I’ve known and worked with Rizwan (Riz) Virk for many years now as a tech entrepreneur (including founding Play Labs @ MIT), a researcher (I was on his PhD committee), an academic and educator (he’s a Research Associate with ASU’s Future of Being Human initiative), an author — most recently of the highly successful The Simulation Hypothesis, and as a public commentator on tech and the future (including appearing a couple of years ago on the Joe Rogan Experience). He also writes for the substack The Simulation with Riz Virk. Riz is currently working on a series of essays on AI myths, and he graciously allowed me to post this one on myths surrounding AI and jobs here. Hope you enjoy it — Andrew Maynard.
AI Myth: AI Will Take All Of Our Jobs
A persistent myth during the current AI wave has been that “AI will take all of our jobs”, resulting in mass unemployment. The reason I’ve classified this as a myth is that it’s a perfect example of the premise in this series of essays I’m working on: that predominant narratives about AI aren’t entirely true. Rather, like myths, they have kernels of truth in them. AI will in fact result in the trimming of some jobs. The problem with this myth is that it ignores the reality of how technology improves productivity and the allocation of work. This narrative, like many others about AI (see “AI Will Kill Us All”), is already starting to look too simplistic as AI changes existing jobs and creates new kinds of work.
The Rise of the Myth
In July of 2025, during a trip to Washington, the CEO of OpenAI, Sam Altman said that “entire categories of jobs will disappear”. In the 2025 statement, Altman also added how AI was already better at being a customer service representative and got better results than most doctors in certain fields of diagnosis.
This kind of statement, coming from the head of the company that is most associated with the current wave of AI, wasn’t atypical for this era, and was only one of many which fueled this myth in the popular imagination.
A few months earlier, in May of 2025, the CEO of Anthropic, Dario Amodei, said that “AI was on course to potentially destroy 50 percent of entry-level white-collar jobs within the next few years,” according to Martin Ford in the revised edition of his book, Rise of the Robots. Amodei also predicted unemployment in the 10%-20% range. To put this into perspective, the only time in the past century with more than 20% unemployment was during the Great Depression (unemployment went from 3% in early 1929 to ~25% in 1933).
But it wasn’t just the AI company CEOs that got into the act. A month earlier, in April, 2025, the CEO of Ford Motor Company, Jim Farley said something similar: “Artificial intelligence is going to replace literally half of all white-collar workers in the US.”1
Surprisingly, this concern about AI was present even before the current crop of generative AI products hit the market. Back in 2019, when Altman was still head of Y Combinator, and just before he took the helm at OpenAI, Altman predicted “AI will probably replace most of the jobs people do today. ”
This sentiment wasn't uncommon in the past decade. Books exploring automation and the future of work appeared in the mid-2010s, including the first edition of Rise of the Robots (Ford, 2015), The Second Machine Age: work, progress, and prosperity in a time of brilliant technologies (Brynjolfsson & McAfee, 2016), and Machines of Loving Grace: the quest for common ground between humans and robots (John Markoff, 2015). These concerns echoed earlier times in American history when workers worried that automation (first machines, and then later computers) would take all of their jobs, resulting in, you guessed it, mass unemployment.
Which Jobs Will Go? Which Will Stay?
Before we explore the history, it’s worth saying that, like all of the AI Myths in this series, the kernel of truth in it is that AI will be able do many jobs better than humans and is already better than at us at doing certain types of office work. The underlying idea is that our economy has become an economy of desk jobs, and AI, especially with the release of new models every few months, as we’ve seen with recent releases of Open AI’s GPT Sol 6.1, Anthropic’s Claude Sonnet 5.5 and Google Gemini 4 Argon (to name just a few), can do almost anything we can do on a computer.
Adding to this general narrative are the predictions of AI doomers, like my friend Roman Yampolsky, who went extremely viral on the Diary of a CEO podcast for saying that only 5 jobs will be left after the job apocalypse, which would occur in the next 24 months (this was in September 2025):2
Emotional work: Therapy, counseling, and human care.
