Despite significant technological advancements, many experts are raising doubts about whether artificial intelligence (AI) can fulfill all the needs of a modern global economy or deliver its promised vast economic benefits at an affordable societal cost.
Anthropic’s Findings Temper Job Apocalypse Fears
In March, Anthropic, the developer of the chatbot Claude, published an analysis to address concerns that intelligent robots would redefine human existence and eliminate the demand for human labor. This report offers a more measured perspective compared to earlier predictions made by Anthropic’s co-founder, Dario Amodei.
Previously, Mr. Amodei had asserted in May of last year that AI could eradicate half of all entry-level jobs within one to five years. In January, he stated that AI would likely become a “general labor substitute for humans.” Later, in June, he warned about the risk of “a world where the economic trade-off dial is stuck on the hypergrowth, hyper-inequality setting.”
However, the recent Anthropic analysis indicated that the technology’s actual impact thus far has not matched expectations. The report noted: “We find no systematic increase in unemployment for highly exposed workers since late 2022.” Furthermore, it pointed out that deploying AI technology “remains a fraction of what’s feasible,” citing that Claude handles only 33% of tasks within the computer and math category, while theoretically capable of managing nearly 100%.
Economic Reality vs. Technological Hype
The broader economic picture suggests caution regarding AI’s immediate disruptive power. Although spending on data centers is increasing rapidly, productivity gains have not reached the dramatic levels suggested by tech proponents. In fact, labor productivity during the initial years of the AI era was slower compared to the information technology boom that began in the mid-1990s.
Even prominent figures associated with the field are reassessing its job-killing potential. Sam Altman, a public face of OpenAI, stated in May that he no longer believes in the “kind of jobs apocalypse that some of the companies in our space advocate or talk about.” This sentiment is echoed by MIT economist David Autor, who observed that “A lot of people have noticed that the world is not changing as fast as they predicted.”
This shift has moved public discussion away from a catastrophic narrative toward one that emphasizes history’s complex relationship between automation and human employment. Consequently, the tech-heavy Nasdaq index, which had been heavily driven by AI stocks, experienced an approximate 8% decline since its peak in early June.
The O-Ring Argument: Value of Remaining Tasks
One key theoretical critique applied to the “AI will do everything” narrative is known as the O-ring argument. This concept draws an analogy from the failure of the space shuttle Challenger on January 28, 1986—a structural failure caused by a simple rubber O-ring that did not function correctly in cold temperatures.
The analogy suggests that since AI cannot perform every task flawlessly, it will consequently boost the value of all remaining tasks. This increase could benefit high-skilled workers if AI takes over the low-level components of their jobs, or conversely, it might elevate opportunities for lower-skilled workers by assuming more expert duties.
A recent academic study supported this view, noting that “despite strong substitution at the task level, overall employment effects are modest, as reduced demand in exposed occupations is offset by productivity-driven increases in labor demand at AI-adopting firms.”
Challenges Beyond Capability
While some experts remain optimistic about AI’s future—with Nobel laureate Daron Acemoglu noting that “Insiders are as gung ho as ever” regarding the advent of artificial general intelligence (AGI)—several practical and economic hurdles temper the enthusiasm.
Economist David Autor points out a critical limitation: “Not everything is a computational problem.” While AI excels at language replication, it lacks the ability to connect that language to real-world physical reality and remains prone to significant errors. Furthermore, the political acceptance of the technology is wavering; seven in ten Americans oppose constructing AI data centers locally, citing concerns over energy demand which drives up local electricity costs.
The economic sustainability also raises questions. The International Energy Agency projects that power usage from data centers will more than double by 2030, reaching roughly 945 terawatt-hours—a total exceeding Japan’s entire energy consumption. Moreover, the rapid depreciation cycle of AI models means companies developing these systems, according to Acemoglu, “are never going to make money… They are losing hundreds of billions of dollars every year.”
The consensus among emerging analyses suggests that while AI is advancing rapidly and its progress shows no signs of slowing down, the technology may ultimately fail to deliver vast economic benefits at a price point society deems acceptable.
