Bubble or not, the AI backlash is validating one critic’s warnings

This was followed by a technology sale, as abandoned investors sent the value of the S&P 500 by a decrease of $ 1 trillion. Looking at the increasing dominance of this index by technology shares that have largely turned into Amnesty International’s shares, the nerves were that the mutation of artificial intelligence was turning into Dotcom Bubble 2.0. Certainly concerns about artificial intelligence trade are not the only factor for market transport markets, as it is clear from the S&P 500 that picks up a five -day loss series on Friday after the almost mysterious Jerome Powell comments in Jackson Hall, Wyoming, to a hint from the Federal Reserve Chair towards cut off the interest rate on any September.

Gary Marcus has warned of the limits of large LLMS models since 2019 and a warning of the potential bubble and doubtful economy since 2023. His words have a special weight in particular. The cognitive scientist who has turned into a long -time researcher in automatic learning since 2015, when he established engineering intelligence. This company was obtained by Uber in 2016, Marcus left shortly, and worked in other startups of artificial intelligence while providing audio criticism of what he sees deadlines in the area of ​​artificial intelligence.

However, Marcus does not see himself “Cassandra”, and he does not try to be. luck In an interview. Cassandra, a character of the Greek tragedy, was a character that uttered accurate prophecies, but was not believed until it was too late. “I see myself realistic and as a person who expects problems and was right for them.”

Marcus attributes the swaying on the market to the GPT-5 above all. He said it is not a failure, but it is “disappointing”, “disappointment”, and this “really awakened many people. As you know, GPT-5 was mainly sold, such as AGI, and not only” he said: “It is not a terrible model, it is not as if it was bad, but” it is not the quantum jump that many people led. “

Marcus said this should not be news for anyone who is noticing, as he argued in 2022 that “deep learning strikes the wall.” Marcus certainly wondered frankly about an alternative when the IQ bubble turned. He said luck Certainly, “crowd psychology” is happening, and he thinks every day about John Mainard Keynes quote: “The market can remain a solvent for a longer period than you can remain rational”, or Wile E. Coyote from Looney Tunes after the road running off the edge of the abyss and suspension in the field, before it falls on the ground.

“This is what I feel,” says Marcos. “We are outside the abyss. This is not logical. We get some signs from the past few days that people will finally notice.”

Building warning marks

The bubble talk began in the heating in July, when the chief economist in Apollo, Torsten Slok, released widely reading and influencing Wall Street, an amazing account while screaming on the bubble advertisement. “The difference between the IT bubble in the nineties and the artificial intelligence council today is that the top 10 companies in the S& P 500 index today are more exaggerated than in the 1990s,” he wrote, warning that the front lineage forward and amazing fees for companies such as NVIDIA, Microsoft, Apple and Meta were discovered from its profits.

In the weeks that followed, the disappointment of GPT-5 was an important development, but not the only one. Another warning mark is the huge spending of databases to support all theoretical future demand for artificial intelligence. Slok also dealt with this topic, as it was found that the contribution of the data center investments to the growth of GDP was the same as the spending of consumers during the first half of 2025, which is noticeable because consumer spending constitutes 70 % of GDP. ((Wall Street MagazineChristopher Mims presented the account weeks ago. New York Times Introduction to August 19, on the pretext that “it is not certain how much the meaning of artificial intelligence.”

This is important about the face, according to the political world Henry Farrell, who argued in Financial times In January, Schmidt was a major voice that constitutes a “new consensus in Washington”, which is partly dependent on AGI “around the corner.” On his authority, Farrell said that Schmidt’s article shows that his previous collection of assumptions was “clearly collapsing”, while he was challenging that he was relying on informal conversations with the people who knew him in the intersection of foreign policy of the continuous current and technical policy. Farrell’s title for this post: “Twilight of Tech Unilingrality”. He concluded: “If AGI bet is bad, then many of the logical basis for this consensus collapses. This is the conclusion that Eric Schmidt seems to be coming.”

Finally, the atmosphere in the summer of 2025 turns into a reverse reaction from artificial intelligence. Daril West warned at Brockens In May, the wave of public and scientific opinion will soon turn against the masters of the universe in artificial intelligence. Soon after, Fast company The summer is expected to be full of “male intelligence”. By early August, Intuition The colloquial “Clunker”, which is widely applied to artificial intelligence incidents, has identified, especially in customer service that has become diluted.

