Scott Robinson

May 23, 202322 min read

The Coming Decimation of the Job Market

Updated: May 24, 2023

You probably see some, if not all, of these headlines every day at this point:

AI is going to cause a global financial meltdown

AI can do many jobs better than humans

Companies looking for ways to replace workers with bots

And, most concerning of all,

AI is coming for your job!

With something as big and new and groundbreaking as generative AI, it’s no surprise we find ourselves drowning in media hype. But it is big, and new, and groundbreaking, and so we really have no choice but to bite the bullet and sift through it all. Just how big a deal is generative AI, what can we realistically expect to change, and how do we prepare for that change?

Well, if you’ve seen generative AI for yourself, you’re aware that it’s a very big deal. You’ve thought the implications through, and can tell that it’s going to be a game-changer. We’re at a serious inflection point, not just in the evolution of technology, but for our fundamental economic models.

And, yes, it’s coming for your job.

But this has happened many times before. New technology comes along, new processes and methods replace old ones, and jobs go away. Then new jobs come along and replace them.

Disruption is nothing new.

The question isn’t whether AI is coming for your job; it absolutely is. The question is, is this time really any different?

Master and Mule

AI philosopher Calum Chace says yes, this time is absolutely different. And he rallies a wide phalanx of other experts around his argument in his book Artificial Intelligence and the Two Singularities.

Roger Bootle, a London economist and advisor to the British House of Commons, likewise agrees, and musters not only elegant arguments but compelling data to support his case.

The rising chorus of voices singing, Yes, this time is different! present a range of reasons why, but the biggest of these rattles cages past to present, overturning even beloved Adam Smith. And that argument can be presented as follows.

In the beginning, the work of human beings was performed by human beings. And that was that. If something needed doing, human beings did it. There was no other source of labor, and the labor was self-directed.

Then human beings discovered that certain animals could be put to work: horses, oxen, et al. This changed the equation considerably: now there was a new source of labor, but it was one that couldn’t self-direct; it needed guidance.

Revolution: Agriculture!

Fields of grain no longer had to be hand-plowed by humans; they could be plowed by mules, and the humans took up a new role – pointing the mules in the right direction.

Management was born!

After endless centuries of this dynamic – animals doing the heavy lifting, humans managing the work – a new kind of production emerged: the making of things in vast quantities. This was work that couldn’t be done by animals, because it was too complex and required dexterity – not just muscle power. But keeping human beings coordinated in these processes required that the new factory floors include not just the workers, but managers to keep them pointed in the right direction.

Revolution: Industry!

The Master-and-Mule dynamic prevailed, but now it played out with human mules overseen by human masters.

This spilled out of the factory and into the broader business universe, where the rapidly advancing societies of humanity began concocting ever-more complicated and abstract ways of making money and spending time. Finance. Insurance. Lawyers. In all of them, Master-and-Mule prevailed.

And as technology proliferated, new machines were created that could do the work in the factori es. It became cheaper to turn work over to a machine than to a human being.

Revolution: Automation!

By the hundreds of thousands, human beings were phased out and machines were phased in. Master-and-Mule prevailed – but now the Mule was a machine.

In all three of these revolutions, the same thing changed: the Mule. At first, the Mule was an actual mule (Agriculture); but then the Mule was a person (Industry), and then a machine (Automation). Still, the thing that changed was the same – the Mule.

This time is different.

As AI begins taking up cognitive tasks – routine decision-making, the generation of insights, engagement, and now even human interactions – the Master is beginning to change. The management of processes central to many businesses are now within the purview of AI.

Examples include back-end administration in banks, health insurance offices, and law firms, among others. And in almost every industry, processes that include interactions falling within a modest range – payroll processing, order processing, compliance checks – are increasingly automated.

And that’s just what conventional AI is doing. As generative AI overturns entire professions and begins weaving itself into almost every business role where information-gathering and human communication are involved, the percentage of time spent on tasks it can do will reduce the number of human beings needed to fill roles of all kinds, across the board.

In short, the advent of ubiquitous AI will change the Master-and-Mule equation forever.

Machines are already the Mule; now machines are taking over for human beings as the Master.

Master-and-Mule will be, increasingly, machine-and-machine.

So, yes, this time is different...

