[NA]AI bubble

Let'sgoflying!

Let'sgoflying!

Touchdown! Greaser!
Joined
Feb 23, 2005
Messages
21,872
Location
west Texas
Display Name
Dave Taylor
Home Airfield
KMRF
Do you think we are close to one?
Are you getting out, or planning to short?
 
I would say that we are entering the beginning of a bubble, but the bubble can inflate for a while before it bursts. Lots of companies want to jump in this AI thing, but don't know how to do it or what to automate.
Lots of very advanced tooling are being released, but companies are struggling with basic use cases still. So there's room for growth.

Agentic AI frameworks are stabilizing now and the platforms are getting a lot better. This aspect of AI is what's going to grow like crazy for the next 2-4 years. I don't see this bubble bursting yet.
 
Let'sgoflying! said:
I have no answer & AI only offers generalities, no specifics
Click to expand...
Yeah
About a year and a half ago I was looking for public companies that I could take the opposite approach to today's discussion and jump on and ride the hype wave a little, but found that either most of them were privately-funded, or the already-large usual suspects like Microsoft or Google.

I do have holdings in some [more physical] automation/robotics companies (think manufacturing and medicine) but those aren't really part of the gen-ai craze. I'm still long on those.
 
Let'sgoflying! said:
Did it say this when you did it?
Click to expand...
Actually I didn't bother. I would not trust the response anyway.

The thing is, yes AI is overvalued at least based on its current use case. However that is also true for almost any new promising technology.
 
The bubble will burst right about the time companies lay off half their software developers thinking AI will replace them.
 
dcat127 said:
Actually I didn't bother. I would not trust the response anyway.

The thing is, yes AI is overvalued at least based on its current use case. However that is also true for almost any new promising technology.
Click to expand...
I have a boss who is constantly harping at me to buy XRP cryptocurrency because it's going to be such a huge valuation shift for the US and banks to replace SWIFT. Keeps saying it's gold-backed (it isn't). Says if it goes the way it's supposed to go (according to whatever insider info he things he has) he might retire after the first of 2026. He keeps saying he has "researched" this stuff for several years. The problem is, I don't think he'd understand what/how blockchain works even if he read the source material on it. His research is largely based off of whatever articles/links he's clicked on that are obviously advertising the crypto. I hope it works out for him, but it reads like a pyramid scheme to me, lol.

It seems similar with AI where the people predicting how world-changing it is for businesses at this moment are trying to apply it way before it's developed enough to be useful. Right now, it's often too susceptible to bad data, which ends up with GIGO. I think we're two decades out from having AI really be a threat to anyone's jobs.
 
Define AI bubble. Like many things, it's probably oversubscribed right now. Too many people attempting to profit off too little unique technology. I predict those that this is their only business will fail for lack of income or get absorbed into larger entities. Those like Meta and Alphabet will shrug and say, "Well this isn't making us as much as we thought it would, but it does help the business as a whole."
 
flyingron said:
Define AI bubble. Like many things, it's probably oversubscribed right now. Too many people attempting to profit off too little unique technology. I predict those that this is their only business will fail for lack of income or get absorbed into larger entities. Those like Meta and Alphabet will shrug and say, "Well this isn't making us as much as we thought it would, but it does help the business as a whole."
Click to expand...
Initial hype reminds me of the transparent efforts of companies trying to capitalize off of crypto any way they could. (Though I think certain applications of gen AI serve useful purposes and are here to stay)

My wife's company (a public company) redid their corporate logo after they found that part of two of the letters could be highlighted to spell out "AI". I'm not lying :rofl::rofl:
 
Core AI technology isn't a bubble, but a lot of the peripheral businesses trying to hop on the bandwagon are a bubble. Can't go more than a day or two without seeing a company raise money for a dubious biz model based on AI enabled tech.

I think the only downside BigTech has is that they did these massive capital outlays for AI, and then their product gets commoditized almost the second it's released. That's great for customers, bad for them.
It's why all the AI stuff I build (and that's pretty much 75% of my work these days) is setup to integrate with all the major providers. Switching costs are zero. Performance is comparable between vendors. Similar model params are usually toggleable for control.
That means all you gotta do is throw a switch to flip models -- as a result, none of them have a lot of pricing power.

IMHO: the big winners are the people selling the hardware and infrastructure. They get paid whether all this works out or not. And since it's a massive arms race at both the corporate and sovereign level -- the future looks pretty bright for them.
 
ArrowFlyer86 said:
That means all you gotta do is throw a switch to flip models -- as a result, none of them have a lot of pricing power.

IMHO: the big winners are the people selling the hardware and infrastructure. They get paid whether all this works out or not. And since it's a massive arms race at both the corporate and sovereign level -- the future looks pretty bright for them.
Click to expand...
Those are both important observations.

