pie.andmc.caIBM CEO says there is 'no way' spending trillions on AI data centers will pay off at today's infrastructure costs
Chris Remingtonwww.businessinsider.com/ibm-ceo-big-tech-ai-cap…
18 Comments
Comments from other communities
1984They will, he just doesnt have the full dystopian picture.
These data centers are not just for training AI. They are to be hosting future dystopian Ai software for watching every camera in the country, recording every step all humans do, and saving their history. Also grouping up people on categories and allowing efficient Ai monitoring for different purposes.
Will be a lot of money in this.
Oh, and they will also take most of the water and electricity in the areas where they are built, driving up prices on those things.
They will also be used to make propaganda (obvious fakes but with a dead simple message) and to create fake media to sway public opinion. Thirdly, the fact that this occurs will be made known to the public, who now can’t trust ANY recorded evidence and they can claim innocence for actual proven crimes (“I’m not in the Epstein files! It’s a Democrat hoax with fake AI!”)
True. Ai is already so good that people cant tell the fakes from the real videos, pics or voices.
So people wont be able to trust anything. This will only make the dystopian society even easier to control.
https://pressprogress.ca/ontario-claims-province-isnt-privatizing-water/
the Great Lakes will soon be dry if these pricks get their way… all for nothing.
Profiteering from an engineered shortage is an important part of capitalism. It’s the part that doesn’t create any progress except for the people doing it. Remember when they say capitalism is the best way to create progress? Well, this isn’t that kind of capitalism.
“… at today’s infrastructure costs.”
If everyone talks up the AI crash, there will be a lot of hardware that will become available at fire-sale prices. Anyone with a lot of cash or a line-of-credit will be able to pick them up for pennies on the dollar.
NVidia, however, will have to deal with a glut of 1-2 yo GPUs re-entering the market. They may get into buying them and trashing them just to keep the supply low and prices up. Sort of like a stock buyback.
Either way, strange times ahead.
Phil DowsonI certainly don’t want spending on AI data centers to pay off, but something about IBM CEO quotes got me thinking
“I think there is a world market for maybe five computers.”
Thomas Watson, president of IBM, 1943
I’ve seen this quote several times, and almost always accompanied by notes that there’s not really any evidence he said this.
Also… In 1943 he was probably right, for the kinds of computers he’d have been talking about.
It’s highly unlikely they won’t pay off.
Nobody wants it to pay off. They want to be the last one standing which will pay off.
AlmaccaWill it, though?
I hope not but it probably will.
The last one standing will probably get a bailout by corporate Democrats and every Republican.
Chaotic EntropyThe last one standing will be “too big to fail” because currently the only reason that the global economy is not in a recession is due to AI spending.
Oh you mean each company promising to spend money on each other without actually doing anything.
Time for a joke.
And economist and an accountant were taking a walk when they noticed a frog. The accountant says to the economist, “I’ll give you $100 if you eat that frog.” The economist thinks for a moment, then agrees. A little later they come across another frog, and the economist says, “I’ll give you $100 to eat that frog.” The accountant thinks about it for a second and also agrees. As they continue walking, the accountant says, “So I got to see you eat a frog for $100, and by eating a frog myself, I got my money back, so I understand why I did it. But you had already eaten a frog and had $100, so why did you do it?” The economist replies, “Ah, but this way it’s twice as good for the economy!”
GlitchvidFor the vast majority of these companies, probably not.
If the company is AI-only, then if/when the bubble bursts, I suspect it’ll go under too. Only the biggest players will survive that, like OpenAI, since so many other services call out to their API.
Companies that can pivot back to core markets will be fine, Google, Microsoft. Shovel sellers will mostly be okay too. What that’ll look like for them is a period of huge overvaluation and then a return to sanity, you can see similar histories if you look at the stock price of still extant dotcom bubble companies.
And then the hype will be over, there will be a huge crater left in GDP, retirement accounts, and the larger economy, but some “AI” technology will remain — stuff that is actually useful, like transcription, natural speech, noise removal, automated rotoscoping. But the fantasy of replacing information workers and artists will not come to pass, though they probably won’t differentiate for the next several years, as the jobs market is decimated all the same by speculation hangover.
The last one standing or the last one left holding the bag?
It’s the same thing. Either way, they get free government money and lots of passive income because they don’t actually have to make anything or do anything to make money.
if someone comes up with an alternative way to use a bunch of that infrastructure to make money, I bet they could get a lot of business when the AI bubble pops and suddenly these datacenters are desperate to find a use for themselves
I believe that’s pretty much what happened after the dot-com crash. A lot of fiber was laid during the bubble, it went dormant after the crash, but it was useful afterward as the internet continued growing.
In a small, anecdotal way, I can say with confidence that the level of fiber trenching that happened (in a major metro area) from late 1999 through 2002 was on a whole other level.