The Beginning of the End of Human Labor

Robots were actually the beginning of the end of human labor, and they’ve been around for several decades; but, per the Master/Mule argument above, the automation of human cognitive tasks, beyond physical labor, heralds the true starting point of what AI experts have long called the Fourth Industrial Revolution, and more recently the Second Machine Age. Formally, those definitions are as follows:

The Fourth Industrial Revolution: describes the shifts in technology, industry and society that result from ubiquitous connectivity (the Internet) and smart automation (AI).1

The Second Machine Age: describes the automation of human cognitive tasks, rendering humans unnecessary to production processes.2

We can pin down the starting gun for this new era in 2012, when deep learning neural network methodology went from theoretical to practical. The seminal event was the unveiling of AlexNet, a convolutional neural network model that broke all records in image recognition. That’s when deep learning became the new foundation of AI for commercial development.

In the 10+ years since, many deep learning-based applications have emerged: facial recognition systems; natural language processing; self-driving vehicles, at least in closed environments (factory floors and shipyards).

But early in 2023, when ChatGPT and other generative AI engines were introduced for commercial exploitation, it became clear that we had arrived in a completely new place: AI can now not only do things we do with our brains, but creative things we do with our brains.

It can write essays, emails, reports, ads, and blog posts in a split second; it can generate custom images in any style, any context, on demand; it can spontaneously create video and music. This output isn’t particularly sophisticated, mind you, but neither is most of the workaday writing and art we crank out at our desks. It just has to be understandable, clear, and accurate – which generative AI can do. (And, of course, in 10 years it will do it 100 times better.)

In short, it is poised to take over a vast proportion of work – cognitive work – that until this point in history could only be done by humans. And it’s here, it’s freely available (and ridiculously inexpensive), and it’s already being widely applied and will be deeply disruptive.

More questions arise: how fast will generative AI grow? How quickly will it become commonplace in business? Exactly how disruptive is it going to be?

What the experts say

Estimates as to how many jobs will be taken over by AI, what professional domains will be affected, and when it will happen are understandably variable.

The McKinsey Global Institute has predicted that half of all jobs will be automated by the year 2055.3 It also stated that 60% of existing jobs can be at least 30% automated today, that between 400 and 800 million jobs will be lost to automation by 2030, and that less than 10% of those who lose their jobs will be able to shift to new occupations.
The World Economic Forum reported that the ratio of human to machine labor in the major industries is 7/3. It predicted that this would shift to 6/4 by 2022, and stated that by 2025, about 80 million jobs would become automated.
Goldman Sachs issued a report in March 2023 stating that in the US and Europe, 2/3rds of jobs are exposed to some degree of AI automation, and that globally, 300 million jobs are potentially at risk.4
Gartner placed the automation replacement estimate at 1/3rd of all existing jobs, and stated that this would occur by 2025 (that’s unlikely, but the actual replacement ratio won’t be trivial).
A report from the Organisation for Economic Co-operation and Development stated that 14% of jobs across 32 European nations are already 70% automatable. Another 32% are 50-70% automatable, putting 210 million jobs at risk.5
In February 2023, the career planning site Zippia reported research that placed the numbers at 20 million additional manufacturing jobs lost in the US by 2030; an industrial robot population increase of 40,000 per year between now and then (adding to the current population of 310,000); a total job loss of 73 million by 2030, across industries; and that, at present, 25% of US jobs are automatable.
Futurist Nicholas Badminton has estimated that all jobs everywhere, of every kind, will be replaced by AI or robots within 120 years;6 that translators will be out of work in 2024; authors by 2049; and surgeons by 2053.
Forrester estimates that nations in the Asian Pacific are at greater risk than in North America and Europe. It forecasts 63 million jobs will be lost to automation, with another 247 million placed at increasing risk. It estimates European job loss by 2040 at 12 million.
And Elon Musk has said that AI could be outperforming human beings by 2030.

These estimates, of course, are only estimates, though they come from serious and credible sources. Even so, they vary widely.

What we can be sure of is that whatever the eventual numbers turn out to be, they will be big ones. What happens then?

“If we get the fourth industrial revolution right, digitalization will benefit the nearly 10 billion humans inhabiting our planet in the year 2050,” said Siemens CEO Joe Kaeser at the 2018 World Economic Forum. “If we get it wrong, societies will be divided into winners and losers, social unrest and anarchy will arise, the glue that holds societies and communities together will disintegrate, and citizens will no longer believe that governments are able to fulfill their purpose of enforcing the rule of law and providing security.”

The stakes are high.

The Jobs That Are Going Away

The opinions of experts vary on the when and the how many of the jobs that will go away. They're pretty much in step, however, regarding the order in which those jobs will go.