Microsoft is microsoft because once you adopted windows (and to a lesser degree, MSDOS) it was extremely expensive to switch. There is, at least currently, almost zero vendor lock in on the AI in common use. Further, it's hard to see where lock in could be built into a model. In recent years Microsoft got its second or third wind by becoming an enterprise company (thanks to Balmer). I just don't see what the equivalent move could be in AI.

OTOH, when you're doing a data center build to power AI... That's real cost and real lock in.
 
172andyou said:
Those are both important observations.

Microsoft is microsoft because once you adopted windows (and to a lesser degree, MSDOS) it was extremely expensive to switch. There is, at least currently, almost zero vendor lock in on the AI in common use. Further, it's hard to see where lock in could be built into a model. In recent years Microsoft got its second or third wind by becoming an enterprise company (thanks to Balmer). I just don't see what the equivalent move could be in AI.

OTOH, when you're doing a data center build to power AI... That's real cost and real lock in.
Click to expand...
Lock in could be integration with existing workflows or apps. Take CoPilot - easy integration with 365, for example. As soon as people/teams build agents to perform tasks using existing applications and have established workflows using those agents then there is a cost to migration.
 
Baked Potato said:
Lock in could be integration with existing workflows or apps. Take CoPilot - easy integration with 365, for example. As soon as people/teams build agents to perform tasks using existing applications and have established workflows using those agents then there is a cost to migration.
Click to expand...
As @ArrowFlyer86 points out, the integration points are mostly swappable. If Claude is giving the best results for my agent today and ChatGPT tomorrow, it should be pretty trivial to replace it. So AI vendors will be forced to compete based on capability and cost instead of lock in.
 
Baked Potato said:
Initial hype reminds me of the transparent efforts of companies trying to capitalize off of crypto any way they could. (Though I think certain applications of gen AI serve useful purposes and are here to stay)
Click to expand...
I always said that anybody who came up with a problem that blockchain was the solution to, either didn't understand the problem or what blockchain was or both.
 
172andyou said:
As @ArrowFlyer86 points out, the integration points are mostly swappable. If Claude is giving the best results for my agent today and ChatGPT tomorrow, it should be pretty trivial to replace it. So AI vendors will be forced to compete based on capability and cost instead of lock in.
Click to expand...
I see - but for fear of conflating the models vs the agents/clients, I believe the former is easily swappable while the latter may lend itself more to vendor lock-in. One can easily build an app where you can swap LLMs at will, but managing integrations with the end user software, the APIs, security models, etc may pose a greater challenge.
 
Baked Potato said:
Lock in could be integration with existing workflows or apps. Take CoPilot - easy integration with 365, for example. As soon as people/teams build agents to perform tasks using existing applications and have established workflows using those agents then there is a cost to migration.
Click to expand...
In addition to this is the trust/security factor. I just don't have the same level of trust in an AI company that started in the last few years and is "moving fast and breaking things" to prioritize the security of our data over one whose business model depends on it (e.g. Microsoft).
 
Baked Potato said:
I see - but for fear of conflating the models vs the agents/clients, I believe the former is easily swappable while the latter may lend itself more to vendor lock-in. One can easily build an app where you can swap LLMs at will, but managing integrations with the end user software, the APIs, security models, etc may pose a greater challenge.
Click to expand...
There is probably at least a kernel of truth in that. Where I see it differently is that I see the LLM vendor as the AI vendor and the application vendor as an AI user. Finally the human using the application as an end user or indirect user. So I see the LLM vendor as a commodity. The application vendors often already have lock in, and will continue to have that. And the end users will just see their applications get new features.

For example, I don't see ChatGPT replacing Salesforce. And I see Salesforce being able to use ChatGPT, Claude or any other LLM to provide features in the salesforce applications. Salesforce, of course, has amazing lock in. But that won't benefit the LLM providers.

Of course, the AI world is way bigger than just LLMs. But since they are the ones consuming all the oxygen right now...
 
Sam D said:
In addition to this is the trust/security factor. I just don't have the same level of trust in an AI company that started in the last few years and is "moving fast and breaking things" to prioritize the security of our data over one whose business model depends on it (e.g. Microsoft).
Click to expand...
I don't think you're wrong, but there is another way of viewing this. You could also say, "I don't have the same level of trust that an inertia bound bureaucracy like microsoft is going to be able to execute an AI plan, with new technology and different skills, better than a nimble startup that doesn't have the baggage of a large company culture".

We should know which of these two viewpoints is correct in about a decade. :D
 
Sam D said:
In addition to this is the trust/security factor. I just don't have the same level of trust in an AI company that started in the last few years and is "moving fast and breaking things" to prioritize the security of our data over one whose business model depends on it (e.g. Microsoft).
Click to expand...
Nothing epitomizes the wisdom of this mindset like looking at the crypto fintech start-ups. All of them were your best friend and playing above board until the dominoes started falling and the executives were hauled off to prison (FTX, Celsius, Terraform, CZ, ..).
If you're building an application that will handle sensitive data, you might take slightly lower performance if it means you can sleep soundly at night knowing your major vendor is not doing something egregiously bad w/your information.
 
