Is the AI Bubble about to burst?

General
1

Hello everyone!

Following some responses to a previous topic made by me about AI Voice “Acting”, particularly the ones by @Demetrius_Dixon, I started to wonder whether the AI situation is just a bubble waiting to burst or it will actually be the future like literally every company is saying.

As he pointed out:

the economic AI bubble propped up by investors with the promise of AGI will pop. And I believe it WILL pop sometime in the next few years.

and the recent RAM (and other components) shortage has been showing how much maintaining these LLMs actually costs to the normal customers, so that may speed up this bursting if it ends up happening, and:

ChatGPT/OpenAI is RIDICULOUSLY unprofitable and the magic is starting to wear off for some of the general populous. None of us know if it will even exist in 5 years. The general tech will stick around, but not in this wild west/arms race type deal.

I’m starting to believe that the companies will (finally) start to understand that Artificial Intelligence will not be forever the cash cow they intend it to be and (combined with the general backlash that AI gets everyday) it’ll start to cost them more than it earns.

It’s evident though that companies have already spent billions in implementing AI everywhere they can, even in places where it makes everything less efficient and much slower (ahem, File Explorer with Copilot integrated), and so they might not be willing to cease most of the work on what would then maybe be the biggest money waste of all time.

AGI (Artificial General Intelligence) and even more ASI (Artificial Super Intelligence) are NOWHERE near being achieved even with all the work that has been put on AI (feeding AI its own contents may even be throwing us steps back tbh) and so investors may stop betting all their money on companies that put all their work on AI, contributing on the decrease of earnings from AI.

To address the other side of the coin though, we have a living example of what people thought was a bubble that then became one of the biggest things to ever exist: internet.

Many people though “Yeah this is just some niche thing that will be used by military and maybe some nerds but there’s no way everyone will start using it” and well, look where we are now (quite literally since this is a topic in a forum lol).

What do y’all think? I’m genuinely curious on what will happen, because as much as I think AI could be useful in some cases, it has caused more damages than good things.

Sorry if my English is not really good, it’s not my main language.

5 Likes
2

Please remember, when I say “AI bubble”, I mean it in a purely economic and logistical sense. Not technological. Not that I support it.

To my understanding, the definition of the word “bubble” in economics is when something is being valued higher than its actual value. Emphasis on ACTUAL value.

Every company seems to be injecting AI into everything to see what sticks, but the tech right now is still finding its TRUE place in society and our lives.

Also, I’ll emphasis this again, the logistics of long-term AI use haven’t been discovered yet. I’m not even close to an expert, but how will the physical infrastructure like data centers be funded long-term? How will AI companies curb pollution and energy use? How will the next generation, the one born into a world with advanced-ish AI, grow up?

These are all LONG-term questions that take a whole lifetime to answer, but AI companies (minus Google) only have a couple years AT MOST to burn investor money for their next breakthrough. Nobody knows it they’ll survive that long. OpenAI is already looking into putting ads into ChatGPT and an adult mode that allows NSFW to attract more users. (Although those haven’t been rolled out yet as of this reply)

Eventually, the hype will die out, the economic bubble will deflate/pop, the companies with no long-term value will vanish, and AI as we know it will become a tool like any other for those who do or don’t use it. But that doesn’t mean there isn’t a bubble. It can be both. Have you ever heard of the dotcom bubble? Once that burst, the internet stabilized.

Thanks for quoting me, BTW. Our future seems like just a gamble of sustainability at this point.

4 Likes
3

To be clear in this point, the internet was a bubble - and in March of 2000 it burst. specifically, the stock market adjusted itself and most companies that invested in the internet lost money at that time. I remember it well because I got laid off because of it in 2001. Now that took over a year, but I spent the next 9 months looking for a job (before ironically getting hired back by the same company).

The internet did not go away. And many companies rose from the ashes. Apple and Microsoft survived, and new companies like Amazon, Google, Facebook, and Twitter grew from that time. Now that’s how people make money. But back then internet investment grew faster than the adoption by normal consumers. There were no customers.

AI is experiencing the same problem. No ROI. When the bubble bursts, AI isn’t going to go away. It’s just going to shift the marketplace as companies go out of business, and new ones replace them. It’s also going to cause the market to self-correct. There are multiple companies now valued at multiple Trillions of dollars now, when NO company have ever been valued that high in history before the AI boom ChatGPT started.

6 Likes
4

It is very hard to talk about this whole topic as it is very complex: technically and economically obviously, but also socially. I hope we all agree at least that “AI” as a term, as well as the economical development in the last three years was heavily marketed as something it is not and can not be.

There is no AI here. Basically everything is a large language model with a set of transformers, adapters and generative algorithms connected to it and then packaged and labeled as “Video Generator AI”, “Image AI”, “Music AI” and so on. There is no intelligence here.