Obviously, the more routine and simple the tasks required by the job, the easier it will be to automate and the sooner it will go. There are tens of millions of such jobs – in particular, jobs where consistency and unvarying task execution are very much the point (as in the service industry). There are jobs that consist of only one, or very few, tasks (as in data entry). There are jobs where the only moving parts are the collecting of information and the dispensing of information (clerks).

The service industry, to be sure, is vast and varied – from retail to fast food to hospitality and travel planning to help desks to janitorial services. Lots going on there, and some of it can be intimidatingly complex. But all are about the delivery of information or a simple, consistent product; all involve collecting predictable information from the customer; all deliver a fixed and consistent set of expected outcomes. That’s automation territory.

These five occupations, then, top the list of those on the way out:

Service Industry/Retail

Data Entry

Clerks/Administrative Assistants

Drivers

Manufacturing

With just these three – all of which are already being automated at a steady rate – we will see the removal of millions of people from the US workforce.

Within the service industry, which accounts for nearly 80% of the US workforce, the employment numbers are roughly as follows:

Food servers: 2,200,000
Food preparation workers and cooks: 1,700,000
Retail workers: 4,100,000
Cashiers: 3,300,000
Customer service representatives: 2,100,000
Janitorial workers: 2,000,000
Maids and housekeepers: 860,000

This is only a sampling, but it conveys the scope of this sector’s workforce percentage. The wages paid in these roles are among the lowest.

Andrew Yang, entrepreneur and former presidential candidate, noted that the Obama Administration, on its last day, published a report forecasting the loss of 83% of all jobs in the US paying $20/hr or lower by 2030.

Retail workers are already being let go at alarming rates, triggered by the rise of e-commerce. The “Retail Apocalypse”, in motion now for two decades, has reduced the number of US malls from 2,500 in the Eighties to around 700 today, according to Business Insider. And that number will drop, it reports, to a meager 150 over the next decade. Physical store closures beyond malls: over 12,000 in 2017 alone. This will only increase in the coming years – malls won’t be making a comeback.

Call centers employ 2.5 million people today (2023). That number is going to shrink drastically, as intelligent chatbots continue to proliferate, driven by the astonishing advances in generative AI. Those bots will not only be as effective (if not more) in solving caller problems as their human peers, but will look and sound increasingly life-like – to the point of being indistinguishable, in the very near future.

Anyone who buys groceries is well aware that you can now check yourself out, with no cashier (more than 330,000 cashiers’ positions will be lost in the 2020s, according to the US Bureau of Labor Statistics). And stock management in retail, which my manager Landon did with paper in pen when I worked on his night crew at Kroger’s in my youth, is now almost completely automated.

And when we blame those job losses on e-commerce, we don’t just mean the Internet: poke your head in any Amazon fulfillment center or warehouse and you’ll see robots, end-to-end.

Among clerks and administrative assistants, of which there are currently 4,500,000, almost half will be out of work by 2030 (again, US Bureau of Labor Statistics). There are almost a million receptionists and information desk workers, who will experience even greater levels of replacement as AI improves at conversational speech. Stock clerks and order fulfillment jobs fall into the Amazon paradigm – 1,800,000 are at risk and already being replaced. Bank tellers, already long at the mercy of the ATM: 550,000 remaining and at risk.

One of the biggest occupations targeted for automation over the next decade is not in immediate peril today, but is soon to be decimated, is that of truck driver. Amazon (and others) have been developing driverless vehicle technology for almost a decade, and at present the success rate of those self-driving vehicles is 98 percent. When it asymptotically approaches 100, 3.5 million truckers will find themselves out of work (truck driving is, in fact, the most common job in 29 US states).

This phase-out of truckers will happen in layers. The first mass deployment of driverless trucks will have a human driver on board as fail-safe. Next will come the practice of letting rigs loose on interstates, then turning them over to human drivers as they enter cities. Finally, completely human-free trucks will be on the roads with tele-operators on hand to take over remotely when they are in high-risk urban environments. When the technology completely surpasses human performance under even the most dangerous conditions, these, too, will be phased out.

This will begin in 2024, and will be ubiquitous within a decade.

As for manufacturing, robots have been replacing humans in factories for more than 50 years, so available jobs have been falling in that industry for all of that time (though many of those jobs went overseas – to factories populated by robots). Of those remaining, 2.1 million will go away by 2030, according to a study published by Deloitte and The Manufacturing Institute.