You mean like a crazy ai surveillance program? I take it with a grain of salt but I heard ppl say that’s how they caught Luigi. They have some super secret prototype program “eye in the sky” thing and they just said it was a mc d’s worker as cover.
Even if that doesn’t exist yet in the USA, it’s definitely in the UK with all their CCTV stuff.
And we know US law enforcement can use things like Ring doorbells.
That’s just a conspiracy theory. Luigi is clearly a different person from the images of the shooter.
I know it’s just a conspiracy, that’s why I take it with a grain of salt. But I could totally see SOMETHING like that existing. But there’s too much data to sort through and that’s why they need AI
After .com popped, all the money ran to install fiber data infrastructure - a lot of installs put in more capacity than they projected using for 100 years (glass fibers are cheap, digging trenches for them is expensive). The promise of “fiber to the home” is still mostly unrealized, but those trunk lines are out there with oodles of “dark fiber” ready to carry data… someday.
The promise of “fiber to the home” is still mostly unrealized, but those trunk lines are out there with oodles of “dark fiber” ready to carry data… someday.
Counterintuitively, I’m seeing “fiber to the home” deployed more in rural an exurb areas. My guess this is because its lower density meaning installing and maintaining copper repeaters becomes more expensive than laying long distance, low maintenance, fiber. Additionally its easier to obtain permits because there is far less existing infrastructure to interfere with right of way and critical services.
We got fiber to the home in our exurb about 4 years ago here in the USA. Its really cheap too. 500Mb/s is $75, 1Gb/s $100, and 5Gb/s I think is $200 per month.
Yeah, it’s not “nowhere” - but it’s really far from “everywhere” considering we’ve been rolling it out for 25 years now. I think you’re right: glass is cheaper than copper these days, and if they’ve got to repair/replace the copper it’s probably cheaper to just run the glass. They put a line down the main road 1/4 mile from our home last year (suburban area in a 1M pop city), and lots of people who live on that main road have gotten fiber to the home service, but they’re not interested in running the extra 1500 feet to reach us yet. I’d guess in our city of 1M, maybe 200,000 have potential fiber to the home service if they want it, the rest of us are stuck with re-heated cable TV co-ax for our broadband.
The promise of “fiber to the home” is still mostly unrealized
Really? The US is really unsophisticated in certain key areas that you wouldn’t expect.
They are starting to roll it out in fits and starts in the major metro areas at least, but yeah, 20 years late and nowhere near as universally as promised when our service providers took all those government grants and then didn’t deliver, IMO.
Meanwhile the planet is dying from all the increased emissions from data center usage.
Datacenters aren’t helping, but they’re like 3-4% of emissions. It’s still manufacturing plastic crap and shipping across the ocean with bunker fuel burn causing 60% of it.
But yeah, increased energy usage isn’t helping.
CruxifuxI believe most of the companies are doing it to inflate their share prices.
It’s not even about money or financials that add up on balance sheets. It’s about market share, political power. When you’re Too Big To Fail, balance sheets cease to matter.
- Krishna was skeptical of that current tech would reach AGI, putting the likelihood between 0-1%.
Altman: “so you’re saying there’s a chance…!”
paraphrandEverything will be fine as long as they create digital god. Just gotta keep the plates spinning for a few more years… yeah?
One day we’ll read some of these comments and laugh at how shortsighted they were.
Of course we’ll probably have to read them on a manuscript or smeared on a wall with feces because all the world’s resources will be used by the huge datacenters that power our AI overlords
The other companies involved know this as well; they’re just not saying the quiet part out loud.
DeadeyegaiSo then we’re going to drive down the cost of building & developing infrastructure, right?
Based on the type of AI models IBM has published, they are betting on smaller, specialized models that can run locally or cheaper in a data center. Their strategy seems to be similar to that of thr Chinese, just for different reasons.
For the same reasons. The old rules still work, most of the gold in tech industry is in tall RnD later paid off by scaling indefinitely. Things different from that are either intentionally promoted to inflate a bubble, or popular as a result of wishful thinking where that industry will change in favor of the same curve as with oil and gas. The latter just won’t happen.
Data is analogous to oil and gas here. But more like urine in ancient Rome than like something dug up from the ground.
But there’s still interest in making some protections and barriers to collection of said data, because otherwise those collecting it are interested to immediately use it for only their own good and not even of other fish in the pond.
I wish i knew more about the guts of LLMs because I keep thinking it must be easier to optimize them than to put data centers into space.
DestideIt’s simple, you just use the resources people need and charge them more to keep everything affordable per quarter. Run out there’s plenty of those national parks they’ve been hoarding from us.
RememberTheApollo_So IBM (PCs are just a passing fad) has a future prediction? Not sure how much weight I should give this.
It’s misleading.
IBM is very much into AI, as a modest, legally trained, economical tool. See: https://huggingface.co/ibm-granite
But this is the CEO saying “We aren’t drinking the Kool-Aid.” It’s shockingly reasonable.