Creative work: Art, music, and writing driven by human expression or niche appeal.
Handcrafted work: Luxury physical items or craft goods specifically made by humans.
Oversight and regulation: Human auditors, ethicists, and policymakers attempting to monitor or manage AI systems.
Human companionship & engagement: Teaching, entertainment, or interactions rooted directly in human-to-human preference.
Of course, since Roman made that prediction, it’s clear that AI has become even better at creative work, emotional work, and human companionship than many humans, so that would leave not meany jobs left at all!
But to truly consider whether these are the only jobs that humans will be able to do, let’s look more closely at predictions about specific jobs — for simplicity, I’ve organized them into Blue Collar, Desk Jobs, and Human Interaction jobs. Some of these, the narrative claims, will be replaced by robots, some of these by AI programs like Claude and ChatGPT, and some by some combination:
Blue Collar
All blue collar jobs in factories, etc.
Street cleaners
Retail workers
All Truck Drivers and Uber drivers
· Desk jobs
Data entry and analysis
Software developers
Mathematicians
Other scientists
· Human interaction Jobs
o Doctors
o Therapists
o Surgeons
o Teachers (K-12 as well as secondary education)
o Home health care professionals
This is only a fraction of the jobs that have been predicted to disappear, and I won’t be able to even address all of these in this short piece. But I would like to hone in on a few that have been played up on in the press and see what they can reveal to us about this particular myth.
One that is currently in the news is that we will no longer need mathematicians. Earlier this year (in 2026), Open AI, using millions of dollars of compute, and thousands of agents, solved one of the “Millennium” problems (a fluid dynamics problem associated with the Navier-Stokes equations). These were problems which had been labelled as particularly difficult. After that groundbreaking discovery, on October 7, OpenAI announced that its AI had solved another 370 open math problems. In the weeks since, many mathematicians have been in “despair” according to some reports, and others reacted with “fury”. The one thing everyone seems to agree on is that “Math will never be the same”, though it’s too early to say exactly how academic mathematicians lives will change.
The Curious Case of Software Development
To see why I have classified “AI will take all of our jobs” as a myth, let’s start with my old profession: software developer. When I graduated from college at MIT in the early 1990s (my degree title was “Computer Science and Engineering”), computer science in general and programming in particular was the future. Back then, we wrote code by hand in low level (assembly), or mid level (C) languages which were compiled (for faster code) or interpreted (like BASIC or, later, Java and Python). Moore’s law was in effect, with computer processing power doubling every two years, so it became possible to use higher and higher level abstractions and programming languages to get output that could only have been achieved by programming in C when I was a student.
Throughout the 1990s, 2000s and 2010s, as computers made their way into more and more aspects of our lives, there was an almost insatiable demand for computer programming, and this gave rise not only to new jobs in the US, but began to lift many elements of the third world out of poverty by creating centers of offshore data processing and computer programming, most notably in India.
Software Development followed a pattern that has been evident in jobs during our modern technological era — the introduction of new technology into every part of our lives coincided with a rise of the profession that builds and maintains this new technology. This was true with radio, automobiles, television and other appliances, and of course with computers. Like these other industries, we should expect that software engineering would also go through a cycle, where it moves from being a specialty skill to a commodity. Today, while appliance workers and car mechanics haven’t disappeared, they can’t be called the same kind of high-growth industry that they were in the in the early and mid-twentieth century, and have certainly become more commoditized.
However, software was unique in that computers weren’t so much purpose built devices (such as televisions or automobiles) as much as technical platforms that could be used to build applications for any industry — and as computers went from large mainframes to mid-range computers, and eventually desktop computers and even mobile phones (a smart phone is basically a mobile programmable computer), computer software became important in almost every single industry, creating a large class of “software people” everywhere, from banking/finance to manufacturing to healthcare to the airlines and universities and well. The propagation of these computing devices among consumers created industries of apps, and now there are literally millions of apps and an industry that didn't exist before the Apple app store went live in 2008.
