We Just Don’t Know the Impact AI Will Have on Jobs

I like to think I know HR and workforce issues pretty well. I’ve spent a large part of my life researching them. But the more I read about these new technologies, the more I think we just don’t know the impact AI will have on jobs. We’re going to have to learn as we go.

Driverless Cars Are Coming, Till They’re Not, Or Are They?

Just a few years ago, the standard thinking was that a lot of blue-collar jobs would soon disappear due to the advent of driverless vehicles. But it turns out that driving is a lot harder than it looks and AIs just can’t handle the “edge cases” well. Sure, they can handle 98% of driving, but they can’t cope with the unexpected. For example, there’s the case of the self-driving car in San Francisco that didn’t obey an officer’s command to halt when it was in an area where firefighters were working. The cop smashed in the car’s windshield to stop it. That’s the kind of edge case that goes viral.

So, all those ride-sharing and taxi jobs are safe, right? That’s become the new conventional wisdom. For now.

“The white-collar employee’s future is more threatened than the Uber driver, because we still don’t have self-driving cars, but AI can certainly write reports,” Martin Ford, author of Rule of the Robots: How Artificial Intelligence Will Transform Everything, told the BBC.

Well, pardon my skepticism ,Mr. Ford, but I doubt you really know. Right now, there are hundreds of driverless cars in California, many in San Francisco. In fact, there are over 1,400 such vehicles registered in that state, up from just 900 last November, according to the Department of Motor Vehicles. In some parts of San Francisco, these vehicles have become a pretty common sight.

Now, maybe this surge is temporary. Maybe the public will turn against these cars, especially if there’s some well-publicized gruesome accident costing multiple lives. Or, maybe these cars will start appearing in city centers all over California and then beyond. If someone can pay a fraction of a price of an Uber for a ride in a driverless car, you can bet a lot of people will be willing to give them a try.

The point is, we just don’t know for sure. Mr. Ford is wrong in that we do have self-driving cars, we just don’t have a lot of them yet. The future is, as they say, already here but still unevenly distributed. Things could change…quickly….or not.

Now White-Collar Jobs Are Expected to Evaporate, But Will They?

The new conventional wisdom is that blue-collar jobs are safe while white-collar jobs are on the chopping block. Are all those software development jobs, for example, going out the window because the new generative AIs are pretty good at writing computer code?

Maybe. There are stories of programmers who are able to boost their coding productivity by three times or more as they leverage ChatGPT or other AIs. Let’s assume for a minute that the productivity claim is true. Does it mean that 2 out of 3 programming jobs are now expendable?

Could be. Or, it may be there’s a whole lot of development work out there that companies couldn’t get to because they just couldn’t afford to hire enough developers. If these professionals are three times more productive, then companies stand to earn more money per worker more quickly and can afford to hire more programmers.

Consider the Case of the ATM

Consider the case of the bank teller. For a while, the conventional wisdom was that automated teller machines, or ATMs, would cast bank teller jobs into the dustbin of history. But that’s not what happened. Since 2000, in fact, teller jobs have grown a little faster than the the labor force as a whole. The impact of the ATM machine was not to destroy teller jobs but to increase them.

James Pethokoukis writes, “What happened? Well, the average bank branch in an urban area required about 21 tellers. That was cut because of the ATM machine to about 13 tellers. But that meant it was cheaper to operate a branch. Well, banks wanted, in part because of deregulation [but also] for basic marketing reasons, to increase the number of branch offices. And when it became cheaper to do so, demand for branch offices increased. And as a result, demand for bank tellers increased.”

Tech Wiped Out Most Farm Jobs, Right?

Now let’s consider the job of the agricultural worker. Farming jobs have become so technologically productive that they are practically the poster child for productive-tech-kills-jobs idea. And, it’s true. We do have many fewer farmers per capita today than we did before the industrial revolution.


Guess what the top role is expected to be over the next five years in terms of net job growth?

