FREE NEWSLETTER
IN A RECENT INTERVIEW, Dario Amodei, CEO of Anthropic, a leader in artificial intelligence, grabbed headlines. Amodei argued that the next generation of AI systems could replace half of entry-level jobs and drive up the unemployment rate to 20%. All of this could occur in the next five years, he said.
Recent data seem to support these glum predictions. Mark Zuckerberg said AI will be as capable as a mid-level programmer by the end of this year. Microsoft announced thousands of job cuts this spring, with programmers disproportionately affected. In an announcement last week, Amazon’s CEO wrote, “It’s hard to know exactly where this nets out over time, but in the next few years, we expect artificial intelligence to reduce our total corporate workforce….”
If these data points are a sign of things to come, it would certainly be concerning. This outcome isn’t guaranteed, though. To illustrate why I see things differently, let me share a recollection from some years ago.
My father was a lawyer, and I remember visiting his office when I was a child. In hindsight, what was notable was there was no computer on his desk. There were no computers anywhere. To send correspondence, attorneys spoke into dictation machines. They then handed the machines’ miniature tapes to secretaries, who literally typed up a first draft on typewriters, using Wite Out to fix errors and carbon paper to make copies. The draft was then marked up in pen, a final copy retyped, placed in a stamped envelope, and mailed.
Because of the amount of work involved, each attorney had a dedicated secretary who spent his or her days typing and retyping documents in the manner described above. Today, everyone seems to do their own correspondence and, as a result, there are far fewer secretarial jobs. But if we look back at historical data, we see that the invention of both the word processor and email didn’t cause any noticeable increase in unemployment. Why not? For starters, transitions like this occur slowly.
Also, new technologies are rarely a net negative. Instead, they tend to create new jobs. Take a look at today’s law firm. It has far fewer traditional secretaries but significant numbers of IT people. (IT was nonexistent in the days of typewriters and dictaphones.)
The transition in office work is the most recent change, but it’s by no means unique. Years ago, many people were employed as Morse code operators. Sam Altman, founder of OpenAI, commented that in the past there were also large numbers of people employed as lamplighters who traveled the city streets each night lighting gas lamps. More significantly, farming used to be a major sector of the workforce. In 1900, 40% of Americans worked on farms. Today, it’s less than 2%.
These transitions were all significant, but none caused the sort of mass unemployment Dario Amodei forecasted. In fact, Amazon said it doesn’t expect AI to result in significant layoffs. Instead, the bulk of the headcount reduction is expected to occur through attrition—the normal course of employees changing jobs or retiring.
Amazon provides another useful data point on this topic: The company now uses 750,000 robots in its warehouses. In theory, those robots would have taken more than 750,000 jobs, but that’s not what the overall employment data show. Unemployment today is near the low end of where it’s been over the past 75 years.
How is it possible that technology has displaced jobs, and yet unemployment remains low? An economic principle known as the Jevons paradox can help us understand this. In the 1860s, William Jevons, a British economist, observed that manufacturing plants had become more efficient in their use of coal and yet, counterintuitively, the demand for coal was increasing.
The explanation: As manufacturers realized they needed less coal to produce the same amount of output, they chose to expand their businesses into new areas, resulting in the use of more coal.
Over time, a second order effect kicked in. These gains in output led to wage increases and faster economic growth. That, in turn, further increased demand for coal.
The adoption of artificial intelligence might deliver the same positive effects, with greater productivity leading to higher wages and faster economic growth without any loss in employment.
Recent comments by the CEO of software vendor Box illustrate how the Jevons paradox might apply to AI. “AI is not replacing existing work that’s being done,” he wrote, “but adding new capabilities to the organization.” Box, in other words, won’t use AI to cut costs; it will use the technology to do more.
He adds: “This means that companies can simply now attack the kinds of problems that just never were economically feasible to solve before…. Yes, there’s certainly opportunity to automate some of the work that we currently do to drive efficiency, but the vast majority of work that we will bring automation to is the work that we just never got around to in the first place.”
Interestingly, despite his concerns about employment, Amodei sees some of these same benefits. In the same interview in which he made his comments about unemployment, he also described some of the potential positive effects that AI might deliver: “Cancer is cured, the economy grows at 10% a year, the budget is balanced….”
The reality is that this is all an open question. In a May interview, economist Daron Acemoglu, who recently won the Nobel Prize for his work on economic development, argues that AI will be able to replace only a fraction of jobs. But he adds, “It’s hugely uncertain, and it’s very difficult to know because it’s a very rapidly changing technology.”
And that just may be the best way to think about AI. It’s all very new and still uncertain. While Amodei worries about a potentially negative impact, that’s just a guess. Economic history suggests it may very well go the other way.
Adam M. Grossman is the founder of Mayport, a fixed-fee wealth management firm. Sign up for Adam’s Daily Ideas email, follow him on X @AdamMGrossman and check out his earlier articles.
Want to receive our weekly newsletter? Sign up now. How about our daily alert about the site's latest posts? Join the list.