I’ve said before that the unexpected ability of LLMs to write code was one of the major inflection points for today’s AI to be more than a sophisticated Google search or summarization engine. Today, AI is very good at writing code at scale, with prompts. You can see many examples of this, including entire games that have been built by prompting AI — in the past, building games would take a whole team of people. Having been in the video game industry myself, I can attest to how revolutionary the ability of AI to write software is.
In early 2024, I decided to have ChatGPT write a small game (in JavaScript so I could run it in a web browser) for me that was modeled after Pac-Man, a game I enjoyed in my youth. It did a decent job in getting in something up and running, but the amount of times I had to keep prompting it to fix things made me so frustrated that I figured out some of the math myself and just edited the JavaScript file.
By the second half of 2026, AI was being used to create games like this spider-man game (Spider-Bench) running on three.js (a web-based game platform) all by one all by one developer/prompter/game designer:
What happened to the Jobs Apocalypse for software and other areas?
But even as AI has become better at writing code, something odd began to happen in 2026. Or rather, something became clear. Both the technological capabilities of AI began to be more clear, and the implications started to be more measurable in businesses and society.
In May, 2026, Sam Altman was forced to admit that, while he had been right about the “technological predictions” made, he had been wrong about the social and economic factors, which included the impending jobs apocalypse:
“I’m delighted to be wrong about this, I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened … “
“I don’t think we’re going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about.”
The “jobs apocalypse” phrasing began to make its rounds as others weighted in realizing that things weren’t progressing “as predicted”. This gets to a larger point in this series, that the ramifications and social & environmental side effects of technology, while very real, are also usually unpredictable. They end up being different from what the “experts” predict.
Just a month before Altman’s quote, CNN had reported that software development jobs, which were thought to be in danger, had simply changed in their nature. The headline was equally bubble-popping: The demise of software engineering jobs has been greatly exaggerated.
In another article, the Director of the University of Washington Paul G. Allen School of Computer Science, Magdalena Balazinska, replied to concerns from undergrad students about their future job prospects. The main message she wanted convey was: “AI is not killing your job options. It’s expanding them.”
It seemed that computer science itself was changing in important ways, but the number of jobs available to those who understand computer science and have adapted to the reality of AI was important: they need to use AI, maintain AI, improve AI and most importantly, oversee and direct AI tasks. Michigan Technological University made this point in an article for undergraduates who also worried about whether computer science was still a valuable field to study: “Entry-level jobs are shifting, not vanishing. Generative AI can automate some programming tasks, but human problem-solving and oversight remain critical.”
TechCrunch published an article a month after Altman’s mea culpa, in June of 2026, saying: AI was supposed to kill engineering jobs, but new data suggests they’re the most resilient. TechCrunch stated how, while many tech companies quoted “AI” as a reason for layoffs, the tech companies were hiring more engineers than before:
“In fact, engineers comprised 55% of all new hires in 2025 across the 12 companies SignalFire classifies as ‘Tech Majors’ — Alphabet, Meta, Apple, Amazon, Microsoft, Netflix, Nvidia, Tesla, Uber, Airbnb, Block, and Stripe. This is a significant jump from 2019, when engineers represented only 46% of new recruits, according to the report.”
A few months later, in September of 2026, the Economist came out with a bombshell article that showed why this is a myth, with a title that perhaps hit at the heart of the matter: The jobs apocalypse is postponed. An AI jobs boom is here.
And In October, 2026, as I write this, the New Yorker came out with an article with a similar title and claim: Will A.I. Still Take Our Jobs? The jobs apocalypse hasn’t arrived. Something else is happening instead.
What’s really going on here, and how is this likely to proceed into the future? A quick examination of how the economy has evolved during past technological revolutions may be in order before we try to answer this question.