172andyou said:
I don't think you're wrong, but there is another way of viewing this. You could also say, "I don't have the same level of trust that an inertia bound bureaucracy like microsoft is going to be able to execute an AI plan, with new technology and different skills, better than a nimble startup that doesn't have the baggage of a large company culture".

We should know which of these two viewpoints is correct in about a decade. :D
Click to expand...
You're absolutely correct. But @ArrowFlyer86 nails it when he says we'll accept lower performance to sleep better at night.
 
flyingron said:
I always said that anybody who came up with a problem that blockchain was the solution to, either didn't understand the problem or what blockchain was or both.
Click to expand...

Heh. I was in a meeting last week working on building a go to market strategy for an AI solution and I texted one of my co-workers and said "This reminds me of the blockchain days - everybody wants it, but nobody knows what it is or does."

I think some version of AI will stick around - barrier to entry is pretty low for menial and easily scripted type tasks already. Where the real benefit will be, is whether the models can be used to infer based on inputs from multiple disjointed data sources (thinking about shortening the time it takes to research an issue - technical or business). I don't see it taking out entire software development or marketing departments as some slideware may lead you to believe, though.
 
Lindberg said:
I'm not worried about AI yet, I'm still counting on my metaverse investments to pay off. I'm starting to lose faith in Beanie Babies though....
Click to expand...
I think Beanie Babies peaked when they bought this - N888TY, the BBBBJ:

0288221.jpg
One of the better looking Boeing paint jobs I've ever seen.

flyingron said:
I always said that anybody who came up with a problem that blockchain was the solution to, either didn't understand the problem or what blockchain was or both.
Click to expand...
There are lots of very smart people doing fascinating work on cryptographic voting protocols. We should be funding and encouraging them, and doing all our elections with paper ballots until everyone currently working in that field has retired.


Blockchain
 
Interesting data point for this discussion. Was talking to a young lady yesterday who is making a music video for her up and coming band. We are in the ATL area, which has become a hot spot for all things related to the film industry over the past several years. I asked her if she was able to get a production team locally (outside the metro) or if she has to go INTO ATL. She said that right now the worker bees in the movie industry are struggling so bad that she isn't having to search hard to find people willing to work on her piddly music video. One aspect is that the production companies are moving physical production out of country to beat the union costs which isn't a new concept. The really interesting one is that the 'soft' skill folks - everything from script writers and special effects engineers to actual background actors & stunt doubles, etc. - are being replaced by AI. She said she is going to be using actors, camera & lighting operators, etc. with incredible resumes in her local-band music video because they are so hungry to have ANY work right now.

So maybe it's making more of an impact that I expected...
 
CJones said:
Interesting data point for this discussion. Was talking to a young lady yesterday who is making a music video for her up and coming band. We are in the ATL area, which has become a hot spot for all things related to the film industry over the past several years. I asked her if she was able to get a production team locally (outside the metro) or if she has to go INTO ATL. She said that right now the worker bees in the movie industry are struggling so bad that she isn't having to search hard to find people willing to work on her piddly music video. One aspect is that the production companies are moving physical production out of country to beat the union costs which isn't a new concept. The really interesting one is that the 'soft' skill folks - everything from script writers and special effects engineers to actual background actors & stunt doubles, etc. - are being replaced by AI. She said she is going to be using actors, camera & lighting operators, etc. with incredible resumes in her local-band music video because they are so hungry to have ANY work right now.

So maybe it's making more of an impact that I expected...
Click to expand...
That's an interesting data point on the music side. The creative folks got burned pretty bad.

I have a friend who worked as a professional digital designer for 15 years. Had credentials including working directly for Oprah and a number of other big name folks. She's not some artsy, tattooed up free spirit fighting the man (artists do have a reputation... ) -- but a college educated, talented artist willing to do the whole corporate routine in order to pay the bills. After 2 years of virtually zero work (it all went to AI) she's given up completely and is now training to become a bartender. You do not want to ask her how she feels about AI unless you have an hour or two to spare.

Another rough one: my brother. He's a town reporter for local/regional publications. He's a good guy and works damn hard to write high quality stuff (a lot harder than I'd work for what reporters get paid...). Won a bunch of awards; any small/mid sized town would be lucky to have him (which is where he's focused before). He got displaced by AI writing these boring articles that cannot ask any inquisitive questions (AI can't file a FOIA request, it can't ask questions about someone's accident, can't capture the feeling at their local music festivals, etc). In the meantime he's taken a low-paying hourly job doing something different to make ends meet while he continues looking for opportunity in his field.

AI definitely has an impact.
 
Last edited:
You must log in or register to reply here.
Back
Top