Data entry workers are not nearly as numerous, but there are about 160,000 of them in the US. The work is attractive because it is easily done from home now. And it is easily automated, so those jobs are going away.

Next come these:

Accounting

Paralegals

Marketing (Telemarketing)

Security Guards

Accounting should not be a surprising entry here. The job is, of course, mostly numbers, and computers are extremely good with numbers. Most accounting processes are very monolithic, and the value of the human component is largely in forensics – explaining the numbers to other humans, when things don’t make sense or aren’t readily clear. Business analytics software currently used by human analysts is already proficient at diagnostic analytics; as AI improves, it will take over this last accounting task. At risk: more than a million accountants and auditors in the workforce right now, along with half again as many bookkeepers and auditing clerks.

A surprising entry here is the paralegal. Those who assist attorneys would seem to be specialists themselves, highly educated and trained beyond a level that would be thought of as easily replaceable. And, of course, paralegals are highly educated and trained. But the fact is, most spend their time searching documents. Their primary task is reading, poring through legal documents of all kinds, because it’s cheaper for them to spend their time doing that than it would be for those they work for. And, of course, almost all documents are now electronic, even transitory ones. Search technology, in the Internet era, is perfected; so it’s a question of scanning documents for the parts the lawyer needs – and AI can now do that. At risk: 345,000 jobs in the US.

Anyone with a laptop, tablet, or smart phone knows that automation has already taken over marketing in most industries. And everyone who’s ever said something like “fitness center” or “canoe paddles” in the presence of their own smart phone will almost immediately see ads for local gyms and sports stores popping up in their social media feeds. It’s AI, of course, pouring the ads into the feeds. And marketing itself is becoming increasingly personalized – a customer management trend for more than five years now. Marketing is already fast becoming the domain of AI for all but the most high-end product and services, where human contact is still essential.

And there’s one branch of marketing – telemarketing – that has been automating for years, but is about to take another leap forward, thanks to generative AI. This new tech can create human voice approximation indistinguishable from the real thing. And in online telemarketing, where you actually see someone, you soon won’t be able to tell a real person on a screen from a fake one.

The need for human security guards is already in steady decline as physical surveillance technology has improved, and owing especially to the Internet of Things – a term that describes physical sensory devices attached to the Internet, such as cameras, sensors, microphones, alarms, and so on. An array of IoT hardware can guard a factory, warehouse, bank, or office complex far more thoroughly than human guards, responding more quickly and correctly when something is out of the ordinary. Human security will be reduced to only those roles where physical intimidation is useful. At risk: more than a million security guards currently working.

We’ve already accounted for around 50 percent of the US workforce with just these occupations. But there is another entire layer of jobs that won’t necessarily be erased by AI, but will go away anyway because they are roles that support the occupations that are AI-related.

Office Managers

Retail Supervisors

Food Service Supervisors

Transportation Support Services

In a nation with dwindling office workers, there will be less need for supervisors of those workers. That’s 1.3 million more jobs at risk. The same applies in food service, where 750,000 supervise the work of the people actually handling the food. In retail, the number is 1.2 million.

In transportation support services, the numbers are staggering (though they do overlap with roles that will be phased out anyway). Consider that the trucking industry is supported by the peripheral industry of truck stops, where accommodations for drivers include restaurants, lodging, and other human essentials beyond fueling. The latter will still be necessary, but all the others go away. That’s 5 million jobs.

Other occupations that will vanish soon include:

Inspectors

Hiring and Recruitment

Procurement

Proofreaders

Medical Diagnostics

Supply Chain Management

Tax Preparation

Loan Officer

Insurance Underwriting

The Jobs That Are Safe

Some jobs, a large number of experts assure us, aren’t going anywhere.

These occupations and professions make up a far shorter list – shorter because they require skills AI can’t easily replicate, or possess a human component that is beyond any technology. Or, in a few cases, because people are always going to prefer a human to a machine in a particular interaction.

The lists below are derived from research from the University of Oxford8 and reports from the US Career Institute.9 We can group these safe havens by the things that make them safe from AI incursion: Complexity, Understanding of Human Beings, Aesthetics, Distrust of Machines, and the need for the Human Touch:

Complexity

Some professions will not easily yield to AI because they are simply so complex that there is no path by which AI can reach them, as yet. That’s somewhat counterintuitive, because AI by its very nature is adept at complexity.

But some complexity is cross-domain: it may encompass intricacies in one area that are matters of rules, combining them with subtleties that have nothing to do with the rules. Or there may be straightforward processes and procedures to be executed, but they must be executed in a changing environment where unpredictable events and conditions may surface.