Great post Adam, and I hope (and in most ways agree) you are correct. That doesn’t negate that there will be some displacement, as workers will need to train, and retrain, in new skills, and in that process there will be those who for whatever reason are not able to make that transition. My concern in this case is that the pace of change may be far faster than anything we’ve seen before, and while I fully expect society will roll with the changes, it may be a much bumpier ride than ever before.
The three horsemen of the job apocalypse are AI, general automation and robotics. Where we will see their destructive power is after the next global financial crisis and the deflation of the “everything” bubble. Instead of jobs rebounding as they have in the past, many jobs will never return as companies find ways of doing a lot more with a lot less manpower. Sounds bad, but it doesn’t have to be. Huge productivity advances are beneficial to society and can make all lives better. The question our future society must answer is, will the advances in living quality be shared or will we continue down the road of haves and have nots.
Excellent article. In the near future, AI agents are going to be part of the work force along with full time employees. This would be a big change. See link
https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-future-of-work-is-agentic?cid=soc-web
Humans have historically been very resilient by finding creative ways to survive and grow. This time will be no different.
As someone who worked in IT for decades, I saw firsthand how automation and technology replaced hundreds—maybe thousands—of jobs through the systems I helped build.
While efficiency improved, many middle-class, non-STEM jobs disappeared. This has widened the gap between high-skill STEM roles and others, contributing to the economic divide we see today.
AI will do it faster!!!
We’ve discussed the book Mastering AI in other HD articles. I highly recommend it as the discussions address job elimination. The book’s suggestions regarding AI-based improvements in education could result in revolutionary leaps forward.
Adam: As usual, a great article from you!
One thing I predict is that AI will make good financial advisors even better financial advisors.
I spent the last 15 years of my financial services career as an academic, teaching, training, and educating financial services professionals. Over those 15 years I came to understand that the real value provided by a financial advisor is to stopping people from acting stupid, when they get scared and confused by the market’s volatility. That’s it! Everything else advisors do can be done yourself, if you choose to devote the time and effort to learning how. Fortunately, for advisors, the vast majority of people will not take the time, nor make the effort.
Yesterday I happened to “discovered” claude.ai and I put it to the test. I discovered quickly that value of the answers you get from AI are directly related to the quality and clarity of the wording of the questions you ask it.
I asked claude.ai a series of questions regarding my taxes for 2025 and got interesting and accurate results. I had to correct it, by restating that my spouse and I were over age 65 and filing jointly, to have the standard deduction corrected to reflect the additional deductions available to those over 65. Once that was done, the results were updated to reflect the correct numbers. I then asked it to consider whether I should consider making a Roth Conversion in 2025, and ifs, in what amount. Again, after clarifying that my annuity income is 72% income tax free, the resulting answer correctly stated the tax implications of the Roth Conversion, and the benefits of doing it.
I then moved on to a series of questions related to whether or not I should consider an investment in MYGAs vs. remaining invested in the market. Again, after clarifying a number of points as I went along, I received a response that, had I been grading it as a response from a CFP candidate, on an exam, I would have given them an A! It was well organized, well reasoned, and provided pros, cons, and suggested alternatives to consider, based on whether Income was needed currently, or not.
All in all, I foresee AI being a valuable tool to those who are good at their jobs, and using AI to make them even more productive and valuable to their clients.
Kevin, thanks for the tip on Claude.ai. I look forward to trying it.
Adam, your articles are excellent, keep them coming. You have a knack for bringing up real info with great background ideas. I am also in the camp that AI will not greatly leave us with high unemployment. Most likely the best way to prepare is to have a better education, regardless of your occupation. Seems to me these changes happened before and will continue to happen and it will all work out for society.
Electronic switching displaced a lot of “plug a cable in to make the connection” phone operators. Imagine waiting for a connection to be made by cable for every smartphone call today.
To me, Adam, history suggests it will LIKELY go the other way. And although implied in your article let me state that just about every technoloical breakthrough increases our standard of living…often dramtaically. At least from a jobs point; I’m not a bit worried about the AI impact.
ATM’s were supposed to dramatically reduce the number of tellers banks would employ.
Surprise. Surprise! SURPRISE!!
There are now MORE bank teller, and in some locations, more bank/financial institutions than ever before.
I wanted to know more about this so I did a Google search. The AI summary said this:
It also linked to a Bureau of Labor Statistics article that forecasts a projected to decline 15 percent from 2023 to 2033.
https://www.bls.gov/ooh/office-and-administrative-support/tellers.htm#tab-6
As with all other tech inflections, Amara’s law will likely apply: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”
Thanks Adam for an interesting article. I was in IT for 44 years and heard lots of predictions about how IT would displace workers. Some of it happened, but as you indicated, it often results in new jobs along with great efficiency.
I go to doctors a lot more nowadays. Doctors bring their laptops into the room, review data and enter data through the course of our appointment. Years ago, I had a doctor who would dictate after our meeting and his staff would type it up. I wonder if this should still be the best way to utilize the valuable time and attention doctors have with patients? I believe this is also true of other professions where we are using high priced people to enter data rather than have lower cost staff do it. In addition, I wonder if this may sometimes distract the doctor from his primary job of diagnosing patients.
