At Naked Capitalism Yves Smith published a paper by Servaas Storm:
The AI Bubble and the U.S. Economy: How Long Do “Hallucinations” Last?
Yves writes:
This is a devastating, must-read paper by Servaas Storm on how AI is failing to meet core, repeatedly hyped performance promises, and never can, irrespective of how much money and computing power is thrown at it. Yet AI, which Storm calls “Artificial Information” is still garnering worse-than-dot-com-frenzy valuations even as errors are if anything increasing.
Storm’s introduction:
This paper argues that (i) we have reached “peak GenAI” in terms of current Large Language Models (LLMs); scaling (building more data centers and using more chips) will not take us further to the goal of “Artificial General Intelligence” (AGI); returns are diminishing rapidly; (ii) the AI-LLM industry and the larger U.S. economy are experiencing a speculative bubble, which is about to burst.
I happen to a agree with the arguments and conclusion.
The current Large Language Models are part of the Generative Artificial Intelligence field. GenAI is one twig on the research tree of Artificial Intelligence. LLMs are based on ‘neural networks’. They store billions of tiny pieces of information and probability values of how those pieces relate to each other. The method is thought to simulate a part of human thinking.
But human thinking does much more than storing bits of information and statistical values of how they relate. It constantly builds mental models of the world we are living in. That leads to understanding of higher level concepts and of laws of nature. The brain can simulate events in those mental model worlds. We can thus recognize what is happening around us and can anticipate what might happen next.
Generative AI and LLMs can not do that. They do not have, or create, mental models. They are simple probabilistic systems. They are machine learning algorithms that can recognize patterns with a certain probabilistic degree of getting it right. It is inherent to such models that they make mistakes. To hope, as LLM promoters say, that they will scale up to some Artificial General Intelligence (AGI) know-all machines is futile. Making bigger LLMs will only increase the amount of defective output they will create.
(Yesterday I watched a video of Jon, a baker in Mesa, in which he mentions how he had asked an LLM to half a recipe he was going to make. It did that correctly for all but one ingredient. The model had divided the amount of water needed by ten. Jon’s test bake had failed.)
But the hype around LLms is real and huge amount of money is flowing into the companies that are building such models. This while none of them has found ways to create sufficient revenue to support such investments. Training and running these models at scale is very expensive. There are simply too few real use cases that would justify paying the cost for them. It may be fun to create and play around (archived) with AI-slop videos on social media. But who is willing to pay for that? Especially when the use of social media is finally sinking (archived).
(For a more detailed discussion of LLMs, their costs, lack of use cases, and on the incestuous structures of the investments that are flowing into them see Edward Zitrons 18,500 words epos here: The Case Against Generative AI.)
There are still hundreds of billion dollars flowing into the already overvalued LLM hype:
AI startups’ aggregate post-money valuation (the valuation after the latest round of funding) soared to $2.30 trillion, up from $1.69 trillion in 2024, and up from $469 billion in 2020, which back then had already set a huge record, according to PitchBook.
…
OpenAI reached a $500 billion valuation in early September, when it offered eligible former and current employees to sell $10 billion of their shares in a secondary share sale to other investors, led by SoftBank, according to CNBC. In April, OpenAI had reached a breathtaking post-money valuation of $300 billion at a funding round when it raised $40 billion, primarily from SoftBank. The sky is not the limit.Elon Musk’s xAI is supposedly shooting for a $200 billion valuation in a $10 billion funding round, according to sources cited by CNBC, which Musk denied on X as “fake news. xAI is not raising any capital right now.” Well, not right now. Or whatever.
Anthropic reached a $183 billion post-money valuation, after raising $13 billion in a Series F funding round in early September, according to Anthropic.
And so on. These valuations of AI startups are mind-boggling. How are these late-stage investors going to exit their investments with their skin intact?
They wont.
Dozens of specialized LLM data-centers are getting build to house a huge amount of expensive chips that lose their values faster than a newly bought hyper-car. All without a real use case for LLMs and without any hope for sufficient revenue to ever sustain the business.
This is bad for the U.S. economy.
The money that is flowing into the LLM hype is gone. It can not be invested somewhere else even when that would make way more sense for the larger society – for example in the revival of manufacturing or in apprenticeship programs. Like during the dot.com boom (archived) in the late 1990s the real economy gets crowded out by a virtual one. Trump’s tariffs will not lead to the revival of U.S. industries if there is no money left to invest in them.
The data-centers being build will need huge amounts of additional electricity which can not be generated within the foreseeable future:
The implications are brutal and stark. Curtailed and costly electricity supply for AI and manufacturing will impair American economic competitiveness, with knock-on effects for household affordability. These impacts are already becoming evident, with wholesale pool prices in the U.S. rising by 267% over the past 5 years, on the back of skyrocketing electricity demand from the AI sector (Bloomberg).
All recent U.S. stock market gains were powered by the LLM hype. When the bubble will burst, the stock market will sink and most people who are, directly or indirectly, invested in the LLM hype will lose a lot of their money.
Unfortunately there is no way to foresee when that will happen or how far the damage will spread.
But we can already see damage to the real economy. Investment in factories for real products gets crowded out and electricity prices are doubling and tripling, hitting manufacturers as well as private consumers.
Why don’t we have ways to prevent bubbles? Or why can’t we deflate them before they become threats to our societies?