If you’re sensing a theme here, then you’ve probably guessed it: agricultural equipment operators. The World Economic Forum states, “Surveys conducted for the Future of Jobs Report suggest that the highest job growth in 2023-2027 will be for agricultural equipment operators, for drivers of heavy trucks and buses, and for vocational education teachers. Mechanics and machinery repairers follow in fourth place.”

But Surely the Writers Are Doomed!

Now let’s talk about the most obvious victims of large language model AIs: journalists and other writers. Why should anyone hire a writer when they can get generative AI to write virtually for free?

Good question. Maybe there will be a vast reduction in such jobs. Why not? After all, journalism jobs have been on the decline for decades. The Center on Education and the Workforce reports, “Projected job losses for journalists are primarily due to newspaper downsizing and closures….[O]nly about 15% of journalism majors become editors or news analysts, reporters, and correspondents early in their careers.”

Now ChatGPT et al. will finish the job of killing off journalism jobs for good. Right?

It could happen. On the other hand, what if all these small-town newspapers that have closed over the years because the Internet nuked their business models suddenly become modestly profitable again because the AIs can inexpensively produce copy for online (or even print!) editions? If that happens, those papers will still need some actual journalists to attend the townhall meetings, investigate important local stories via interviews, track down leads, etc.

It’s possible–though I don’t know how possible–that these new technologies will actually lead to more journalism jobs in the same way that more ATMs were correlated with more bank teller jobs. High productivity is like that. It can be stochastic in its effect, so you can’t always anticipate the economic outcomes of rising productivity rates.

Don’t Pretend You’re Certain About Anything

Look, I’m not saying we shouldn’t try to forecast the future or that all predictions are doomed to be wrong. But, as Niels Bohr reportedly said to the legendary Yogi Berra, “It is difficult to make predictions, especially about the future.” 

Sometimes predictions that seem obvious today can prove to be dead wrong, and there will always be some future events capable of surprising us.

So, stay humble, you futurists, forecasters and would-be guru types. None of us really knows how any of this is going to turn out. And that’s okay. We’ll make our best guesses and then figure it out as we go.

AI Won’t Be Just for White-collar Workers

AR for Everybody!

Over the next six to twelve months, we’re going to see a surge in usage of and spending on augmented reality (AR) glasses that are enhanced by GPT-based artificial intelligences (AI). Why? Because blue-collar workers and prosumers will ultimately get a lot of utility out of them.

For example, the Bing chatbot will likely empower HoloLens 2 users. Open AI’s ChatGPT-4 will probably be feeding directly into Vuzix brand smart glasses. And Facebook’s LLaMA AI will somehow be integrated into the not-yet-released Meta Quest 3 (formerly known as the Oculus Quest brand) and/or Raybans.

Although Google recently discontinued Glass Enterprise smartglasses, they’ll no doubt find AR glasses manufacturers to match up to Bard. Meanwhile, there’s no way Apple will be deprived its entry into the space.

Blue-Collar Work Gets Seriously Augmented

Why do I think this is going to happen? The main reason is that, unless there’s some huge crisis associated with one of these AIs in the meantime (it could happen), people at work finally have a legit reason to wear these gadgets, especially in blue-collar jobs.

I don’t mean that it’s going to happen overnight. It’ll start slowly and become more popular over time. But it’ll come because these AIs will all eventually be able to interpret images. In fact, ChatGPT-4 already has this capability.

This potentially changes the working lives not only of various types of blue-collar employees but of avid prosumers as well.

Bing Creates a Scenario

I asked Bing to create a scenario in which an electronics mechanic uses a HoloLens that’s enabled with an AI-powered image recognition module. I specified the person is an experienced electronics mechanic who already knows how to address basic jobs such as fixing a resistor.

Here’s what Bing came up with:

Alice arrives at the customer’s house and inspects the drone that is lying on the table. She notices that it is a DJI Mavic Air 2, a model that she has not worked on before. She turns on her HoloLens headset and says:

“Hello Bing, I need your help with repairing this drone. It has a damaged circuit board. Can you show me a schematic diagram of this model?”

Bing responds: “Sure, Alice. I have found a schematic diagram of the DJI Mavic Air 2 from the official website. Here it is.”

Bing projects a holographic image of the drone’s circuit board on Alice’s headset, highlighting the different components and their functions.

Alice examines the image and compares it with the actual circuit board. She sees that one of the capacitors is burned out and needs to be replaced.

She says: “Bing, can you tell me the specifications of this capacitor?”

Bing responds: “This capacitor is a 100 microfarad electrolytic capacitor with a voltage rating of 16 volts. It is used to smooth out the voltage fluctuations from the battery.”

Alice says: “Thanks, Bing. Do I have one of these in my toolbox?”

Bing responds: “Yes, you do. It is in the third compartment from the left, labeled C3.”

Alice opens her toolbox and finds the capacitor. She uses her soldering iron to remove the old capacitor and replace it with the new one.

She says: “Bing, can you check if I have done it correctly?”

Bing responds: “Yes, you have. The capacitor is soldered correctly and there are no short circuits or loose connections.”

Alice says: “Great. Is there anything else I need to do?”

Bing responds: “No, that’s all. You have successfully repaired the drone’s circuit board. You can now test it by turning on the drone and checking its functions.”

Alice says: “Thank you, Bing. You have been very helpful.”

Bing responds: “You’re welcome, Alice. I’m glad I could assist you. Have a nice day.”

After Bing provided this scenario, I asked, “How would Bing know what is and isn’t in her toolbox?”

It replied, “Bing knows what is and isn’t in Alice’s toolbox because she has scanned and registered her tools with the HoloLens headset before using it. Bing can access the information from the cloud storage where Alice’s data is stored and synced. Bing can also use the graphics recognition module to identify the tools visually and match them with the specifications.”

Okay, I’m not even going to dwell on the uncanniness of Bing’s ability to write a fiction and then seamlessly fill in any plot holes after the fact. The point is that we can see how an experienced professional might be able to use an AI-enhanced, image-recognition device.

Start with a Smartphone

Most of this will be done by smartphones in the near future. But people often need to work with both hands, so wearing the device like a pair of eyeglasses makes more sense in the long run.

So, yes, smartphones will come first, smart glasses second. The first movers will adopt glasses quickly, the fast followers the next year, and within five years smart glasses will be commonplace.

The Weekend Mechanic

Then there’ll be the prosumers, the folks who want the latest gear to pursue their weekend hobbies, whether that’s working on their car, building their electronic doodads, doing home improvement jobs, or anything else you can think of.

In truth, the prosumer market will probably be more lucrative than the pro market because A) there are more prosumers, and B) professional mechanics and other skilled workers already know how to do their jobs. They’ve developed habits that will be hard to break. But that doesn’t mean they won’t come around if they really believe the tech will help them.

AR Finally Gets Useful

Up until now, most use cases for AR glasses have been pretty limited. With the rise of these new AIs, however, everyone gets access to an enormous electronic brain jam-packed with detailed information.

What’s more, the AIs themselves will get better over time at understanding the real-world problems inherent in these professions. And, unlike most of us, the AIs aren’t likely to forget the details of what they’ve learned over time.

Eventually there will be talk of robots doing these kinds of work. That may happen, but long before then, there will be skilled workers as well as passionate weekend amateurs who are made better via AI-enabled smart glasses.

The Dangers of BS

But there’s also a danger here: that all forms of these AIs so far generate incorrect information at times. Some have even called them BS machines because they so confidently put forward made-up facts.

This is annoying for journalists, who must fact-check everything coming out of an AI. But it could potentially be deadly for people like mechanics, electricians and those who rely on their work.

So either these systems become more trustworthy over time (possibly by relying only on databases specific to certain professions), or workers must continue to rely on current quality assurance procedures, never putting full faith in AIs on the job. If the latter, some will argue against their usage in the first place, and they may carry the day in some professions or organizations. We’ll have to see how it plays out. I imagine there will be a tug of war in both directions.