Comment by digitcatphd | Hacker News Reader

Comment by digitcatphd | Hacker News Reader

I have started to notice this as well over the past several months, in-fact, I would say it is orders of magnitude larger than the Crypto bubble and when it bursts will be significantly more impactful. Right now, everything is propped up on the premise the hyperscalers will grow EPS proportionately to their investment and that ROA is being priced in as a best case scenario (hope) in their share price. Maybe its not ROA at all, maybe its simply FOMO, we keep citing this 'AI Race' as if there is some end objective to 'win' drawing parallels to the nuclear arms race that only resulted in massive wasted CAPEX in decaying nukes sitting in unused bunkers since the cold war. (Not to mention it isn't even clear this arms race played a direct factor in the war beside depleting resources) Everyone is happy right now. Execs get stock bonuses, investors get returns, vendors get contracts. Once this is questioned, and history demonstrates it undoubtedly will be, it will cause a cascading collapse. History is a mathematical truth.

The usable amount of AI, no matter how optimistic you are about current progression, has to be below the projections the entire AI economy is living on right now.

A considerable group of people think AGI or even ASI is right around the corner

Sure, just as a considerable group of alchemists believed the recipe for gold was right around the corner

technically we can turn lead into gold now though it's not economical or scalable. so the alchemist were proven right in the end.

> so the alchemist were proven right in the end.

Except for the fact that it wasn't right around the corner???

Coincidence, it is the same group trying to sell "AI" startup/services.

Modern AI is not an intelligence. Wonder what crap they are calling AGI.

Remove the people with horses in the race and your "considerable" group becomes much much smaller.

I've never gotten a straight answer as to whether AGI is a good thing for humanity or the economy.

Real AGI would be alive and would be capable of art and music and suffering and community, of course. So there would really be no more need for humans except to shovel the coal (or bodies of other humans) into the furnace that power the Truly Valuable members is society, the superintelligent AIs, which all other aspects of our society will be structured towards serving.

Real AGI might realistically decide to go to war with us if we've leaned anything from current LLMs and their penchant for blackmail

That's all been thought of, yeah.

No, AGI isn't a good thing. We should expect it to go badly, because there are so many ways it could be catastrophic. Bad outcomes might even be the default without intervention. We have virtually no idea how to drive good outcomes of AGI.

AGI isn't being pursued because it will be good, it's being pursued because it is believed to be more-or-less inevitable, and everyone wants to be the one holding the reins for the best odds of survival and/or being crowned god-emperor (this is pretty obviously sam altman's angle for example)

Is it believed to be inevitable, or are their careers just dependent on that being the case at all costs.

Those seem the same to me

it is believed by some to be inevitable, yes.

Best case scenario for ASI is that they create enormous wealth and will keep humans around as pets because it costs essentially nothing, like in the Culture series by Ian Banks or The Polity series Neal Asher.

Humans are extremely costly in terms of the resources we need in order to survive, the space we take up, and the messes we make (everything from the shit and piss we generate to the wars/destruction/death/violence we inflict on pretty much everything around us). We'd make for entertaining pets, but we're the farthest thing from low maintenance. Even ignoring that all other animals on earth share many of the problems we'd impose on our owners we stand alone in having done devastating amounts of damage (some irreversible) to ourselves and our planet.

I'd hope that they'd keep a few of us around, but it's hard to see the logic in them keeping all of us and allowing us the freedom to live and breed the way we do right now.

"Humans are extremely costly in terms of the resources we need in order to survive"

No. With the kind of wealth ASI can generate keeping 10 billion humans alive with a very good standard of living is like a human owning a cat.

Don't a considerable group of people think the rapture is right around the corner?

https://www.pewresearch.org/short-reads/2022/12/08/about-fou...

Always.

A surprisingly sticky belief when taught young enough. Anxiety about having been left behind every time you come home to an empty house is just a bonus.

Thankfully education about physics, first principles, and critical thinking got me out from under it. Hopefully they can do the same for the rest--if we get them young enough.

If it was that close I think they would have made well over half their money back by now . . .

You've got to figure that the mega-capitalists have so much more discretionary accumulated resources relative to government than ever.

Other than government, is there anybody else who can loosen the purse strings a little bit and have it not act as a temporary stimulant as long as it lasts?

Whether they wish it would last or even provide any benefit to the average person, seems like there are plenty who wouldn't wish more prosperity on anyone who doesn't already have it :\

The only real way for long-term growth would be to plant seeds rather than dispense mere artificial stimulants.