Still, the technology driving it is still amazing and magical. And that technology will remain after this current phase is over - bubble bursting or maybe less disruptive economical adaptation. We have these amazing tools now, because we have had the access to a mass of data (text, images and music) for the first time in history - and we used it all up. It has fed the machine, so to speak.

That also makes a comparison to the genesis of the internet and its development so interesting. The internet is - for the most important part infrastructure. The economy built on top of that is something different. Complex, yes, but it mimics an economy we head before, just digital. And many companies that survived the initial crash (as @dragonforge said) have either built parts of the infrastructure or changed their product a lot.
This might happen for this whole LLM/AI thing, too. The interesting part is not OpenAI, ChatGPD, Claude or Copilot as services. The interesting part is what those companies built or lock up right now.

Google, Microsoft, Amazon, Oracle, Apple and all the others are building new power infrastructure, a new layer of internet infrastructure (as in data centers, server centers, and computation centers) that will host the technology of the future. We might not pay enough attention there right now, because of all the stuff going on - but that’s where most of the money is going. Besides electrical energy, network infrastructure and computational power the major resource is data. As I mentioned above: our current tech is so good, because those companies had virtually 30 years of digitalized data analyzed, sorted, tagged and tokenized to train the current models. This alone made a huge part of the leap those companies were able to achieve. But also, this volume is now fed and gone. This will slow things down a bit now - and we see that happening, right? Most big steps now are nothing compared to the first years. The training of the models slows down and the really useful changes come from updating the generators and tools behind the LLMs. Hard-coding special cases etc.

The new data coming in is, of course, also a lot, as we keep filling the internet with more reddit posts, more tik tok videos and instagram images and more youtube videos of all sorts. And, of course, by uploading our codes to GitHub and solving our problems on StackOverflow. This, also, we should not forget. People pay a lot for Copilot, Claude and other tools, but at the same time correct the often faulty data or provide new one. We should not forget that they all sold or simply took all that data we and people before us actually generated. And this will not likely change. Weather this is good or bad is another topic.

To give a little bit of perspective I recently got: there is a big run on people in academia right now to work as fact-checkers and experts in very niche fields. There are many topics people need 8+ years of education on - chemistry, biology, physics have all become fields of research that are so granular and specialized (and, very much unstandardized in terms of research data) that there is simply not enough data to train a model with. To counter that, experts are hired to evaluate prompts and output on very specialized topics - practically fact-checking and hard-coding correct answers. This is more akin to paid wikipedia than anything AI.

I guess, what I’m trying to say is: this whole thing is overhyped and oversold, yet the technology driving it is amazing in its own merits. It comes with a high price though and will demand more in the future. There certainly is a bubble, and it will burst. Judging by the current situation of global markets, it might just burst at the same time with other disruptive events and we might misread it. Hard to tell. But we should pay attention to who builds the new infrastructure sold during this AI momentum: all around the globe data centers pop up, new power plants and power lines need to be set up, more fiber, more hardware and more exceptions from taxes, law and environmental safekeeping. There should be an alarm going off somewhere, when the big five tech giants start to get into government-level infrastructure territory.

3 Likes
5

Oh, and one more thing on bubbles. Maybe @dragonforge-dev rememembers more on that topic: I am not certain how many people were laid off during the dot-com crash in 2000 in preparation for the crash or because of it. Because what is happening right now is, tens of thousands of people are losing their jobs directly or indirectly because of this AI hype. Software engineers and artists in all fields do come to mind.

A few days ago there were some numbers about Spotify or deezer: of all of those 100.000 new songs per day (wtf?) almost half are now generated.

In a controlled and representative study it was observed that 97% of 10.000 surveyed people can not tell which of three songs was generated and which were not.

All music publishing companies in Europa are down-sizing significantly. I worked in the music industry a lot and many of the people I know - be it musicians, tech staff or managers - are now finding themselves without work.

Or think of OP’s “voice acting” topic: most voice actors do not live off of highly artistic and controlled jobs for games or movies (those pay well, but very unfrequently) but earn their money by doing contracting work in marketing areas. Commercials, corporate communication and stuff like that. This is done by elevenlabs and others way cheaper at the moment. Without the need of additional middle-men like sound engineers and editors. Or even an agency or production company in between.

It is a major shift - not just culturally but mainly microeconomically.

People without jobs don’t earn money and can’t buy stuff. And don’t pay taxes. And don’t pay social securities (in countries in which this applies) so don’t stabilize the system they are in. This is bad. Very bad.

Coming back to the main topic: At the current scale and global complexity this is a very serious issue. We have had a lot of “unprecedented” crashes in the last two decades but this tops it all. Combined with the general instability right now this might not just be a simple bubble but a proper crash - and, as Littlefinger said: “Chaos is a ladder.”

