a collapse.
This reminds me of some of the earlier USSR 5 year plans in the 20s and 30s.
Going to the US for the first time as a European this year, it felt like what I would imagine the Soviet Union to feel like (which ended before I was born, though I have been to Russia). Capitalism and not communism is the state ideology, but propaganda for it runs through every layer of society and hypernormalization runs wild.
America today felt more opressive than Russia did a bit more than a decade ago. Which may or may not be a surprise to anyone.
US does feel like the flip side of the coin of the USSR (was born there, albeit I was too young to remember any substantive details).
I've lived in both US and Russia for many years, albeit I first experienced US as a young adult.
I will agree that Soviet-style hypernormalization is near universal in the US. I felt this as a young adult, perhaps ~10 years before I actually watched HyperNormalisation. It was clear most of the polemics about "freedom of this or that" was just local bullshit, not to be taken seriously.
I would disagree about russia feeling more oppressive (although I haven't been there since 2009), but in some ways it was more free than the US, I will agree on that one.
I also only went to Saint Petersburg in Russia and New York, Philadelphia and DC in the US, of which I guess Philadelphia felt somewhat alright and New York was a bit of a strange mix. DC was weird, but I'm sure Moscow would have been back in the day as well. Maybe not a fair comparison.
I would say Saint Petersburg isn't too different from Moscow (albeit I've never lived there, only visited).
But I agree in the overall discussion point. US is surprisingly like russia, locals in both countries are into their provincial polemics and axiomatic beliefs.
Why was DC weird for you?
What demand? There is no AI demand.
There’s plenty of demand. CEOs and other senior leadership are demanding that it be shoehorned into anything that uses electricity.
I spoke with AI and it demands more power.
And the blood of the innocents...
Well, the chatgpt app on the play store is on place number one for the most downloaded apps, and has been there for ages. I think that less tech affine people (which don't find their way here) use Chatbots pretty often and that there is demand.
Just to give more context, running ChatGPT is expensive and not enough people are willing to pay for it. So, there is not enough demand for LLMs for it to be actually profitable.. so far.
Last I heard, which I don't have any sources for so could be completely wrong for all I know, but some of the big LLM providers do make enough money off of their users to pay for inference and infrastructure. And the only reason they aren't profitable is because of the insane amount of money they spend on developing new models.
Is this paying demand or demand that exists for free services? Now I dont know as much about AI and usage as you guys, but I don’t see any AI companies making profit.
I mean we all tried it when it first came out. How much is actual use changing month to month? I would be curious how fast or slowly it is growing.
I really wish companies had to pay for the actual costs. Maybe they wouldn't force it on people so much.
An AI feature I would actually use would be a toggle in duckduckgo that hid pages which appear to be generated by LLM.
Bonus points if it can provide this service in a cheaper way than using an LLM itself.
It's being artificially created but unfortunately it's there. Companies and startups are just slurping the kool aid
From the general public there is no real demand.
There is demand from people in leadership positions who have fallen for the hype to force it onto everyone else.
Pretty much everyone I know uses AI daily, so there definitely is demand for it. You're living in an information bubble and don't know how commonplace it truly is.
Sounds sustainable.
In case you ever had any doubt, banish it now. We are blatantly and openly ruled by the dumbest motherfuckers on the planet.
On the flipside, rejoice, for now you know, without any doubt, that even on your worst days, you are smarter than the most powerful people alive. Take what solace from that that you can.
If someone says 10x it's probably bullshit. 100x it's definitely bullshit. I don't even have a term for 1000x
If someone says 10x it's probably bullshit. 100x it's definitely bullshit. I don't even have a term for 1000x
The term we're looking for is "fantastic investment opportunity".
1,000x = KiloBullshit
100,000x = MegaBullshit
1,000,000x = GigaBullshit
the next 1000x in 4-5 years
At the risk of stating the obvious, Ars is working backwards from this metric to get their headline "double every 6 months." 2^10 = 1024, to get that number in 5 years means doubling every half-year.
But Google didn't set incremental 6-month deadlines for 5 years straight, they set a single 5-year deadline. Because in 6 months shareholders can call their bluff quite easily, but in 5 years they're hoping everyone is A) distracted by some new disaster, or B) there's a new tech hype cycle they can push. They're trying to stall the bubble popping by pointing to a nebulous future where they magically scale to infinity, and hoping we all forget that they ever made this claim.
We need to turn the entire world into paperclips GPUs
Doubling every year would be crazy unsustainable at Google scale, but to tell employees you need to do so every 6 months seems like a fever dream of someone that doesn't understand what they are asking (or knows they aren't responsible for actually doing it). The old "just throw it over the fence" approach that corporate kool-aid drinkers love because they can take credit and shift blame.
puts a bunch of AI features in, turns them on by default without user’s asking for them, mandates employees use it when ever they can.
“How could this be a bubble? Look at all the demand!”
Double every six months? Yeah. Totally feasible and absolutely not a bubble at all.
Better get some fully functioning fusion reactors to power all that.
Google’s AI infrastructure head Amin Vahdat told employees that the company must double its serving capacity every six months to meet demand for artificial intelligence services, reports CNBC.
Sound awfully like a cancer tumour inside Amin's head.
"My loyal advisor," said the king, "I must grant you a reward for your many years of good service. What is it that you desire?" And the advisor asked the king, "Sire, simply give me one grain of rice on the first square of the chessboard, two on the second square, four on the third square and so forth until all sixty-four squares are filled." "Such a small thing?" said the king. "It shall be done at once." And so the advisor was given all the rice in the kingdom.
So...no different from shareholders. Nothing changes while everything changes.
I'm really high and at first glance saw all the colorful pipes as like a 90s McDonalds play area.
I feel like that interpretation made this whole article ~70% less silly.
We need some massive breakthroughs in AI efficiency soon or we're going to literally run out of RAM manufacturing capacity for all the things. The situation is so bad that in no time at all you're likely to be put on a six-month waiting list to buy a new PC/laptop or mobile phone.
How long do you think this bubble can last realistically? I give it another year tops given that we're already seeing bubble discussion entering mainstream media. That's a clear sign that we've now reached the peak, and that investors are starting to get cold feet. Given that the whole thing runs on free money, investors losing confidence is going to mean the whole house of cards will start coming down.
"The company faces similar constraints serving its 800 million weekly ChatGPT users, with even paid subscribers regularly hitting usage limits for features like video synthesis and simulated reasoning models."
Who is paying for this crap?

all 12 comments