I keep seeing articles like this but does anyone actually think we're not in a bubble?
From what I've seen these companies acknowledge it's a bubble and that they're overspending without a way to make the money back. They're doing it because they have the money and feel it's worth the risk in case it pays off. If they don't spend, another company does, and it hits big they will be left behind. This is at least insurance against other companies beating them.
HN isn't always right. There was massive pushback against self driving and practically everyone was saying it would fail and is a bubble. The level of confidence people had about this opinion was through the roof.
Like people who didn't know anything would say it with such utter confidence it would piss me off a bit. Like how do you know? Well they didn't and they were utterly wrong. Waymo showed it's not a bubble.
AI is an unknown. It has definitely already changed the game. Changed the way we interview and changed the way we code and it's changed a lot more outside of that and I see massive velocity towards more change.
Is it a bubble? Possibly. But the possibly not angle is also just as likely. Either way I guarantee you that 99% of people on HN KNOW for a fact that it's a bubble because they KNOW that all of AI is a stochastic parrot.
I think the realistic answer is we don't actually know if it's a bubble. We don't fully know the limits of LLMs. Maybe it will be a bubble in the sense that AI will become so powerful that a generic AI app can basically kill all these startups surrounding specialized use cases of LLMs. Who knows?
What was promised with self-driving and what we have are orders of magnitude off. We were promised fleets of autonomous taxis - no need to even own a car anymore. We were told truck drivers would be replaced en-masse and cargo would drive 24x7 by drivers who never needed breaks. We were told downtown parking lots would disappear since the car would drop you off and drive to an offsite lot and wait for you. In short a complete blow up of the economy with millions of jobs in shipped lost and hundreds of billions of spend on new autonomous vehicles.
None of that happened. After 10 years we got self-driving cabs in 5 cities with mostly good weather. Cool, yes? Blowing up the entire economy and fundamentally changing society? No.
Waymo showed that under tightly controlled conditions humans can successfully operate cars remotely. Which is still really useful, but a far cry from the promise of everyone being able to buy a personal pod on wheels that takes you to and fro, no matter where you want to go, while you sleep that the bubble was premised on. In other words, Waymo has proven the bubble. It has been 20 years since Stanley, and I still have never seen a self-driving car in person. And I reside in an area that was officially designated by the government for self-driving car testing!
> I think the realistic answer is we don't actually know if it's a bubble.
While that is technically true, has there ever not been a bubble when people start dreaming about what could be? Even if AI heads towards being everything we hope it can become, it still seems highly likely that people have dreamed up uses for the potential of AI that aren't actually useful. The PetsGPT.com-types can still create a bubble even if the underlying technology is all that and more.
They are so-called "human in the loop". They don't have a remote driver in the sense of someone sitting in front of a screen playing what looks like a game of Truck Simulator. But they are operated by humans.
It's kind of like when cruse control was added to cars. No longer did you have to worry about directly controlling the pedal, but you still had to remain the operator. In some very narrow sense you might be able to make a case that cruise control is autonomy, but the autonomous car bubble imagined that humans would be taken out of the picture entirely.
This is absolutely false. In rare cases Waymo needs human intervention, but they do NOT have humans in the loop for the vast majority of their operations.
Cruise control also doesn't need human intervention for the vast marjory of its operation. Yet, the human remains the operator. But it seems we're just playing a silly game of semantics at this point, so let's return to the actual topic at hand.
There was a time where people believed that everyone would buy a new car with self-driving technology, which would be an enormous cash cow for anyone responsible for delivering the technology to facilitate that. So the race was on to become that responsible party. What we actually got, finally, decades after the bubble began, was a handful of taxis that can't leave a small, tightly controlled region — all while haemorrhage money like it is going out of style.
It is really interesting technology and it is wonderful that Alphabet is willing to heavily subsidize moving some people from point A to point B in a limited niche capacity, but the idea that you could buy in and turn that investment into vast riches was soon recognized as a dead end.
AI is still in the "maybe it will become something someday" phase. Clearly it has demonstrated niche uses already, but that isn't anywhere nearly sufficient to justify all the investment that has gone into it. It needs a "everyone around the world is going to buy a new car" moment for the financials to make sense and that hasn't happened yet. And people won't wait around forever. The window to get there is quickly closing. Much like self-driving cars, a "FAANG" might still be willing to offer subsidies to keep it alive in some kind of limited fashion, but most everyone else will start to pull out and then there will be nothing to keep the bubble inflated.
It isn't too late for AI yet. People remain optimistic at this juncture. But the odds are not good. As before, even if AI reaches a point where it does everything we could ever hope for, much of the dreams built on those hopes are likely to end up being pretty stupid in hindsight. The Dotcom bubble didn't pop because the internet was flawed. It popped because we started to realize that we didn't need it for the things we were trying to use it for. It is almost certain that future AI uses that have us all hyped up right now will go the same way. Such is life.
Ah, it’s the old “teleport the goalposts and then change the subject!” strategy.
Just like Waymo, LLMs are already wildly useful to me others, both technical and non-technical, and there’s no reason to think the progress is about to suddenly stop, so I don’t know what you’re even on about at this point.
> and there’s no reason to think the progress is about to suddenly stop
You seem a bit confused. Bubbles, and subsequent crashes, aren't dependent on progress, they're dependent on people's retained interest in investing. The AI bubble could crash even if everything was perfect executed, just because the people decided they'd rather invest in, as you suggest, teleportation — or something boring like housing — instead.
