Most people who are not deeply involved in the artificial intelligence frenzy may not be aware, but AI's merciless marching squealed on August 7th, even becoming smarter than humans and becoming a threat to humanity.
This was the most widely held day when AI company Openai released its GPT-5. This is a high-end product that the company has long promised, and has promised its competitors to launch a new revolution with this innovative technology.
As it happened, the GPT-5 was a bust. It turns out to be less user-friendly and less capable than Openai's arsenal predecessor in many ways. I got the same kind of lyzzyable error when answering the user's prompt, and it wasn't good in math (or worse).
AI companies are now really supporting the American economy and look very bubble-like.
– Alex Hanna, co-author, “ai con”
“The idea was that this growth would be exponential,” says Alex Hanna, technology critic and co-author of the essential new book (with Emily M. Bender of the University of Washington). “Instead, Hannah says, 'We're going to hit a wall.'
The results go beyond what so many business leaders and ordinary Americans expect and even fear AI penetration in our lives. Despite AI Labs not making profits, hundreds of millions of dollars have been invested from venture capitalists and major companies such as Openai's Google, Amazon and Microsoft.
Public companies are hoping to announce AI investments, assert the AI capabilities of their products in the hopes of turbocharged stock prices, and have been promoted as “DOT-COMS” in the 1990s to make investors look even more sparkly.
Nvidia, a powerful chip power maker of AI research, plays roughly the same role as the stock market leader who helps Intel Corp., another chip maker that played in the 1990s, helps support the stock's boule market.
If AI's promises turn out to be as many mi-pirae as Dot-Coms, stock investors could face painful calculations.
A lively rollout of GPT-5 could bring the day of calculation closer. “AI companies are really supporting the American economy right now and it looks like they're in the shape of a very bubble,” Hannah told me.
The development was so disappointing that we put the spotlight on the extent to which the entire AI industry relies on hype.
Speaking just before the GPT-5 was announced, Altman compares it to its predecessor, the GPT-4o. “With the GPT-5 it's like talking to an expert. You're a legitimate PHD level expert in the area you need. Whatever your goals are.”
Well, not that much. When one user asked to create a map of the US labelled with all states, GPT-5 pushed out fantasyland, including states such as Tunnessee, Mississipop and West Wigina. Another urged a model for the list of the first 12 presidents, posting names and photos. Only nine people came up with it, including Presidents Giage Washington, John Quincy Adama and President Thomason Jefferson.
Experienced users of the new version of the predecessor model were especially appalled by the decision to shut down access to older versions and force users to rely on the new version. “GPT5 is scary,” a user wrote on Reddit. “There's a short replies with less “personality” that are inadequate and more offensive AI stylized stories. And there's no choice to use only other models. (Openai quickly became tolerant and resumed access to older versions.)
The high-tech media was not impressed either. Reviewing the “little guy” and the Futurist on the website, Earth Technica called the rollout “a big mess.” I asked Openai to comment on the disastrous public response to GPT-5, but received no response.
This doesn't mean that the hype machine that supports most public expectations of AI has taken a breather. Rather, it remains overdrive.
The forecast for AI development over the next few years, released by what is called the AI Futures Project under the title “AI 2027,” states, “predicting that the impact of superhuman AI over the next decade will be enormous and will surpass the Industrial Revolution.”
The rest of the document that maps courses in the second half of 2027 with AI agents “finally understanding their own cognition” thought it wouldn't be a parody of overly AI hype. I asked the creator if it was, but I haven't received a reply.
One of the issues highlighted by the overwhelming rollout of the GPT-5 is the explosion of one of the most important principles in the AI world. In other words, “Scale-Up” – the fact that artificial general information, or AGI, is approaching reality by awarding technology with more computing power and more data.
This is the principle that supports the huge spending of the AI industry on data centers and high-performance chips. Morgan Stanley estimates that by 2028, more data and more data systolic functions will require approximately $3 trillion in capital. This will outperform the global credit and derivative securities market capabilities. But if AI doesn't scale up, most, if not all of it, will be wasted.
As Bender and Hanna point out in the book, AI promoters continue to attract investors and followers by relying on a vague, public understanding of the term “intellect.” AI bots look intelligent because they have achieved what appears to be consistent in language use. But that is different from cognition.
“That's why we imagine the mind behind the words,” says Hannah.
Certainly, until the 1960s, the phenomenon was noticed by Joseph Weisenbaum, the designer of chatbot Eliza, who was the pioneering designer. Eliza recreated the response of the psychotherapist, convincingly recreating that even subjects talking to the machine knew the subject they believed to have shown emotion and empathy.
“What I didn't realize is that very short exposure to relatively simple computer programs can induce powerful delusional thinking in very ordinary people. Weizenbaum warned that “reckless personification of computers,” or treating it as a companion of some kind of thinking, would create a “simple view of intelligence.”
That trend is being exploited by today's AI promoters. They are labelled “hortography” for the frequent mistakes and manufacturing created by AI bots. This suggests that the bot has a perception that could have been slightly failed. But the bot “doesn't recognize it,” Bender and Hanna write.
The general public may ultimately be making cotton cotton into AI's failed promise. Predictions that AI will lead to massive unemployment in the creative and STEM fields (science, technology, engineering, mathematics) may inspire feelings that the entire company is a tech industry scam from the start.
The prediction that AI will generate a burst of increased productivity for workers is unmet. Many areas experience lower productivity. This is because mistakes and manufacturing do not try to find a way to mission-critical applications, as workers need to deploy and double check AI output.
Some economists dash the cold water more generally to forecast economic benefits. For example, MIT economist Daron Acemoglu predicted last year that AI rose by only about 0.5% in US productivity, and gross domestic product would increase by about 1% over the next decade.
The value of the Vendor and Hannah book, and the lesson of GPT-5, is to remind you that “artificial intelligence” is not a scientific term nor an engineering term. That's a marketing term. And that applies to all the chatter about taking over the world in the end.
“Assistance about consciousness and senses is a tactic to sell you with AI,” writes Bender and Hannah. Similarly, there are stories of billions, or trillions, that take place in AI. Like other technologies, profits go to small executives, but the rest pay the price…unless you're more clearly aware of what AI is and, more importantly, not.