In some sense, it is also a little race of global politics and the world economy. The US economy highly impacts the global economy. The crash will most likely happen there. To mitigate the impact, governments (and companies) will have to get a bit of distance to soften the blow. I think, we can see the opposite. Most are investing into the US market and growing dependencies to US companies.

Last thoughts on what one could do on an individual basis: think local first. Set up your own data infrastructure if you can or learn it. A little homeserver goes far: play with Nextcloud, with Immich, with your own Git solution. Give Linux a proper chance. If you have investments in stocks, double check which might be vulnerable and consider letting go of them. And support your local markets by buying more directly there instead of ordering everything from amazon. Get yourself something semi-expensive but you can still afford that makes you happy and will last a little longer (memories included). This, at least, is my opinion and general expectation. There certainly are different approaches to this. You could also read up on Peter Thiel and embrace one version of a future yet to come. :smiley:

2 Likes
6

The Internet created a bubble, but wasn’t itself a bubble. The Internet had obvious, tremendous fundamental value. The bubble consisted of companies plugging anything and everything into an Internet model without any business plan at all.

AI, on the other hand, has little to no fundamental value, AND companies pushing anything and everything into it have no business plan that isn’t illegal in any other context. Unlike the Internet, AI is a classic bubble kept alive only by the Fear Of Missing Out. Once the fear subsides, the AI bubble will burst like the Hindenburg.

2 Likes
7

@soapspangledgames, what’s your guesstimate on when the AI bubble will pop? I know it won’t be accurate, but I’m curious of what you think?

My prediction is sometime between 2028-2030.

Everyone else is free to chime in as well.

1 Like
8

The AI bubble has been defined as the additional investment caused by the overstatement of the value of AI.
This will likely burst when the assessment of value is more solidly understood.
I would approach the problem of value assessment by looking at what problems AI actually solves first, not as a critic, as an optimist. These include (a) reduced writing time for articles, books, screenplays, etc, (b) reduced time developing software, (c) digital media production. Essentially everything humans can do on computer can be accelerated.
The value of saved time is measured in wages per hour, and since AI system integration is new, there must be some time before it becones widely adopted, and the landscape of attitude can become clear. Also the rate of change or improvement of the qualities that provide labour saving also needs to be assessed. An example of why is the Alpha Zero chess program that was apparently able to learn chess in matter of minutes (citation).
The value of AI in a successful company could be less than the total pay saved per measurable unit workload, otherwise theres no profit.
So the question of when or if the bubble will burst can boil down to whether using the AI, (generally meaning any of the LLM type or deep learning systems types), actually saves time. The presence of AI in industry now really changes everything … for example I read an article recently saying Meta had slashed funding on the Metaverse, and the article suggested it was a multi-billion dollar waste of money, and they had given up on the idea … i think maybe costs are lower because AI makes it easier to develop.
Then eventually the entire computer bubble might burst

1 Like
9

I feel that, this was when my first company went bust too, I had 14 employees and my life savings invested, plus borrowing. It was not a nice time!

It is magical. Remember the Turing test? I think AI passes that and we all just moved on and forgot about it!

The problem it solves is not the toys for us to play with, mostly, it is for industry to cut the wage bill! Yes jobs are created, but so many more are going to go. And just wait till a competent robot can have AI controlling it, bye bye to an entire working class. The problems it causes and will cause could well be horrendous in the extreme. Imagine when countries can field armies of robots, it doesn’t feel too far off to me. And who is the worlds manufacturer today? Not a nice democratic country by any means.

Back to the OP:

Is the AI Bubble about to burst?

Yes, and in the funniest way possible. The billionaire overlords are destroying themselves without even realising it. AI is destroying the very businesses that gave them power in the first place. I just hope they don’t take us all down with them! I am pretty sure though that their fall will only be painful for everyone. I mean, in the end, who exactly do you think is going to pay for all of this. It will be us, the hard working, much shat on, tax paying work force. Its the enshitification of society. And it ‘could’ have been such a different story. AI did not have to be the baddie.

And the saddest thing about all this is that the predictions about global warming and the wars it will cause is all being swept under the “AI will solve it” rug!

I love AI, it is amazing. I also hate it. Or more precisely, I hate what is being done, and is going to be done, with it - by the rich who just want to be richer! (Yes that is my hyphen, not an AI hyphen!)

I think sooner. But it will not be a burst, it will be a deflating anti-climax as all the air is slowly sucked out when they realise there is no-one left to pay for anything.

Just a final note, I am now officially part of the “BUY NOTHING” revolution! I have just three or four subscriptions left that I just can’t see how to live without, and I simply only buy essential items, second hand if possible. It is simply amazing how much you can save when you step off the commercial consumerism treadmill, escape the pyramid schemes and empty promises of vapour ware salesman and start to find genuine fulfillment that has nothing to do with where you eat or what you do.