Unless AI makes the general public way more than capitalists have spent, it wouldn't be worth any increase in cost whatsoever for things like energy or hardware. Even non-AI software could become unaffordable if their labor costs go up enough too keep top people from being poached by AI companies flush with cash.

I bet even the real estate near the data centers gets more unaffordable, at the same time clocking a win for the local economy due to the increased cash flow and tax revenue. Except all that additional cash is flowing out of peoples' pockets not in :\

Looking at the actual job I'm leveraging software to do, we've progressed past the point where I can do everything I want to do in-house on my own hardware. The bottlenecks in my processes are now human, procedural, and there for a reason.

It feels like the AI economy is being propped up by the idea that we're right around the corner from a fabled C-suite promised land where "labor" is unnecessary, but we've already been there for like a year now if you don't need a machine to explain your own business model to you, and it's basically free.

At the moment, every AI service is dealing with capacity issues. Demand is much bigger than supply.

As long as that remains true, don't see how this bubble will be popped

I’m sure someone has numbers but I do wonder how many of their users pay and whether that covers the free users. It can still be a bubble even with high demand, if you are burning money to serve those users, because you hope one day you will rule the galaxy.

I don’t really have a strong preference, so I just use any service where I’m currently not rate limited. There are many of them and I don’t see much difference between them for day to day use. My company pays for Cursor but I burned through my monthly quota in a day working on a proof of concept that mirrored their SDK in a different language. Was it nice that I could develop a proof of concept? Yes. Would I pay 500 dollars for it from my own pocket? No, I don’t think so.

It’s like those extremely cheap food and grocery delivery apps, they made their food cheap, no delivery fees for a while… of course everyone was using it. Then, they started to run out of VC money, they had to raise prices, then suddenly nobody used them anymore and they went bankrupt. There was demand, but only because of the suppressed prices fueled by VC money.

> I do wonder how many of their users pay and whether that covers the free users.

It doesn’t cover the free users, but that’s normal startup strategy. They are using investor cash to grow as quickly as possible. This involves providing as much free service as they can afford so that some of those free users convert to paid customers. At any point they can decide not to give so much free service away to rebalance in favour of profitability over growth. The cost of inference is plummeting, so it’s getting cheaper to service those free users all the time.

> It’s like those extremely cheap food and grocery delivery apps, they made their food cheap, no delivery fees for a while… […] they started to run out of VC money, they had to raise prices

That’s not the same situation because that makes the product more expensive for the customers, which will hit sales. This isn’t the same as cutting back on a free tier, because in that situation you’re not doing anything to harm your customers’ service.

> As long as that remains true, don't see how this bubble will be popped

That's what everybody was saying in February 2000.

I'm having bad flashbacks to 2008. Then I had a series of wallet-destroying medical events, went bankrupt, lost my house, got divorced, and have been dealing somewhat successfully with PTSD from that ever since.

Fortunately, this time around, I have a partner who is gainfully employed and has valuable skills that could be used in other companies. I am somewhat successful at consulting, plus a small pension. So the chances of going completely bankrupt are small, but not zero.

But oh fuck, I am having nightmares.

> I'm having bad flashbacks to 2008.

I was college age at the time, but yeah, this is how I recall it feeling.

There's seemingly a universal malaise around that comes up even in casual conversations.

Seemingly everyone is looking for a job with no success, every third story is about some sort of layoffs, and now even the press is starting to discuss the housing bubble.

I've been waiting for the proverbial other shoe to drop for some time now. Clearly, something's gotta give and the cracks are getting too large to ignore now.

User retention is a more important metric. Everyone is silent there. That is directly tied to MRR, long term viability and ROI for investment. If those were positive I'd expect them to be crowing about it, but they aren't.

Capacity just means there is currently more demand than supply and that might be a number of negative factors driving that: users with no ROI (free users), too rapid growth, poor efficiency etc etc.

Respectfully, starting one's argument with a factually invalid statement is not a good way to argue against a bubble. If by 'every' you are referring to the foundation model providers, this is not 100% of the 'AI service' market and even then, I would argue that a lot of this demand will need to be answered to by companies measuring ROI after the FOMO or unreasonable expectations get settled in and right now, my primary argument is this ROI is driven by speculation rather than empirical measurement.

I have spoken with many companies and nearly all of them, when speaking about AI, have gotten to the point they don't even make any sense. A common theme is 'we need' AI, but nobody can articulate 'why' and in-fact they get defensive when questioned. It is almost perfectly parallel to the 'we need blockchain' argument or 'we need a mobile app'. That isn't to say those are not useful technologies, but the rapid rise, steep decline, then gradual rise is a theme in tech.

