The LLM bubble

[NickZ]

The core of the article names the fear of an upcoming bubble, on AI.

At the same time, key players in the industry acknowledge there is probably pain ahead for AI investors. OpenAI Chief Executive Officer Sam Altman said this week that he sees parallels between the current investment frenzy in artificial intelligence and the dot-com bubble in the late 1990s. When discussing startup valuations he said, “someone’s gonna get burned there.” And a Massachusetts Institute of Technology initiative released a report indicating that 95% of generative AI projects in the corporate world have failed to yield any profit.

Altogether, it’s enough to make credit watchers nervous.

http://archive.today/oZmKO paywall free article from https://www.bloomberg.com/news/arti...e-ai-boom-and-fears-of-a-bubble-credit-weekly

The LLM was welcomed mostly by the curious crowd of web savvy people. Most of those do just like their followers try out the free version. Similar to the intent to sell email accounts, the intent to sell the LLM services trying to convert viewers/users into paying customers.

Some firms have created a LLM which they use, customer chat-bot for the Media Market Group in Germany, quite recently sold to a Chinese Group. The bot works quite nicely, a friend of mine told me.

https://www.reuters.com/business/retail-consumer/chinas-jdcom-buy-germanys-ceconomy-deal-valuing-it-25-billion-2025-07-30/

How many firms in the UK have succeeded in implementing their LLM?

 

[fisicx]

Old news. The tech industry has been saying this for a while now.

 

[NickZ]

Old news. The tech industry has been saying this for a while now.

How many firms in the UK have succeeded in implementing their LLM?

The question went unnoticed?

 

[fisicx]

The question went unnoticed?

Does it matter? There are thousands of businesses using Agentic and Generative AI. Some will be getting better results than others. I doubt anyone has built their own LLM.

 

[NickZ]

Does it matter? There are thousands of businesses using Agentic and Generative AI.

According to different investigations many use it and most don't have adopted LLMs enough to make their use profitable.

MIT Report is from August 18, hopefully not just put away as old news:

MIT report: 95% of generative AI pilots at companies are failing | Fortune

There’s a stark difference in success rates between companies that purchase AI tools from vendors and those that build them internally.

fortune.com

 

[ThatDevAaron]

Does it matter? There are thousands of businesses using Agentic and Generative AI. Some will be getting better results than others. I doubt anyone has built their own LLM.

+1

Making your own LLM is just too expensive, AI is one of those industries, if you can't burn cash, you can't compete in AI.

When I say BURN CASH, I really mean BURN CASH without hopes of recouping for a long damn time.

 

[Ozzy]

When I say BURN CASH, I really mean BURN CASH without hopes of recouping for a long damn time.

I have successfully built an AI solution that is used in my daughter's business and is time-saving and revenue-generating. I've posted about it before, so no need to recycle the content. This same AI solution is now being looked at by some colleges and schools, and one Local Authority. Not really sure whether I'll give it away or charge at the moment, as I only did it to help my daughter out.

I am now in the planning stage of a new venture. Working with a business partner, I am building an AI service for a specific part of the medical industry with a view to it being distributed through the NHS. I'm not looking to generate profit from this; I'm doing it for philanthropic purposes, so burning cash is part of the plan. We're building a custom LLM with detailed medical knowledge and agents to connect to external platforms as needed, and a custom CSM that will communicate with patients in the appropriate tone.

 

[ThatDevAaron]

I have successfully built an AI solution that is used in my daughter's business and is time-saving and revenue-generating. I've posted about it before, so no need to recycle the content. This same AI solution is now being looked at by some colleges and schools, and one Local Authority. Not really sure whether I'll give it away or charge at the moment, as I only did it to help my daughter out.

I am now in the planning stage of a new venture. Working with a business partner, I am building an AI service for a specific part of the medical industry with a view to it being distributed through the NHS. I'm not looking to generate profit from this; I'm doing it for philanthropic purposes, so burning cash is part of the plan. We're building a custom LLM with detailed medical knowledge and agents to connect to external platforms as needed, and a custom CSM that will communicate with patients in the appropriate tone.

Can you link the post?

 

[fisicx]

According to different investigations many use it and most don't have adopted LLMs enough to make their use profitable.

You seem to be very confused with the terminology. The LLM is what the likes of OpenAI use to answer your prompt. Businesses don’t (usually) adopt LLMs.

 

[NickZ]

You seem to be very confused with the terminology. The LLM is what the likes of OpenAI use to answer your prompt. Businesses don’t (usually) adopt LLMs.

A large language model is a language model trained with self-supervised machine learning on a vast amount of text, designed for natural language processing tasks, especially language generation. The largest and most capable LLMs are generative pretrained transformers, based on a transformer architecture, which are largely used in generative chatbots such as ChatGPT, Gemini and Claude. LLMs can be fine-tuned for specific tasks or guided by prompt engineering.

LLMs are called AI by the hype people.

Large language model - Wikipedia

en.wikipedia.org

 

[Ozzy]

LLMs are called AI by the hype people.

I probably wouldn't refer to Wikipedia for answers here; speaking as someone who spent a few years as a Wikipedia moderator.

AI is a combination of a series of modules. The LLM is purely the dictionary with some rules that dictate tone, language and basically what words go with what and what they mean, so that the AI knows how to communicate.

 

[fisicx]

LLMs are called AI by the hype people.

All sorts of junk is called AI. But still doesn’t change the fact that business do not adopt LLMs. They adopt tools that use an LLM.

 

[NickZ]

They adopt tools that use an LLM.

Businesses are wondering why the AI (which does not exist) is not bringing the money they expected.
The wondering is not new. It happened with all "new" tech coming along.

Now after a few fruitless years the pressure has risen to make it work. Almost 70% of all ads I saw lately have an AI connection. Almost all browsers now work with AI (preprogrammed Software as they where called in the beginning)

Too much money was thrown at it and some people are asking the obvious question. Why hardly anyone turns it into a money making machine. Companies collecting billions to build data-center don't give out numbers of paying customers. Just the amounts every round is collecting.

Faith in God-like large language models is waning

archived 8 Sep 2025 22:15:09 UTC

archive.ph

Original https://www.economist.com/business/2025/09/08/faith-in-god-like-large-language-models-is-waning

WHEN TECH folk talk about the lacklustre progress of large language models (LLMs), they often draw an analogy with smartphones. The early days of OpenAI’s ChatGPT were as revolutionary as the launch of Apple’s iPhone in 2007. But advances on the frontier of artificial intelligence (AI) have started to look akin to ho-hum phone upgrades rather than genuine breakthroughs. GPT-5, OpenAI’s latest model, is a case in point. It has generated even less buzz than is expected from the iPhone 17, Apple’s newest release, which is due to be unveiled on September 9th.

Ending in:

For now, most attention remains on LLMs. Though cloud giants like Microsoft and Google have built extremely small models, the belief that the status quo will persist has helped justify the fortunes they are spending on data centres to train and host the largest models. That may be short-sighted. Given the relative merits of SLMs, Apple’s take-it-slow approach may turn out to be justified in the long run.