PS And my cat died yesterday, passed quietly in her sleep, she was 15 and a half years old, and I miss her terribly. Has nothing to do with AI, just thought I would share.

8 Likes
10

My guess won’t be accurate, as you stated. In my opinion, it should have never gotten this far considering it’s near total uselessness. Given that untold billions are still being poured into it, I doubt it will burst in 2026. I think your estimate of 2028-2030 is plausible.

2 Likes
11

I feel for the loss of your cat. :frowning:

2 Likes
12

My condolences.

2 Likes
13

Agree to disagree. This is semantics. It was the DotCom bubble. Whether you call it the AI bubble or the LLM bubble, or something else - the two are similar. Both are technologies where speed of promise outpaces speed of implementation.

I started studying how to create LLMs in 2022, learned PyTorch, and statistical math. LLMs have lots of great applications when trained appropriately. The technology is sound - just like the internet. It’s just that everyone is throwing money at it now hoping to make it big. Some will, most won’t.

As for when it will happen? It may not in the same way it did in 2000 - specifically because the companies that would be affected are financially too big to fail. Like the banks in 2008. Or the banks in 2023.

4 Likes
14

Yeah that makes total sense, because this tech (maybe unfortunately) has a place in the world, but it definitely shouldn’t be everywhere.

(And yeah I cannot get over the fact that the next windows update has File Explorer preloaded and on top of that has Copilot integrated in it.)

2 Likes
15

Interesting way to see it!

The 2001 lay offs are an event I (maybe luckily?) missed out by some time, that must have been a stressful time for you, 9 months searching for a job are definitely a lot.

But I think the driving factor in the AI economy right now is the Fear Of Missing Out on this new tech that might be hyping it up excessively when it doesn’t really have that high of a general value compared on what the internet was.

It definitely won’t go away, but where it’ll stay will be much different from where the internet is right now, because the internet is effectively everywhere (even quite literally as you can access it in most places).

2 Likes
16

Was file explorer THAT confusing to the Windows devs?!?! Confusing enough to where you need to ask an AI to find your files that you yourself made!

Is it gonna organize my files that I already organized, but then proceed to do it wrong, making me have to manually organize files anyway. Is the AI gonna name my files for me that I already have a naming system for? I bet it’s gonna add a dumb emoji too :zany_face:.

I don’t even have that many files on my computer! Seriously, am I gonna ask file explorer for emotional support now?

3 Likes
17

Exactly what I was thinking, it doesn’t make any sense to have an AI there.

They should have focused on making File Explorer faster (Linux, at least on the command line, is super quick in finding files and even viewing its contents, while file explorer takes many minutes, even though searching on a command line is different, but still).

I wish that the tech industry didn’t rely a lot on these proprietary Operative Systems (yeah the industry uses Linux too, but for web servers) and instead started using Open Source alternatives which are much more efficient for work.

The only reason I haven’t switched to Linux yet (using it on a VM though) is the amount of apps I use daily that are windows only.

1 Like
18

Yeah, just let the AI name and organize my highly sensitive game files. That’s a GREAT idea. (Sarcasm)

I’m still on Windows 10 with a free pass for this year’s security updates. I really don’t want to switch to Windows 11 and get used to a new OS until I absolutely have to.

I don’t really think about what OS I use because I’m not much into computer science. I just want something that works. I don’t see the appeal of Mac, I think Linux is too complicated, and ChromeOS can suck my left one.

Windows 10 just works for me right now, because I’ve been using it for the better part of a decade and it’s the standard for PC software development.

2 Likes
19

Windows 10 is a fantastic OS in itself except some general defects of windows itself.

Dude imagine working years on monkanics and then the project stops working because Copilot thought it would be funny to rename a file “FunnyMonke01” lol.

I don’t really get the appeal on Apple products as a whole, it might be because don’t think I’d be able to use a device where I cannot install whatever I want (including apps and games made by me) without having to do backflips one side loading. Even if though I can see the simplicity that the Apple Ecosystem have and some people, including elders, probably appreciate that.

2 Likes
20

Me neither.

I think the best comparison to Apple’s ecosystem is Nintendo. It’s a closed ecosystem; MacOS/IOS is on Apple devices only and Nintendo games are on Nintendo consoles only.

This creates a lot of brand loyalty. Switching would mean you’d lose out on Apple’s/Nintendo’s exclusive offerings.

Also, it’s the fact that Apple and Nintendo have first movers advantage. Even decades later this still applies. Apple revolutionized the smartphone, which despite me having an android, is literally on my desk right now. And Nintendo has the most iconic video game IP’s of all time.

This means if your a fan of Apple/Nintendo, you are a DIE HARD fan of Apple/Nintendo. And that’s enough to make serious bank.

1 Like