Historical Fears and Hopes: The Case of Automation
The worry about the devastation that machines would wreak on the job market, first industrial machines, then computers, and now AI, has been going on for some time. In fact, the term robot, coined in Capek’s 1921 play, R.U.R. (Rossum’s Universal Robots), is derived from the Czech word for “slave” and directly translates as “forced worker”, as in serf. The plot centered around the robots rebelling. The plot of Fritz Lang’s 1927 film masterpiece, Metropolis, also touched on this issue, as the humanoid robot Maria is created to ease a possible rebellion of the working class.
Fears about being replaced by machines go back to a folk song from the age of industrialization, about John Henry being beat by a steam engine. While the 1920s’ worries were about automation via machines, eventually this would become fears of automation via computers, taking away, you guessed it, everyone’s jobs.
In the 1950s and 1960s there were congressional hearings and blue-ribbon panels that were worried about the loss of human jobs. This came in both utopian and dystopian flavors, just like today!
Keynes, taking an optimistic approach, wrote in 1930 that humans could start working 3 hour days and 15 hour workweeks because of what he called technical unemployment.
On the flip side, MIT Professor Norbert Wiener, wrote in 1950:
“Let us remember that the automatic machine... is the precise economic equivalent of slave labor. Any labor which competes with slave labor must accept the economic conditions of slave labor. It is perfectly clear that this will produce an unemployment situation, in comparison with which the present recession and even the depression of the thirties will seem a pleasant joke.”
Yet, the depression level jobs apocalypse that Wiener and others worried about from automation never came through. Before we explore why not, one objection that’s raised is that Weiner and others were talking about physical machines and automation, not AI as we know it, which is software. Therefore, things might be different this time, right?
In an essay titled, “Relax, Machines Already Took Our Jobs”, Marc Weber pointed out in 2021, there is no fine line between physical automatons and software automaton.
In fact, the first AI game playing program was a physical computer that could play chess, created by MIT professor and father of information theory, Claude Shannon, in the 1950s. You might call it a physical automaton, but Shannon himself envisioned AI that could not only implement the rules of a game like chess, but which could infer the rules without being told. Not only did computerized chess move from hardware to software, the feats that Shannon was describing were eventually accomplished by Google’s DeepMind group as they created AI players of the game Go.
Similarly, Alan Turing’s famous “Turing Test”, which he wrote about in 1950 calling it the “imitation game”, was about a physical computer and not specifically about AI as software. Today, we think of the Turing Test as involving software-based AI, recognizing that the important part of AI is the algorithms and the data and the code, running on general-purpose hardware. There have been many variations of the Turing Test over the years; I even posited in my book, The Simulation Hypothesis, a “Metaverse” or “Virtual Turing Test”, but that’s a topic for another article.
What’s Really Going on With Jobs? A Deeper Look at Software Development
The key point happening with software and other white collar jobs echoes what happened with blue collar jobs from earlier periods of tech innovation. It involves substitution and the creation of new jobs that result from technological innovation. It also involves something that’s called the lump sum fallacy, which I’ll address in a bit.
Let’s zoom in on software development since it is a quintessential desk job, and one I know well. The key point is not that AI isn’t or won’t be taking away some tasks or altering entry level jobs in software, but that the nature of software development has now changed.
I should say: software development has now changed, again.
It’s not the first time this has happened, though it may be one of the most significant changes. When I was a student at MIT learning computer science, we didn’t have debuggers, we had to use “print statements” or other logging methods to try to debug what was happening inside a program. Similarly, we didn’t have open source software that we could easily incorporate to build new software quickly.
The year that I graduated from college with my comp sci degree, Steve Jobs, then CEO of Next Computer, spoke on campus and I went to his talk. He predicted the rise of Object-oriented programming and visual programming — where programmers would be more responsible for putting together existing “objects”. This sounded like something from science fiction but it would happen gradually with the rise of open source.
