In recent discussions about the role of artificial intelligence in the economy, the predictions from tech giants tend to provoke a wide spectrum of reactions. Servaas Storm, a Dutch economist, critically evaluates some of these claims, offering insights that challenge the prevailing narrative about AI’s potential impact. His perspective is particularly critical of overly optimistic projections, suggesting that they might be more hype than reality.
By Servaas Storm, a Dutch economist and author focused on macroeconomics, technological advancement, income distribution, economic growth, finance, development and structural change, as well as climate change. Originally published on the Institute for New Economic Thinking website
Leading generative AI companies are investing staggering amounts of money into the development and operation of their algorithms, ostensibly aiming for what they describe as ‘superintelligence.’ To maintain this excitement and keep investors intrigued, CEOs are inundating the market with a flood of untestable prophecies that range from claims about widespread job losses to dystopian fears of human extinction and, conversely, promises of a world brimming with wealth, scientific breakthroughs, and productivity levels that soar thanks to automation.
In a recent INET Working Paper, I attempt to thoroughly analyze some of the more prominent yet ludicrous predictions generated by Silicon Valley’s leadership. In order to maintain perspective amidst these wildly unrealistic forecasts, I draw upon Bertrand Russell’s (1952) Tea Pot Analogy. He argued that those who make unfounded claims should bear the burden of proof, rather than others being tasked with disproving them: “If I were to suggest that between Earth and Mars there exists a china teapot revolving around the sun in an elliptical orbit, no one could conclusively disprove my statement if I were careful to specify that the teapot is too small to be observed by our most powerful telescopes. Yet if I were to assert that because my statement cannot be disproved, human reason must accept it, I would rightfully be labeled as nonsensical.”
This blog post aims to summarize eight highly debated assertions regarding the economic impacts of AI, enticing readers to delve into the more extensive analysis found in the paper. A note of clarification: my focus here is exclusively on generative AI as exemplified by commercial closed-source Large Language Models (LLMs) and does not extend to smaller domain-specific AI tools or other machine-learning methodologies.
Claim #1: AI ‘superintelligence’ is just around the corner
Across the industry, tech CEOs propagate remarkably similar narratives, each speaking to a supposed near-future where ‘superintelligence’ heralds a transformative time of prosperity, curing humanity’s greatest ills. Elon Musk has suggested that Artificial General Intelligence (AGI) could be achieved as early as December 2026, with various estimates from industry leaders placing AGI’s arrival anywhere from 2026 to 2030. Regardless of the accuracy of these forecasts, the concept of AGI remains perpetually ‘just around the corner.’
Teapot Alert #1: AI ‘superintelligence’ is neither present nor imminent
Claims regarding the immediate emergence of AGI resonate with a strong sense of déjà vu. Many have lost count of the times Silicon Valley has predicted the imminent arrival of autonomous vehicles, quantum computing, and humanoid robots, often finding that these timelines serve more as fundraising tools rather than reflections of technical accuracy. It is important to note that the assertion of imminent AGI is fundamentally flawed, not merely due to misguided timelines, but because the foundational approach to generative AI cannot, by its nature, breed ‘intelligence’ or AGI.
Current iterations of large language models (LLMs) lack true intelligence, causal reasoning capabilities, and any semblance of a ‘conceptual model of the world.’ They are, more accurately, advanced predictive pattern matchers and retroactive text generators. Prolific models like GPT-5 and others exhibit hallucination rates between 3-8% on controlled factual queries but are prone to much higher errors (up to 50%) during complex reasoning tasks carried out on unfamiliar datasets. The ‘hallucination’ phenomenon endemic to LLMs cannot be ameliorated simply by scaling; it is inherent to the models themselves.
Leading cognitive scientists have long argued against the notion of scaling LLMs as a pathway to superintelligence. Many industry insiders, including figures like Yann LeCun, have now recognized that advancing LLMs’ capabilities through sheer size and speed is not a viable strategy. While LLMs excel at recognizing and generating patterns, they inherently struggle with causal understanding, lack a functional ‘model of the world,’ and represent a bottleneck in the quest for intelligence.
Claim #2: The AI boom is not a bubble
Within the tech industry, there is a consensus that massive investments in AI infrastructure are laying the groundwork for a future filled with economic abundance and transformation. The belief is that these infrastructures will support machine-learning algorithms that will revolutionize business models, enhance productivity, and spur growth. Investors are reassured that the foundations of this AI industry are robust, asserting that potential issues related to financial maneuvers among tech corporations won’t lead to collapse.
Teapot Alert #2: This AI boom is, in fact, a bubble
Numerous unexpected challenges could arise, especially if LLMs fail to deliver on their overly manufactured promise. Compounding this are the looming economic challenges—soaring energy prices and a tightening monetary policy—characterizing the current U.S. economy. Concerns surrounding the health of private credit markets, which heavily support AI and tech sectors, reveal an ominous backdrop. Lloyd Blankfein, the former CEO of Goldman Sachs, suggests a storm may be brewing, noting that we are approaching a reckoning concerning the tech industry’s current trajectory.
A primary reason for skepticism regarding the sustainability of the AI investment cycle is the lack of a clear path to profitability for any AI firm. Companies in the space are grappled by escalating operational costs and are under pressure to provide free or low-cost services to grow their user bases. Unfortunately, this model is proving generative of losses due to overwhelming competition, especially from lower-cost providers. The situation is forcing many companies to impose restrictions on LLM access, limiting usage during peak times.
Claim #3: AI will lead to a white-collar jobs apocalypse
According to some forecasts, AI is expected to decimate white-collar positions rapidly, rendering vast swathes of an educated workforce obsolete. Dario Amodei, along with other tech leaders, has articulated fears of a swift and devastating impact on office jobs, suggesting significant portions of the U.S. work force may face unemployment in a so-called ‘great disemboweling’ caused by AI.
Teapot Alert #3: AI will reshape jobs, not eliminate most of them
The underlying purpose of these alarming predictions appears to be to instill fear of AI in workers while emphasizing the supposed transformative power of LLMs to investors. History does not support these gloomy prophecies; a previous study indicated that only a fraction of jobs faced significant automation risk. Current data supports that most U.S. occupations remain largely unaffected by emerging automation technologies. Indeed, more than 40% of U.S. positions have not shown relevant AI usage data, indicating low exposure to AI innovations. Even in roles counted among high AI-exposure occupations, data suggest a trend toward redefined responsibilities rather than outright replacement.
A telling example can be seen within the field of radiology, where positions have remained stable despite the advent of AI. Meanwhile, structural biologists are now dedicating less time to straightforward tasks such as determining protein structures, redirecting their focus toward analysis and validation, resulting in the creative growth of new job titles and responsibilities.
Claim #4: AI will boost labor productivity
This belief arises as a natural follow-up to fears of job loss; proponents suggest that increased usage of agentic AI tools will lift overall labor productivity significantly, setting the stage for massive economic advancement.
Teapot Alert #4: Net productivity gains from AI may be minimal
Studies show that while some occupations will experience pronounced growth due to new machine-learning tools, overall productivity growth will likely not see major increments. A recent survey of economists gives credence to this, estimating only modest productivity increases for the coming decade. The overall financial impact of AI investments during this period is likely to see disappointing returns, suggesting that the risks associated with these new technologies could outweigh their benefits, resulting in hidden costs and inefficiencies.
Moreover, many across the workforce report that managing AI-generated tasks often consumes equal or greater amounts of time than what is saved by automation, further fueling counter-productivity. Employees find themselves bogged down dealing with ‘work-slop’—residual errors and inaccuracies from AI outputs that require correction. Such situations create mistrust in AI tools and ultimately contribute to lower overall morale and productivity.
Figure 1 How Much Time Do You Think You Save Each Week by Using AI (Percentages) Source: Ellis (2026). Note: Totals may not add up to 100% because of rounding
Surveys reveal that AI tools are costing companies valuable time instead of saving it, with employees spending excessive hours correcting the ‘slop’ produced by AI systems. The overreliance on AI technologies has cascading negative effects on other areas too, like the legal sector, where courts are overwhelmed by improperly-generated documents. An examination of how AI is implemented in various sectors, such as healthcare, indicates a deterioration of essential skills and a decline in critical thinking.