History says: short -term pain, long -term gain

John Thornehil from Financial times She offered some perspective on the bubble question, and advised readers to prepare for a crash, but to prepare for the future “golden age” of Amnesty International. It highlights the construction of the data center – an amazing investment of $ 750 billion in Big Tech during 2024 and 2025, and part of a global fame that is expected to reach 3 trillion dollars by 2029. Thornegore turns into financial historians for some comfort and some perspective. Over and over, it shows that this type of profitable investment usually leads to bubbles, dramatic collision, and creative destruction – but the solid value is finally achieved.

It is noted that Carloota Perez documented this style in Technological revolutions and financial capital: the dynamics of bubbles and golden ages. Amnesty International has identified as the Fifth Technology Revolution that follows this style, which began in the late eighteenth century, and as a result, the modern economy has now become an infrastructure for railways and personal computers, among other things. Each one had a bubble and shattered at some point. Thornehil was not cited in this particular column, but Edward is a consultant who trusted similar patterns in classic Satan takes the maximumA noticeable book not only for its discussions on bubbles but to predict the Dotcom bouquet before it happens.

Owen Lamont from Acadian Asset Management in November 2024, when he said that the moment of the main bubble was passed: a large number of market participants saying that prices are very high, but they insist that they are likely to rise.

Wall Street banks are not largely invited to a bubble. Morgan Stanley has recently witnessed an enormous efficiency of companies as a result of Amnesty International: 920 billion dollars annually to the S&P 500. Investors have warned against expecting a period of “CAPEX” accompanying the establishment of the data center, but also confirmed that the adoption of artificial intelligence expands beyond expectations, pointing to the liquefaction of Chatgpt from Openai, Alphabet’s Gemini, and AI CRM.

Bank of America Research wrote a note in early August, before the launch of the GPT-5, where he saw artificial intelligence as part of the “sea change” the productivity of the worker that will lead to the “innovation installment” of the S&P 500 companies. , To convert people into operations, and that artificial intelligence will connect this. “I do not necessarily think it is a bubble in the S&P 500,” I told luck In an interview, before I add, “I think there are other areas in which the bubble is slightly similar to.”

SUBRAMANAAAN Smaller Companies and Perhaps Permanent Endowment as fields “may have been categorized with great force,” SUBRAMANAAAAAN MS and Perhaps Perhaps Potential Endowment as fields “Possibly have been ranked with great force,” She is also anxious about the risk of diving in databases as well, noting that this represents a transformation into the most assets approach, rather than the appearance of light assets that are increasingly characterized by the highest performance in the American economy.

“I mean, this is new,” she said. “Technology used to be very assets and spending money only on research, development and innovation, and now they spend money to build these data centers,” adding that it sees that it determines the end of the asset light, the presence of the high margin and converts it mainly into “very intense in assets and more manufacturing than it was.” From its point of view, it calls for a decrease in the stock market. When asked if it was a bubble, if not a correction, she said, “It has started to happen in places,” and she agrees to comparison with the railway mutation.

Mathematics and ghost in the device

Gary Marcus also referred to the basics of mathematics as a cause of his anxiety, with a value of approximately 500 artificial intelligence worth $ 2.7 trillion. “This is not logical in relation to the amount of revenues [in]He said: “Marcus was martyred by reporting a billion dollars in revenue in July, but it was not profitable. [invested]”

So if Marcus is correct, then why did people not listen to it for years? He said he warns people this For years, he also described it as a “naive gap” in his 2019 book Restart artificial intelligence And controversy in New Yorker In 2012, this deep learning was a ladder that would not reach the moon. For the first 25 years of his career, Marcus trained and practiced as a cognitive science, and learned about “the embodiment of the people who do it.” [they] Look at these machines and make a mistake in attributing them to a really unaware intelligence, and it is human that does not really exist, and they end up using them as a companion, and they end up thinking that they are closer to solving these problems more than they are already. He said he believed that the amplifier bubble is in a large part of it because of the human motivation to drop ourselves on things, which is being trained in the cognitive world not to do it.

“These machines may seem like human beings, but” do not work like you, “said Marcus. I mean, it is almost tragic. ”

“People love artificial intelligence technology because it looks like magic. He feels magic and falsely … the truth is that she has not changed the world much yet, but I don’t think it is something to be rejected.” She also became with her herself. “I really use Chatgpt more than my children. I mean, it’s interesting to see this. I use Chatgpt for everything now.”