Nor did we have visual programming environments. Visual Basic was one of the first visual environments for programming in the early 1990s. Rather than reduce the need for software developers, I remember how, at one point, there were an estimated 1 million people programming Visual Basic. In a sense, it had expanded what people could do in a short period of time, and expanded the pool of people that could be “programmers” — you no longer needed to learn about managing memory in C or C++, or even object oriented programming. Rather, you could code visually and fill in some high level, interpreted code, that was a heck of lot easier to build and program. This led to a whole expansion of visual programming environments for IT departments in the 1990s like PowerBuilder and Delphi.
All of these things, debuggers, reusable components, visual programming, and open source, have changed software development considerably. In fact, when I was coding in the 2010s, I rarely needed to write code from scratch; I could easily find open source and examples of algorithmic code that I might need on sites like Stack Overflow, etc.
And yet, the availability of these didn’t reduce software engineering jobs. Instead, they changed the expectation of what the software developer was supposed to do and the production code. They also expanded the people who could program, opening it up to more people.
By the time I did my last production-level programming, almost 10 years ago, software engineering had already changed from my student days. Building software already involved patching together many existing modules and writing code to control and put them together on the cloud. While the “assembly” method of programming didn’t happen in the way nor in the timeframe that Steve Jobs predicted in the early 1990s, it did happen eventually.
While this reduced some of the need for entry level programmers, the real result was increased productivity of a programmer. With the enhancements to tools and existing modules, I could single handedly code something in 2016 that would have taken a team of programmers in the 1990s. Enhanced productivity and expansion of access (i.e. more people can easily write code) were hallmarks of the changes in software.
Now we see the same thing happening with AI. AI is able to pull its knowledge of existing open source code and its uncanny ability to do what programmers did manually in the old days (take English requirements, turn them into lists of actions to perform, called pseudocode, and then turn those into runnable code) has changed software development forever. It means more time has to be spent on oversight, on coordination, on figuring out how to make different components work together, rather than on writing code manually. Most of all it has, even more than major improvements like Visual Basic, expanded the number of people who are writing software.
Sure, the old types of programmers are going the way of the past, since AI can write way more code, much more quickly than we could. But the need for people who understand software, and understand how to use AI to get software built, has gone up. And the number of people who can suddenly build things has gone way up too.
In short, the reason that AI hasn’t led to a software jobs apocalypse is because of productivity and expansion of access, which will lead to much more software being created than ever before. There’s also, as the Economist article about the jobs apocalypse pointed out, creation of new jobs that weren’t there before. Thus, we might expect an AI job boom for software as much as we expect an AI jobs apocalypse.
You can probably perform this same analysis with the other categories of jobs that were supposed to be gone, including oncologists or other medical professionals etc., that were supposed to be obsolete by now. You might have to wait several years after the doom predictions, but the same logic applies.
The Nature of the Job and the Lump Sum Fallacy
This isn’t the first time that technology has fundamentally changed the nature of work. There’s the famous example of the “computers” at NASA — who manually performed calculations for the moon mission. Today, computers and machines perform these calculations.
Weber also pointed out past jobs that had already been automated but which didn’t result in an apocalypse, including: meter readers, telephone operators, the secretarial pool, etc.
The need for punch card operators or elevator operators come to mind, as do typists or word processors. Even the introduction of desktop computers changed what an office worker was supposed to do — and they were expected to produce more because of desktop software than they might have been able to do with manila file folders and paper.
This is a natural part of the progression of technology, and every new tech wave creates new types of jobs and leads to changes in existing job descriptions, in addition to making some older, more rote jobs unnecessary.
There’s a key term that’s been applied to this — that predictions of job doom usually rely on what’s called the “lump of labor” fallacy. This “lump of labor” or “lump of jobs” fallacy assumes that there is a fixed amount of work — a lump “zero sum” if you will, and that all that we can do is to divide up this “fixed pie”.
As explained by the Federal Reserve of St. Louis, this fallacy is explained as:
“Machines, robots, and artificial intelligence are completing tasks that had been performed by human workers. The lump of labor fallacy would say that automation displaces human workers and results in fewer jobs.”
This is explained because, as one sector is automated, jobs tend to shift into other sectors, or to other responsibilities within the same sector. One often used example is agriculture:
“In the year 1900, 41 percent of the U.S. workforce worked in agriculture. After a century of technological change in that industry, the number stood at 2 percent (in the year 2000).”
And yet, if we look at total employment from 1900 to 2000, it went up significantly along with the growth in population during that time.
Where did the new jobs come from? Partly in the automation sector, partly in the services sector, and many jobs from new innovations that opened up new jobs which didn’t exist before. For example, when software development became a thing, so did QA development, and software documentation, and project management for software. We can see how each innovation creates new needs within an industry, and often creates new industries altogether. Prompt engineering wasn’t even a thing four years ago, nor was managing AI agents. Social Media manager wasn’t a job 20 years ago.
This “lump of labor” fallacy has been borne out again and again, and explains why Norbert Weiner’s predictions of unemployment due to automation in the 1950s that were close to Great Depression levels didn’t work out.
Is AI Different? AI optimists vs. AI doomers
But, you might ask, isn’t AI different? In some ways, this is asking the same question that’s been asked in the past, which is, this time, it’s different, isn’t it?
Today’s AI doomers are following a similar path. Even as headlines were starting to emerge that the AI jobs apocalypse might be postponed, AI doomers abounded, and they weren’t just in the big tech industry. Verizon CEO Dan Schulman, said, in April of 2026, that it’s likely that “Unemployment will climb to 20 to 30 percent in the next two to five years, he claims, and many manual laborers will eventually be usurped by humanoid robots.”
AI doomers like Roman Yampolskiy point out that AI will be able to do every single desk job better than humans, and Robots will eventually be able to do every single job better than humans, therefore almost all jobs will be gone.
This is an extension of the argument made with machines and automation. In the 1920s, a time of great industrialization and automation (out of which the aforementioned science fiction like R.U.R. and Metropolis emerge), this concern was there. Half a century later, Weiner and others worried about it with the emergence of computers more broadly.
It’s important to note that, when it comes to jobs and AI, there are AI optimists as well as AI doomers. Peter Diamandis, founder of the X Prize, envisions a future where AI may take some entry level jobs, but that will leave room for more meaningful jobs, higher level jobs that require human judgement and oversight of AI.3
In Law & Liberty, Dr. David Hebert explains why this fallacy comes up again and again, and what it misses:
“This is why the predicted catastrophes in the labor market never arrive. The hand-loom weavers displaced by the spinning jenny were a genuine disruption. But the textile industry that followed employed more workers, not fewer, producing vastly more output and at lower prices. In doing so, this generated the demand for goods and services that had previously not existed.”
Is this playing out with AI?
We can already see through the software market that jobs are changing and that AI agents are creating new opportunities. There are also many new companies today, which will become the Googles and Facebooks and Apples of tomorrow, that are being formed. The amount of venture capital flowing into new AI ventures dwarfs even previous boom cycles of tech investment, including the dot com boom. This is both because AI is seen as such a transformative technology, but also because this will result in new types of jobs.
To use perhaps an old science fiction example, even in Isaac Asimov’s Robot stories, the replacement of human workers with robots occurred first in dangerous situations, and then created new types of jobs that didn’t exist before, including robot psychiatrists. Do we need AI psychiatrist? This I’ll address in a different essay, but I just wanted to point out that new technology always creates a need for new types of jobs that didn’t exist before.
The Real Problem We Face: the Kernel of Truth in the Myth
The real issue, then, is not that AI will take all of our jobs, because as the lump of labor fallacy shows us, new technology will create new jobs.
However, we cannot deny that previous waves of automation have results in reduced employment in existing sectors — the farming sector being a perfect example.