Progress alone isn't enough to retain interest. Just like the case before, the internet progressed fantastically through the last 90s — we almost couldn't have done it any better — but at the same time people were doing all kinds of stupid things like Pets.com with it. While the internet itself remained solid and one of the greatest inventions of all time, all the extra investment into the stupid things pulled out, and thus the big bubble pop.
You're going to be hard-pressed to convince anyone that we aren't equally doing stupid things with AI right now. Not everything needs a chatbot, and eventually investors are going to realize that too.
I ride waymos on the regular. It's not tightly controlled conditions. It's some of the hardest roads in the bay area.
>While that is technically true, has there ever not been a bubble when people start dreaming about what could be? Even if AI heads towards being everything we hope it can become, it still seems highly likely that people have dreamed up uses for the potential of AI that aren't actually useful. The PetsGPT.com-types can still create a bubble even if the underlying technology is all that and more.
What I see more of on HN is everyone calling everything a bubble. This is a bubble that is a bubble. It's all a bubble. Like literally, Sam Altman is the minority. Almost everyone thinks it's a bubble.
Hard is subjective. Multiplying large numbers is hard for humans, but easy for machines. I'd say something like the I-80 through Nebraska is one of the easiest drives imaginable, but good luck getting your Waymo ride down that route... You've not made a good case for it operating outside of tightly controlled bounds.
More importantly, per the topic of conversation, you've not made a good case for the investment. Even though it has found an apparent niche, Waymo continues to lose money like it is going out of style. It is nice of them to pay you to get yourself around and all, but the idea that someone could invest in self-driving cars to get rich from it is dead.
> What I see more of on HN is everyone calling everything a bubble.
Ultimately, a bubble occurs when people invest more into something than they can get back in return. Maybe HN is right — that everything is in a bubble? There aren't a lot of satisfactory answers for how the cost of things these days can be recouped. Perhaps there are not widely recognized variables that are being missed by the masses, however the sentiment is at least understandable.
But the current AI state of affairs especially looks a lot like the Dotcom bubble. Interesting technology that can be incredibly useful in the right hands, but is largely being used for pretty stupid purposes. It is almost certain that in the relatively near future we'll start to realize that we never needed many of those things to begin with, go through the trough of disillusionment, and, eventually, on the other side find its true purpose in life. The trouble is, from a financial perspective, that doesn't justify the spend.
This time could be different, but since we're talking about human behaviour that has never been different before, why would humans suddenly be different now? There has been no apparent change to the human.
Autonomous cars did have a bubble moment. They were hyped and didn't deliver on the promises. We still don't have level 5 and consumer vehicles are up to level 3. It doesn't mean it's not a useful or cool technology.
All great tech has gone through some kind of hype/bubble stage.
You aren't living it. I live in SF, I ride waymos every freaking day. SF has some of the tightest roads and hardest driving. Just living in SF I already know they can scale to the entire country.
"bubble" is just an oversimplification and is overused. Uber was a bubble. Video games were a bubble. Streaming services were a bubble. Self-driving was a bubble. The reality is much more complex and nuanced.
LLMs are legitimate AI for the first time, and have real use cases and have changed things across myriad industries. It's disrupting education in a big way. The google AI search thing is becoming useful. When I look at products on Amazon, I often ask it's AI review thing (Rufus?) questions and it gives me good answers, so good that I don't really validate anymore.
There's massive, intense competition, and know one can predict how it is going to go, so there probably will be things that are bubble-y that pop, sure, but it's not like AI has hit a permanent plateau and we are as far as the tech is going to go. It's just getting started, but it'll probably be a weird and bumpy path.
There's a very real possibility that all the AI research investment of today unlocks AGI, on a timescale between a couple of years and a couple of decades, and that would upend the economy altogether. And falling short of that aspiration could still get you pretty far.
A lot of "AI" startups would crash and burn long before they deliver any real value. But that's true of any startup boom.
Right now, the bulk of the market value isn't in those vulnerable startups, but in major industry players like OpenAI and Nvidia. For the "bubble" to "pop", you need those companies to lose big. I don't think that it's likely to happen.
I think we are in a bubble, which will burst at some point, AI stocks will crash and many will burn, and the growth will resume. Just like the dotcom bubble definitely was a bubble, but it was the foundation of all tech giants of today.
The trouble with bubbles is that it's not enough to know you are in one. You don't know when it will pop, at what level, and how far back it will go.
>They're doing it because they have the money and feel it's worth the risk in case it pays off.
If the current work in AI/ML leads to something more fundamental like AGI, then whoever does it first gets to be the modern version of the lone nuclear superpower. At least that's the assumption.
Left outside of all the calculations is the 8 billion people who live here. So suddenly we have AGI--now what? Cures for cancer and cold fusion would be great, but what do you do with 8 billion people? Does everybody go back to a farm or what? Maybe we all pedal exercise bikes to power the AGI while it solves the Riemann hypothesis or something.
It would be a blessing in disguise if this is a bubble. We are not prepared to deal with a situation where maybe 50-80% of people become redundant because a building full of GPUs can do their job cheaper and better.
From what I've seen these companies acknowledge it's a bubble and that they're overspending without a way to make the money back. They're doing it because they have the money and feel it's worth the risk in case it pays off. If they don't spend, another company does, and it hits big they will be left behind. This is at least insurance against other companies beating them.