James Mickens named-and-shamed this viewpoint, which he called "Technological Manifest Destiny".

Others have observed and pointed out his prescience before:

https://news.ycombinator.com/item?id=22364699

alternately, they're too polite to say that to grow further they need to extract more value from labor and this is the biggest hope of that right now.

I'd love to see the stats though, how much capacity is used for slop vs how much is used for actual productive tasks. If half of the capacity is used by bots on social medias and scammers it doesn't mean much for the economy

Probably more than half is used to try to undercut good work with bad work

Well we'll keep scaling to meet the "demand".

Teachers are demanding not to do the work that is teaching. Lawyers are demanding not to do the work of lawyering. Engineers don't want to do coding and leaders don't want to steer the ship anymore unless it's towards AI.

Alllll the "value" is bullshit. Either AGI arrives and all jobs are over and it's eternal orgy time, or at some point the lazy ai-using losers will get fired and everyone else will go back to doing work like usual

I can already see the campaign message for the Donald trump of 2125,

Make teachers great again.

Make center for kids who can’t read good and wanna learn how to do other things good too.

Volunteer at your local public library

Watch Zoolander

This kind of black and white thinking never pans out. The truth is always nuanced.

Eternal orgy time is not possible, will never happen. And if AI is useful, which it is, it will never be abandoned. Somewhere in the middle is the real prognosis

I'm not so sure about the middle. In many more things than ever, groups tend to cluster closer to the extremes.

It may "balance" around the middle, and expect it to be noticably different than now, even without much of an actual middle.

Or the middle could end up being just another group, maybe or maybe not the most prominent one.

What percentage of that capacity is being put towards useful things and what is being put towards entertainment? The products main selling point is to make us more efficient at our job and if it’s being primarily used as entertainment which everyone I know who uses an LLM besides programmers falls into that category, then I’d say the expected profit from them is a bubble.

Are those services actually operating with profit to benefit from this demand? Or are they serving that demand while taking losses and showing unrealistically inflated metrics for demand?

I think it’s more like the “crash” that never happened with Uber/gig and crypto. I can’t tell if it’s just luck with one craze seamlessly enabling another in some cosmic coincidence or smart people being able to absorb the excess capital and capacity to pivot gracefully.

It’s basically similar to biking on rolling hills with each subsequent hill being a lot taller than the previous one. You come down the hill hard, but just before you crash to the ground, another hill comes up, and you let the momentum carry you through, and you also get a gush of wind in your back.

That’s what’s been happening since 2012 - it’s a great 13-year run now. I’ve stopped betting against it.

At least we will have a huge amount of data centers if this bubble burst. Insane compute overcapacity, as well in chip manufacture

Reminds me of all the optic fiber infrastructure that was built during the dot com bubble

Yes, this bubble is much better than crypto/nft. Same as mentioned railroad rush - capex burned, but lots of stuff have been built and left after burst.

But fiber optics are different, they stay good for a very long while and you just need to add better transceivers at the ends to upgrade them, it's questionable whether the current GPUs are even worth powering up once 2nm ones are out, and we also don't know how long they last, electromigration might start killing them in 5 years.

GPUs might become obsolete after the AI bubble bursts, but tons of supporting infrastructure, including non-GPU servers, will be in the market for pennies.

I worry about correction in chip manufacture after this – will the fab industry get even more concentrated? What if the remaining fabs scale down and set us back a decade or two cutting edge chip research? Will GPUs become extraordinarily expensive?

It is a bubble. It will collapse. The only thing that might cushion the collapse a bit is that most of the capital is from large tech companies that can absorb the fallout and pivot to the next thing. The hardware and infrastructure should be able to be leveraged for other things.

Oracle really does seem poised to be the winner here.

It was a shrewd move by Oracle.

"History is a mathematical truth"

No, history is a web of lies written by the winners, just like your daily news.

The truth always comes out. Always.

One thing that's different about AI is it's a greater percentage applicable in the real world today than most bubble things that are more hype and speculation mostly by non-technical people who are attention farming.

This statement is unrelated to the funding in the space, which is not going to be mis placed, only a question of how much.

If anything, this might be a realer version of a dot com boom.

That's why every layoff is now "because of AI", it's the perfect cover up for, "the forecast isn't looking great". No no, we're about to 10x!

And lemme guess, we'll all pay for it again. Yaaaaay. We voted for this!!!!

Orders of magnitude? Like 100x?

Reply on news.ycombinator.com