How concerned should we be about AI? Fears of widespread unemployment have always been raised by technological advancements, but are these worries justified?
As Josh Tyrangiel notes in The Atlantic, Dario Amodei, CEO of the AI firm Anthropic, stated in May 2025 that the technology his company is advancing could eliminate half of all entry-level white-collar jobs and increase unemployment to 10%20% over the next one to five years.
According to Ford CEO Jim Farley, AI will eliminate half of all white-collar jobs within ten years. According to OpenAI's Sam Altman, it is only a matter of time until a billion-dollar business is run by a single employee.
Predictions of catastrophic outcomes following the introduction of new technology are nothing new. However, the makers of the new technology and those who are most eager to adopt it are the ones who are prophesying doom rather than the usual suspects.
So, are they correct? According to Tyrangiel, AI is obviously already changing the nature of work. Recently, companies like Meta, Amazon, Walmart, and JPMorganChase announced layoffs because of "automation."
According to three researchers from the Stanford Digital Economy Lab, since late 2022, entry-level positions that could be disrupted by AI have already decreased by 13%. Therefore, even though it's too early to be certain (other factors could explain the decline and the evidence is sparse and mixed), the transformation may already be under way.
Economists assure us that if that change happens gradually and the economy adapts swiftly, we might be alright or even better off overall. However, the consequences could be severe if AI instead causes a quick reorganization of work, condensing years of change into months and impacting about 40% of jobs globally, as the IMF predicts.
Can we truly benefit from AI?
Which will it be? As technology analyst Benedict Evans has noted, humans have been automating tasks for 250 years. History demonstrates that every automation wave has eliminated entire job classes and produced new ones. Some people may find the process unpleasant, but overall, it has led to increased prosperity.
We are confident that this time is not likely to be different because of two economic concepts. The first is the "lump of labor fallacy," which is the false belief that there is a set amount of work that must be done and that people will have less work if some of it is done by a machine. However, if using a machine to make shoes, for example, becomes more affordable, more people will be able to purchase shoes, giving them more money to spend on other things. We will also discover new things we need or want, and new jobs will be created.
The second idea is the Jevons Paradox. People in the 19th century were concerned about what would happen when the coal used to power the Royal Navy ran out. The optimists reassured them that as steam engines become more efficient, they will use less and less coal. Not at all, according to economist William Stanley Jevons: if we make steam engines more efficient, they will be less expensive to operate, and we will use them for new and different purposes. As a result, we will need more coal.
According to Evans, this paradox has long existed in connection with white-collar employment. Clerks could produce more than ten times as much in the 1880s thanks to typewriters and carbon-copy paper as they could in the days when they had to manually copy documents one at a time. The outcome for clerical employment was a significant increase in the number of clerks hired. If one clerk can complete the work of ten, you might want to take on more of the tasks that clerks perform, like handling more inventory or conducting more analysis. Because of the new technology, you may be able to create a different and more effective business.
When digital spreadsheets were introduced much later, they could perform tasks that would have previously required a team of accountants to complete over the course of a week with just a single click. Accountants now have more jobs.
According to Noah Smith's report on Substack, the most recent study on the effects of AI on employment appears to confirm that this is indeed the case this time. Researchers at US tech start-ups Ramp and Revelio Labs, Ara Kharazian, Lisa Simon, and Ryan Stevens, looked at private data to find out what happens when businesses use generative AI. They hire more people, which is the solution. Entry-level employment increased as well. According to Smith, it appears that AI is "still mostly a complement to human labour rather than a substitute" for it. AI is "behaving pretty much like a normal technology" for the time being.
Software is all that AI is.
That's the typical pattern, and according to The Economist, it would be "unprecedented in human history" if AI really began to advance at the rates feared and with the expected consequences. Because the diffusion of new technologies always happens slowly, they have never spread quickly enough to cause widespread unemployment for extended periods of time.
As Tyrangiel notes, AI is just software, which helps explain why it is unlikely to be different this time. Additionally, "people hate it almost as much as they hate change" when it comes to software. AI must access data and be integrated into current systems before it can completely change a business."trade secret of the majority of Fortune-500 companies is that they continue to use heavy-duty, industrial-strength mainframe computers for essential operations, which virtually never malfunction and thus cannot be replaced." Big changes involving many people with strong opinions about the "right" course of action would result from integrating such legacy technology with AI. As time goes on, months and years go by, and "the CEO still can't understand why the miracle of AI isn't solving all of their problems."