Lawyer

Construction Management

EMT/Firefighter

Civil Engineer

The practice of law, for instance, involves the application of highly complex bodies of rules, judgments, structured accommodations in execution of policy and mitigation of conflict. This is a significant challenge for human minds to do well, let alone AI; but it is complicated further still by the fact that this very intricate system of social governance must exist within the framework of messy, less-structured human behaviors and emotions, within an even more inconsistent convolutions of the human social lattice. Put simply, the practice of law is deep complexity nested within deep complexity.

The same is true in the utterly dissimilar domain of construction management. The undertaking of building great structures involves rules, intricate processes, meticulous planning, best practices – all within the often dicey clouds of environmental complications and sometimes fickle human push-back – another potentially very complicated juggling act.

EMTs and firefighters must apply very rigid and specific remedies, often incomplete, under very chaotic conditions. Civil engineers must integrate optimized planning within economic and environmental constraints, buffeted by social and political complications.

These complexities make these professions a safe haven from AI for some time to come – although, as with the professions below, AI can be a very capable assistant in practicing them well.

Understanding of Human Beings

Some professions are so completely about human behavior and emotions that there is, at present and in the foreseeable future, no way for an AI to master them. These are professions where the problems themselves being experienced by human beings are not readily reduceable to data, and where the subtleties of interaction in engaging them are beyond AI learning.

Social Work

Psychology/Psychiatry/Therapy

These professions are superb examples of work that is so immersed in human understanding and contact that AI couldn’t provide it, even if it could “understand” humans in the way that behavioral health professionals and social workers do. And while each of these professions is deeply steeped in science and empiricism, there is an undeniable intuitive aspect to their practice. The human beings who undertake these roles as their life’s work often succeed through their own strong sense of the emotional states of other human beings. That’s not AI’s domain, at least not for a very long time.

Aesthetics

Some professions may be within reach of AI and robots, but incorporate artistic attributes that are, as yet, beyond AI’s reach – and which, even if achieved by machines, will still be preferred rendered by human hands by their consumers.

Some Design/Artistic Roles

Some Writing Roles

Some Musical Production Roles

Architects

Landscaping/Interior Design

Tradespeople

These professions will remain in human hands indefinitely to some degree, simply because human beings often prefer the handiwork of other human beings. Here in the early 2020s, as this is being written, generative AI – ChatGPT and its myriad siblings – can write and create art and music convincingly, at least well enough to meet the mass content requirements of the Internet. But content produced by human artistry will always have a place, even if competition in that quarter grows more intense.

This extends to the design of homes, which can certainly be automated; but habitats designed by those who live in habitats will always resonate with at least some portion of those who need such designs. This is even more true of those who craft interiors for those habitats, and who sculpt the landscape around them; yes, machines can and will do these things, and do them well, but the human aesthetic will always have a place.

A very interesting case is that of human craftsmanship. A friend recently argued that custom cabinetry, for instance, cannot be roboticized anytime soon, because the range of high-dexterity skills required of human carpenters won’t exist in robots until entirely new forms of AI come to pass (AGI in particular). What is already changing, in the meantime, is the move to modularizing in home construction and other environments where tradespeople have traditionally ruled: cabinets won’t be built along with the building, but will be separately constructed – by robots – for drop-in installation. So many tradespeople will, in fact, be out of work.

But not all. All tradespeople who work with their hands and build custom products of all kinds will remain valued for their ability to create-to-order, adding human touches that would never occur to an AI. That niche will thrive, no matter how much of the general work is automated.

Distrust of Machines

There are several professions that may be within reach of AI, and which AI might actually do better – perhaps astronomically better – than human beings. They are, however, roles that human beings will overwhelmingly prefer to remain in the hands of human beings, to the point of being rejected if deployed by machines.

Politicians

Medical Professionals

IT/AI Technologists

These are jobs that AI could probably do much better. Sci-fi writer Isaac Asimov wrote of machines that run human affairs in the future, handling all the big decisions – effectively replacing the politics we depend on today.

And it wouldn’t take much for those machines to do a better job than the nincompoops and charlatans who routinely slither into office today. Even so, it is hard to imagine that most of us would embrace AI leadership over society, or legislation generated by machines – even it was demonstrably more fair, equitable, and effective.