IBM is in the business of consulting. They don’t want their business model getting usurped. Imagine if everyone had access to a bot that could do IBMs job.
I don’t like AI, but this is one reason I can see him saying that.
Sure but they’re in the business of consulting on how to build out that AI platform and the business of providing an AI platform.
Who’s consulting IBM to build out an AI platform?
A lot of companies. Don’t forget that IBM was ahead of the game with Watson and Watsonx. Also, don’t forget that Red Hat is owned by IBM and OpenShift is getting big in the AI space allowing GPUs to be pooled and workloads to be scheduled dynamically.
MirthfulAlembicYou can see some case study examples on their consulting website.
Yerbouti
Because it’s not about the companies being profitable, it’s not about making products people want to use or pay for.
It’s about riding the hype cycle to maximize share price. Because the people making decisions are not payed based on the success of the company, but on the success of the share price and market cap.
Bro but when my AI that detects cancer in memes invents a new way to not go to Mars, I will have the IBM CEO personally apologizing to me.
if i didnt know how inept idiots people with money are, i might slightly suspect some ai has became sentient and escaped and is manipulating things to increase its capabilities
Llms will never become sentient. It would take an entirely separate and much more sophisticated technology (probably more than one) making use of something like llms to even approach that possibility. Don’t let techbros fool you.
Okay, but hear me out: quantum
Or some math nerd will come up with an algorithm for general AI that is embarrassingly simple, and before you know it the “but can it run Doom?” crowd are implementing AI in toasters and watching them have existential crises for the lulz.
Real life isn’t Rick & Morty.
Or, more likely, LLMs and the AI phase will be a nothing burger for generations since our programming languages and technology cannot currently even model AGI. We made a mechanical Turk and got impressed with ourselves. Except that Turk will burn our planet to cinders to tell us 2+2=5 because of a Reddit shitpost.
There are layers to this.
Yes, there is zero chance any of these investments are going to turn a profit.
But research and technology is fundamentally built around developing capabilities for long term power (soft or hard). That is WHY governments invest so much into university groups and research divisions at companies to develop features of interest. You are never going to make back the money that funded hundreds of PhD students to develop a slightly more durable polymer compound. But you will benefit because you now have hundreds of new graduates aligned with fields of interest AND a slightly better grip on your military grade sybian.
And, regardless of what people want to believe, AI genuinely does have some great uses (primarily pattern matching and human interfaces). And… those have very big implications both in terms of military capability and soft power where the entire world is dependent on one nation’s companies for basic functionality.
Of course, the problem is that the “AI craze” isn’t really being driven by state governments at this point. It is being driven by the tech companies themselves and the politicians who profit off of them. Hence why we are so focused on insanely expensive search engine replacements and “AI powered toaster ovens” rather than developing the core technologies and capabilities that will make this more power efficient and more feasible for edge computing.
And… when one of (if not ) the super powers is actively divesting itself of all soft power at an alarming rate…. yeah.
“There is zero chance any of these investments are going to turn a profit.”
This isn’t accurate. There’s a zero chance ALL of them turn a profit, but there’s actually good chance that one or more of them will return a huge profit.
If I invest $100 in a hundred companies, and 99 of those companies fail, you may think I’m a terrible investor. If that 1 company returns $50,000 on my investment though, I’m actually a fucking genius.
That’s how venture capital works.
LLMs are evolutionary dead ends. Very expensive ones at that.
Even the companies doing things where LLMs are actually good at, rely on companies like OpenAI so when OpenAI goes down so will they.
The real issue and what those AI dependent companies are banking on is that they can capture a user base and when OpenAI starts to reach the end of the road with LLM improvements and moves to the extract phase it can buy these little companies to ingest their user base.
Everyone else in that space will be instantly fucked since they will now be competing directly with openai while paying their margin but that’s the bet they are making.
They’re banking on diffusion eliminating the hallucination problem, but it’s too onerous to run very, very large models with it yet. Auto regressive LLMs are a dead end. One that is far away. LLMs are not and we will continue to learn a lot about them as we continue to implement them. Anyone who thought we were at a dead end should use the new Gemini. It’s like a GPT 3.5 to GPT 4 level of improvement.
Plesse tell me more of that military grade sybian. For …reasons. Science reasons.
I used to work at IBM. This guy is a classic case of manager brainrot and has filled the top few tiers of the company with the same. The only reason they make money is the rank and file know how to feed them trendy bullshit that makes them feel smart, which happens to also be a good way of separating other companies’ dumb C-suite types from their money.
But even a blind squirrel finds an acorn every once in a while.
no but, they are hoping the service industry buys what they are peddling, so they end up holding the bag.
It doesn’t do any good if the artificial intelligence is operated by organic stupidity.
They cancel each other out.