According to Evans, it "sounds like classic tech solutionism, but turned from utopia to dystopia" that "one magic piece of software" will instantly transform everything and overcome the complexity of actual people, actual businesses, and the real economy. According to Zeynep Tufekci in The New York Times, reality appears to be quite different. Businesses that tried completely automating tasks like customer service have failed. As a result, con artists have been persuading chatbots to cede control of important features by offering refunds or amazing discounts, like a brand-new car for £1. The McDonald's order-taking bot turned out to be "wildly dysfunctional."
It's important to realize that these incidents are caused by the technology operating as intended, not by mistakes. Present-day AI systems are "not reasoning machines"; instead, they merely generate likely responses based on the data they have been trained on. They lack both intelligence and common sense. AI is capable of "doing many things with astounding efficiency," particularly if those tasks are structured, formal, and subject to real-time testing and verification. "Good old-fashioned human intelligence" is still necessary for the majority of jobs, which are just not like that.
According to The Economist, this does not imply that the "job apocalypse" won't occur. Perhaps things will be different this time. It's possible that technology will change in ways that are not yet foreseeable. If this is the case, you may recognize the end of the world by these indicators: in the US, the frontier economy of the world, sharply increasing productivity coupled with weak real-wage growth. This would manifest as a rise in GDP per capita above the upper limit of 2.5 percent, which has historically been the norm in frontier economies, and a concurrent increase in corporate profits, indicating that capital, not labor, was benefiting from increased output. Another indicator of a recession would be significant job losses across many industries. The jobs that disappear during the upcoming recession will "give a hint about the shape of the AI world to come."
Is there a real indication that any of this is taking place? Not really. According to The Economist, the job market "certainly is not cracking yet." "The percentage of working-age people in the OECD who are employed continues to break records, unemployment in the club of mostly wealthy nations is only 5%, and America employs more people than ever in AI-exposed industries like law." Prior to ChatGPT's introduction in late 2022, which ignited the AI revolution, American graduates have been having difficulty finding employment. There won't be much disruption, according to many economists. The Bureau of Labor Statistics in the United States predicts that between 2024 and 2034, the nation will add 5.2 million jobs, resulting in a 3% increase in total employment.
Your job cannot be done by a robot.
Skepticism has wider causes. Due to the fact that using AI is more expensive than beneficial, the biggest users have been rushing to reduce its use. Remarkably few people regularly use the technology, and the percentage of OECD businesses that have embraced AI is still quite low (roughly 20 percent), though numbers for both individual use and business uptake vary greatly between studies based on what is considered to be significant. As Evans has argued, the fundamental issue here is that most people simply don't know what AI is supposed to do for them. You can type text into a box and receive a text response. The text is frequently accurate in general but inaccurate in detail. How many lives will be changed by that? As Pablo Picasso wisely observed in 1968, "Computers are useless." All they can do is provide you with answers."
AI is likely to make some jobs easier and faster for some people rather than completely replace them. According to Evans, jobs are typically a complicated web of elements that we might not even be able to clearly describe. For instance, you may understand why a chatbot will never be able to perform your job, but you will be impressed if someone claims that it can already perform the duties of a doctor or a lawyer. The harsh reality is that we have no idea what exactly goes into jobs that we are certain will disappear tomorrow, let alone how, if at all, AI will alter them.
Fear itself is what we should be most afraid of. According to Robert Shiller in The New York Times, a recent survey revealed that 70% of Americans think AI will limit their job options. There could be economic repercussions from that fear alone. There is a chance that fear will "help birth the reality" when millions and millions of people base their economic decisions on pessimistic expectations. Instead of promoting alarmist narratives in the hopes that the media attention that follows will demonstrate the strength of their most recent AI model, Silicon Valley leaders should learn to do better. If the outcome is an economy immobilized by fear and recession, they will have a harder time selling their products in the future.
Leave a comment on: Will AI truly eliminate every job we have?