The same is true of medical professionals. AI can already out-doctor some doctors (radiology is already a dying field, because AI does it so much more accurately), and surgeons will yield to robots within a generation. But on the whole, we trust human doctors – and human nurses, even more so – because they are human. The empathy a doctor or nurse can offer is something that is as important to us as the actual treatment, and human-to-human empathy isn’t on the AI table.

Finally, there’s AI itself. We already have AI that can produce effective computer code faster and more flawlessly than human application developers. It will get better and better, moving forward. But do we want AI that is designed by AI? That’s the stuff of science fiction nightmares. Even it was better, even if it was utterly trustworthy, there is little chance that we’d be willing to live with it.

Human Touch

As above, there are roles that human beings want to see filled by other human beings – but out of a positive dynamic, not one of distrust. There are some services we receive from other human beings that are a pleasure to engage in precisely because we enjoy the human interaction, and wouldn’t want to give it up.

Event Planners

Bereavement

Recreational Therapist/Fitness Instructor

Veterinarians/Pet Care

Professional Sports

That empathy factor is even more present in professions where the human contact isn’t just part of the service – it's something we seek out. It’s easy to imagine an AI that specializes in wedding planning being cheaper than a human wedding planner and just as effective, but part of the fun of wedding planning is working with the wedding planner, who specializes in making a series of overwhelming and stressful tasks seem fun and exciting.

At the other end of the spectrum is bereavement. When someone we love dies, we must turn to someone who will take care of the details we find emotionally distressful and doing so with compassion. An AI could easily fill this role, but wouldn’t offer that compassion.

And would you turn your pet over to a robot for either surgery or daily care when you’re away? AIs could (and will) do these things, but do you want to put your pet through that?

Finally, sports.

Robots will play tennis better than we do in our lifetimes. And basketball. And baseball. That’s going to be thrilling, and we will enjoy it. But will human sports go away? Almost certainly not, because we will clearly see, when it happens, that robots playing baseball is not even remotely the same thing as a human game. All of the emotional resonance that we feel when we look out at the field at the bottom of the ninth, with bases loaded, as the pitch is thrown and every player on the field is utterly wound up in the tension of the moment, will completely capsize the bobbing emptiness of the nothing we feel when we observe robot players in the same scenario, as emotionally indifferent to the situation as they would be in the dugout.

The Limits of Retraining

At the beginning of the 20th century, almost half of all Americans worked in the agriculture industry in one way or another. Today, the number is less than 2%.

Those millions of workers migrated into factories, which boomed for decades. And when those jobs began to be shipped overseas, they shifted into service industries.

Now the service industries themselves are going away. Where are the tens of millions of workers tallied above going to go next?

One of the stock answers to that question is, “This is how displacement works; old jobs go away, and even more new jobs will come along to replace them.”

That’s been provisionally true through the displacements outlined above; but the automation revolution is different-in-kind. There is, in fact, nowhere for those workers to go.

Andrew Yang, again:

“We actually are getting rid of the most common jobs in the US economy, filled by high school graduates, and then replacing them with a handful of jobs for higher-skilled people in different places. And then we’re pretending that the first population is somehow gonna access the new opportunities, when the odds of them getting up and moving to Seattle and becoming a web designer or logistics manager or big data scientist or something are essentially zero.”

The follow-up to that stock answer is that we could have the government offer programs to retrain those displaced workers, which was done to offer transition assistance to the displaced factor workers of the previous revolution.

“Government-funded retraining programs had a success rate of between zero and fifteen percent,” Yang continued. “This is what actually happened to the workers of Michigan and Indiana and Ohio. And so, if you say we’re going to retrain these people, then you also have to come up with a way for us to become amazing at something that right now we’re really, really bad at.

“And if you were an employer, would you rather employ a 50-year-old former truck driver with health problems who got some certificate program, or would you rather hire a 25-year-old kid who went to community college, is probably cheaper, has lower expectations, and his skills are natively gonna be a little fresher?"

The follow-up to retraining is, then, retraining for what? There aren’t 3.5 million jobs needing to be filled in the tech industry, and that’s just the truckers; what about the tens of millions of other displaced workers? Even if you could effectively retrain them, what would it be for?

As for the new jobs arising as AI takes hold, yes, there are going to be many, and they are very exciting, and will be discussed in an upcoming chapter. But all those jobs together represent only a small fraction of what will be lost. So, in the face of the coming jobs upheaval, there is plenty for an industrious person who is determined to do well to work with – but competition will be fierce. We can take nothing for granted.

In Summary...

Stay away from these occupations – or, if you’re already in one of them, consider moving in the very near future: