Human Web CollectiveIBM CEO says there is 'no way' spending trillions on AI data centers will pay off at today's infrastructure costs
www.businessinsider.com/ibm-ceo-big-tech-ai-cap…
56 Comments
Comments from other communities
IBM has a surprisingly sane approach to LLMs:
Small models
Economically trained
Apache licensed open weights
Geared for tool usage/RAG
Well documented
Legal, licensed training data
Open experiment artifacts, including A/B tests
See: https://huggingface.co/ibm-granite
No nebulous promises, no existential hype, no scorching the environment, no underbaked user facing disasters. Just plain locally runnable tools.
It’s so sensible it hurts. This is what all “AI” companies should be doing, albeit with a little more budget, and more modern architectures than dense GQA.
defunct_punkMost corporate-facing businesses have some sense of sanity because they know every purchase of their product is going to be scrutinized to hell by a team of MBA accountants. They’ve got to actually meet expectations versus consumer business which can be sold (and defended) on marketing And hype alone
I mean, a lot of internal corporate “AI” systems I’ve seen are ChatGPT wrappers, and total messes. But they get approval and funding.
Another odd thing is that IBM’s hardly ‘selling’ here. These are basically open source contributions for PR, with the hope that they’re used internally or in their other businesses I guess. But it’s quite low-key for PR.
This is fine🔥🐶☕🔥They also made IBM Watson which was not a commercial success. They’re already familiar with ’overpromise and underdeliver’ nature of AI.
“Never interfere with an enemy while they are making a mistake”
We can quibble about if they are true enemies but these AI companies are certainly not our allies.
I hope to live long enough to see Facebook/Google/Amazon be acquired by an investment firm and sold for parts.
quick_snailMeh, better they get ordered to become Nationalized, all their IP licensed GNU, a budget allocated for their maintenance, data harvesting stripped, and anyone can submit PRs.
True public infrastructure.
I don’t know whether that holds true when they’re damaging so many things in the process. If it was only themselves they were hurting, sure…leave them to it.
I want to see private equity do a Toys’R’Us on em; buy em and leverage them to death.
A fitting end.
SalJesus Christ, when even IBM, the father of computing, says “AI is not gonna give you infinite money”, you KNOW the technology is doomed. Altho I’m more inclined to believe the AI CEOs will be like “shut up boomer we’re gonna be trillionaires and you won’t be!!”
Well, good luck with that once the investors realize you’re selling snake oil!
LungI really don’t understand why lemmy hates ai so much. I get massive productivity gains every day from AI. I agree with this exec that we probably won’t hit AGI using current tech or maybe ever, but we still have:
- incredibly useful image generation that dramatically speeds up the work of designers
- coding agents that, when used skillfully, generate usable code, and review code pretty well in PRs
- suno, which can generate stems that my musician friends are regularly using instead of having to hunt through the Internet for days for the right trumpet melody
How this translates to profit is unclear since there is a race to the bottom, companies are choosing to give away the tech to gain market share rn, but the technology is undeniably useful, and already rolled out inside of every major tech company
I get massive productivity gains every day from AI
You mean heavily hallucinated code you’re gonna spend more time fixing than if you did it by hand?
incredibly useful image generation that dramatically speeds up the work of designers
AKA: Stealing from hard working artists to not have to pay them. As an artist, go fuck yourself.
suno, which can generate stems that my musician friends are regularly using instead of having to hunt through the Internet for days for the right trumpet melody
Again, stealing from hard working artists (this time musicians). Hard fucking pass.
Also, you definitely sound like you wrote this with AI. Mods, send this clanker to the scrap heap.
Jared White ✌️ [HWC]
I really don’t understand why lemmy hates technology which regurgitates slop based on data stolen from creators against their will, erodes trust, degrades mental health, wreaks havoc on the labor force, destroys the environment, and temporarily enriches a handful of fake billionaires with fake money…gosh I just don’t understand!
I hate AI because it demands that I hate it. It has been forced down my throat on every other platform to the point where I can only really escape it on Lemmy. The art it creates is soulless, the environmental impact is going unchecked and uncharted, the constant expansion of data centers is putting a huge strain on local infrastructure, and the LLMs and front facing products are just the ad campaign. The real horror of AI aren’t spoken about; The expansion of the police state, The Military Industrial Complex creating autonomous death machines, and the entire western economy is being held up by a bubble that everyone is actively ignoring.
So far my direct experience with ai is
- minor efficiency improvements as another programming tool. It can help with syntax, boilerplate, scaling out unit tests BUT IS NEVER. A FINAL PRODUCT
- MAJOR time sink from junior developers who let ai run amuck without putting in the effort to make it work
- TECH DEBT from hundreds of excessive unit tests poorly written for maintainability and with no added value
- INSTABILITY from junior developers letting ai commit slop to the default branch, wasting my time trying to back out thousands of lines of nonsense
- MISMANAGEMENT where I felt forced to run my writeups through ai to “translate” into something recognizably generated so I could fulfill mandates to use ai
If your juniors can commit slop directly to main the problem isn’t them or the AI, it’s your process.
Mandates to use AI are the stupidest fucking thing I’ve ever heard. I quit a company over it last year.
I don’t get it. Do managers think the guys plowing the fields saw someone go by with an ox an cart and went “I keep hearing about these ox carts, but I need my manager to tell me to use it before I try”? No. The ox and cart was better than plowing the field by hand, obviously so.
If the domain experts don’t feel like a tool makes them productive, forcing them to use it isn’t going to make them productive.
OMG the mandates. We’re got those too. It’s ok to have the option, it’s bad to have the mandate.
I’m reminded of the criticism that middle management likes AI because all MGMT has to do is summarize and report up, something AI is actually pretty good at. And they think they’re smart, so they think AI should be able to do other jobs too, except they’re so detached from production that they don’t understand how far it falls short.
Lemmy is rather over-the-top for AI hate. I agree it can be damned useful, but can’t see it ever paying off for these companies.
My bitching revolves around two items:
1) It’s a fucking environmental disaster. Even without these data centers, we’re draining aquifers that took 1,000s of years to fill and major cities all over the planet are about to completely run out of water. And that’s not even touching on the power usage, pollution and how it’s affecting residential rates!
2) It’s a fucking economic disaster. Have a look at the Buffett Index, total stock market valuation vs. GDP. It’s over 200% ATM and wasn’t close to that for the Great Depression, the dotcom bust or 2008. We are in deep shit.
TrickDacyWhat community do you think you’re in? Block the community if you have a problem with the reason for its existence.
Dogiedog64I occasionally have these spells where I forget what I’ve previously learnt, and choose to try to let an ai help me out with some particularly boring code, think something mundane but laborous. So I succumb to the temptation.
And every fucking time I come out having wasted time and still having to do the thing, or at any rate, something equally boring and mundane, myself manually, because either the result is so ridiculous that it requires extensive cleanup/refactoring or just won’t get there at all.
And each time I merrily keep forgetting this, learn nothing from it beyond the short term, because I seem to be hardwired to try and skip the boring stuff.
There’s no escaping the boring stuff.
Thing is, what these models really do, is give a small dose of escapism under a fragile illusion of productivity, so for a moment there you feel good, “released”, though only postponing the boring fucking work and annoying your future self.
But that escapism sells well, because we yearn it. Same with entertainment. Games. The escape from mundane, boring and annoyingly repetitive life.
For a method of escapism, it’s so fucking inefficient and resource-intensive, we should punch ourselves in the dick just for falling for that each time.
While having this mindset, knowing rationally what I know, I still expect to fall for it in the medium term again, because fuck me if some work just isn’t boring.
And every time the excursion leaves me feeling so ashamed and dirty, I truly would deserve a punch right in the dick. And perhaps need one, to actually make this stick…
I’ve not heard a single capable person ever come out of using the current agents, or “ai”s in general, and not have ended up just wasting time and resources and feeling dirty and embarrassed for it.
The only thing good about them is the potential, which, at least currently, never materializes.
GoldholzFully agreed. The huge front runners will never recoup the cost of their training, especially since the result is just being used to train the more efficient models.
IBM doesn’t know how innovation works in corporate America, news at 11.
Of course the entire investment won’t be recovered. No one expects it to. It’s a race, and to the victor goes the spoils. Most of these efforts will fail spectacularly, but a small handful will succeed spectacularly.
- a few companies will make huge profits
- everyone else will have huge tax write offs, offsetting corporate taxes for years
- the rest of us lose
That’s great and all, except *no one is fucking succeeding.* They’re having to resort to revolving money to stay afloat and OpenAI has current debts going at over a 1.5 trillion.
Sure but it’s reasonable to say it’s too early.
My company is one example of a paying customer. Someone is earning a significant amount of recurring revenue from us as a customer. This is something new this year
Those providing companies had better be looking at the usual
- new customer rates
- income growth
- investment
- forecasts of profitability
While we have no information as to whether they are properly looking at it from a business perspective, it’s reasonable.
Presumably the forecasts are insane, as the only way to make sense of that. But time is the solution. Business people are looking at the numbers going up over time: that reality will fix insane forecasts. The bubble will burst when the business people get hit in the face with that reality.
In my future utopia, the homeless are living in hollowed-out data centers.
AI is a military/political control/fossil energy project. Purpose of spending is purely for dominance ambitions, and government will (over)spend what is needed to use the datacenter time. AI budget/regulation items are already part of military funding bills. Remember, if the US does not bankrupt itself in pretense of supremacism, China wins.
The last “novel” thing IBM did was helping the Nazi’s to create one of the most efficient genocides in history. In the last few decades though they just buy existing IP and keep it alive as long as it can make them money
Nobody wants it to pay off. They want to be the last one standing which will pay off.
Will it, though?
I hope not but it probably will.
The last one standing will probably get a bailout by corporate Democrats and every Republican.
The 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!”
Big if, because you expect a US company to win.
All companies are US companies as long as they
brivemake campaign contributions. Business has no borders.For 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.
That seems optimistic, but I’ll take it.
Worked for the US after WWII.
Well… for a while at least.
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.
The people, of course
We know. Ffs.
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.
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.
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.
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.
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
Trillions of dollars worth of compute mining dogecoin
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.
I heard ram pricing is high. There’s their use, an economic one.
So you’re saying mining crypto is gonna come back into fashion?
No shit
Everything will be fine as long as they create digital god. Just gotta keep the plates spinning for a few more years… yeah?
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.
You can see some case study examples on their consulting website.
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
I 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.
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.
Altman: “so you’re saying there’s a chance…!”
There is literally a chance if you run into a wall you could pass through it.
Just like there’s a chance you actually won that Canadian lottery you never signed up for.
So 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.
The other companies involved know this as well; they’re just not saying the quiet part out loud.
It’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.
Just ask for more tax breaks. poof Problem gone.
Don’t worry consumers will pay for all their poor decisions.
Now guess where that money comes from?