Claim #5: AI is already destroying jobs in high AI-exposure sectors
Recent studies suggest that AI is having a measurable adverse effect on the job market, particularly for younger generations, with a reported net loss of around 16,000 jobs per month in the past year. Notably, entry-level jobs have declined sharply, causing concerns for younger workers trying to enter the labor force.
Teapot Alert #5: Blaming AI for low hiring rates misses the bigger picture
It’s premature to entirely attribute low hiring rates among recent graduates to AI. The broader macroeconomic environment also needs to be considered. Following extensive low interest rates during a period of rapid growth, companies began increasing hiring significantly. However, the current financial climate—marked by tighter monetary policy—has resulted in widespread job reevaluation, with many companies rolling back their previously expansive hiring levels.
Figure 2 Hiring Rates in Major Industries (Percent Change during November 2021-February 2026) Source: FRED database, based on BLS data from the Job Opening and Labor Turnover Survey (JOLTS). Notes: The hiring rate is defined as the number of newly hired workers as a percentage of already employed workers in a month. The red bars indicate industries with the highest AI-exposure.
Claim #6: AI will eliminate ‘bullshit jobs’ and create better opportunities
Some posit that AI’s emergence will phase out meaningless roles, particularly in administrative and managerial functions, liberating employees to engage in more fulfilling work. However, this characterization is shallow.
Teapot Alert #6: In reality, AI will likely create more ‘bullshit jobs’
With AI taking on numerous tasks, there will be a concurrent rise in supervisory jobs aimed at overseeing AI outputs and managing the errors that inevitably accompany them. These new roles—often considered ‘bullshit jobs’—will encompass everything from AI oversight to compliance and ethical review, ultimately inflating administrative layers rather than diminishing them. For instance, organizations may establish AI ethics boards and task groups dedicated to ensuring AI systems function responsibly, adding layers of oversight to maintain stability in the workforce.
Furthermore, the increased use of AI will likely lead to expanded surveillance mechanisms within the workplace. Companies could find themselves employing more personnel purely to monitor employee performance and maintain compliance with AI-generated directives, which ultimately fosters a culture of distrust and anxiety rather than innovation and motivation.
Claim #7: Anthropic’s Mythos is too dangerous to be released publicly
Anthropic’s AI model, Mythos, has been touted as so advanced that it can identify countless vulnerabilities in software systems. Initially, this has led the company to withhold its public release over concerns for cybersecurity.
Teapot Alert #7: The real issue lies in strengthening cybersecurity, not in the capabilities of Mythos
While Mythos highlights genuine flaws in current cybersecurity practices, Anthropic’s claims about its extraordinary capabilities are inflated. Security systems in the real world involve multiple defensive layers, and the findings are based on exploring vulnerabilities outside the actual environments in which they exist. Thus, rather than merely blaming advanced AI for security failures, efforts should focus on improving the standard of cybersecurity available.
Claim #8: We need to shift toward ‘pro-worker’ AI
Some economists, including those from MIT, advocate for policies that would steer AI development in a direction that enhances worker value, aiming to create new tasks and opportunities for employees.
Teapot Alert #8: Commercial AI is unlikely to become pro-worker
Calls for ‘pro-worker’ AI do not adequately address the ingrained tendencies of capitalism towards automation and control. Historical precedents reveal the dismantling of labor unions and other worker protections in the face of technological advancement. Moreover, existing practices among major corporations often focus on harnessing AI as a tool for exerting control rather than advancing worker rights.
The Takeaway: A Call for Realism About AI
Current perceptions regarding the economic implications of AI require reevaluation. The notion that AI will lead to transcendent intelligence, widespread job losses, unprecedented productivity gains, or radical technological progress may be overly optimistic. In reality, AI’s aggregate impact on the labor market will be more subdued, creating some redundancies while simultaneously reshaping jobs and introducing new roles to manage AI systems. However, looming costs and inefficiencies will likely overshadow any productivity gains, leading to an unsustainable trajectory for current investment trends.