The real issue, the kernal of truth in the myth, that we do need to address, is will AI replace jobs so quickly that there won’t be time for the economy to adjust? This is a real short-term danger that could happen if companies proceed with automation too quickly.
Dr. Hebert describes this in his article:
“A charitable interpretation of the fear this time is not that AI will replace jobs permanently, but that the disruption will be so large and so pervasive that the rate of job loss could far outpace the abilities of entrepreneurs to imagine and create new sorts of work for these displaced workers.”
Let’s turn back to the software industry, the industry I know well. There were similar concerns with the hiring of offshore developers (in places like India, China, Pakistan, Vietnam, etc.) in the 2000s. Yet, the need for supervisors who understood software development here in the US increased, and many companies found that initially moving jobs offshore, while it saved costs, resulted up in productivity issues. AI can only do what’s it being directed to do and will often, like offshore programmers, get things wrong. Humans need to be in the loop.
We are seeing similar problems with AI today. AI has promised a lot, but managing AI to get that productivity is not as easy as it seems. Just recently, I met with a startup that was creating software to try to manage AI agents that were writing code. This is a much more challenging problem than it may seem at first — but it is an extension of the simple example I gave earlier of trying to get ChatGPT to write a Pac-Man type game. I had to spend way more time directing the AI, finding issues, than writing code. Just as prompt engineering didn’t exist a few years ago, AI agent manager didn’t either.
Boston Consulting Group put out a report this year (in April 2026) that AI will “reshape” jobs more than “replace” jobs in the near term. The report points out how, while AI may be replacing some call center employees, it is also augmenting call centers and creating new jobs. The report also points out, just as the others have done, that software jobs have increased significantly from Jan 2020 to the end of 2025, which includes the first 3 years of the AI boom.
Conclusion
We can see the automation doomers narratives in the past didn’t quite play out as expected.
Instead, we had rising productivity in addition to loss of jobs in specific automatable sectors. In an example that Silicon Valley AI enthusiasts like to often use, lamplighters or elevator operators no longer exist as jobs, but we have more jobs total in the economy than we ever have.
Similarly, as mainframe computers were rolled out in the 1950s and 1960s to automate finance and other “counting” and “database” operations, this didn’t lead to an economic downfall, but created a whole new field of “computer operators”. This became doubly so when personal computers, starting with the Apple II and IBM PCs made themselves a common sight in business small and large.
Overall, those who learned to operate the new technology found a new world of economic opportunity, and it created many new opportunities for those that couldn’t have operated the older computers. As the internet came about in the 1990s, demand for software developers and implementors would outpace almost all other sectors of the economy. Coincidentally, software development was thought to be among the first jobs on the chopping block, and yet, the number of software positions is going up.
AI will definitely reshape jobs, and will take some entry level jobs. But the lesson, just as with previous technological advances, is to adapt to the new jobs that will use AI rather than simply being replaced with AI.
As Dr. Hebert points out, the lump of labor fallacy is not just about people assuming that there is a fixed amount of work, it is also about the fallacy that humans wants are fixed. As technology increases, so do the markets and desires for new products, new technologies and new services, and in the end this, along with increased productivity, is what drives economic growth.
Those who believe we heading into a job apocalypse suffer from the pessimist’s lack of imagination that characterized earlier eras. The same will be true of AI. In short, an AI jobs boom will likely outweigh an AI jobs apocalypse, avoiding the global depression that AI doomers are worried about.
Rizwan Virk is founder of Play Labs @ MIT, author of The Simulation Hypothesis, and is a member of ASU’s Center for Science and the Imagination and Future of Being Human Initiative. This article one of a series Riz is working on focused on AI Myths. Read more at The Simulation with Rizwan Virk
Ford, 2026
Note: this particularly summary is, in fact, AI generated, by Google gemini, perhaps adding credence to Roman’s claims.
See, for instance, The Death of the Company.