I don’t know. When I type something it helps me better focus on the information I’m receiving. But I have ADHD.
My son the doctor now has an AI support that listens to his dialogue with a patient through an appointment and delivers a fairly accurate note that summarizes the actions and plans and needs minimal editing!……Saves him a lot of time
In the last decade, my wife and I have seen orthos for injuries, new knees, hips and now, for me, a new shoulder. We see differ docs in the same clinic, and they are accompanied by a “scribe,” who records the visit. I assume the doc reviews and signs off on the trascription at some point. CNAs and techs interview, take vitals and enter into the computer. The doc goes from patiet to patient with scribe in tow.
I think I am a believer in our capitalist economy from a macroeconomic perspective. So, I am not too worried about the effects of AI on total employment numbers. However, as these changes ripple through the economy, there will be micro-economic dislocations. Some people who were making 6 figure salaries writing programming code will have to transition to other occupations often at lower salaries. Coders are smart though, and perhaps most will figure out a personal solution. There are other groups of workers who might have more difficulty. There are 3.6 million Americans, for example, employed as commercial truck drivers. Should a true AI truck driving program come along, I’m sure that trucking companies would be happy to rid themselves of their human drivers. What advice could we give to these drivers when they come to their exit interviews?
So, there will be disruptions. Like folks who used to work at furniture factories in N. Carolina whose jobs were sent to China and other foreign countries, there will be economic casualties.
The description of what happened in law firms is apt. But a difference is that Westlaw and Lexis have the correct answer. AI makes stuff up.
In the mid-1980s my dad found that he could compete with and beat large law firms by computerizing his office. The relevant part for this discussion is that even using a 300 baud modem, his connection to Nexis/Lexis meant that he could do complex searchs, get all the results, and then search through, sort, and synthesize these results himself. He saw all the data. This led to winning arguments. Large law firms, on the other hand, had fleets of paralegals doing research. They sometimes missed key information. They were siloed from each other. The attorneys at the next step didn’t see everything. Nexis/Lexis was expensive, but only when it was being used. Paralegals needed to be paid every day, and new ones trained when hiring, and were a pain to deal with as people. He trimmed his staff.
AI is still more like paralegals than a real database. This week my institution implemented a chatbot for my resource (I provide services for research scientists). Testing it yesterday, I found that it is impressive. But it’s also wrong too often to be truly useful. I asked is where a key piece of equipment is located, and it answered the wrong building. I pointed out it was wrong, and it pushed back that I was wrong! Then I asked it a key detail about the use of electronic gain with new photon counting detectors, and it gave wrong answers. The correct information is on the internet, including step-by-step instructions on my website, but the AI got it wrong. It was correct for older technology tubes, but we’ve been using these newer detectors for more than ten years and it has been difficult training humans to use them right. This AI answer would only derail the mission.
At the same time, we desperately need to hire a new person. We found a real human expert, he even understands the example I just gave, who wants to work with us and who is younger, so we would be training the next generation, but we have to cut our budget due to decimation of science by the feds, so no new staff.
So, yes, AIs could be contributing to unemployment. And misinformation.
Hallucinations are a big weakness with AI. What good is it if you cannot trust it and have to back track on everything it says?
My wife had a cat scan 2 weeks ago on a Friday. While waiting on the doctor’s response, we used AI to interpret the results. They were interesting and more detailed that the doctor’s eventual response, but we trusted the doctor’s response. The value of AI was to enable us to ask specific questions.
We used to have a small army of men who delivered blocks of ice to people’s back doors. When the refrigerator became affordable the ice delivery industry disappeared. Nobody misses them now. I suppose most of the workers found other jobs, maybe some took to delivering refrigerators. This is part of capitalism. We just have to adjust.
The same with the milkman. He used to supply our home with milk and related products in an insulated container. He delivered in the early morning and we had fresh, cold milk for our cereal a few hours later.
Adam, as always thank you for your thoughtful perspective. So helpful to be able to step back amid the onslaught of hype about AI.
It doesn’t help when bosses seem to be exerting a power dynamic to employees: AI is coming for all jobs; ‘everyone’ is at risk. It also doesn’t help when each week we learn more about implications for AI deliverables (this link describes how well AI tools are advancing to synthesize data about oneself based on searches- talk about surveillance or use for profiles for credit or jobs or …whatever):
What LLMs Know About Their Users – Schneier on Security
The swirl of uncertainty around AI adds to generalized anxiety about job security and hence financial security/planning. Fixed costs – rents, mortgages, homeowners insurance, car prices/insurance, day care – have escalated for many, including my three children in their 20s/30s. Emergency funds will be more critical than ever along with investing in low-cost index funds. And perhaps more support at critical moments from retired parents, if/when they can provide help.
Adam, thanks for your thoughtful piece. I too am optimistic that as a society we will adapt over time, as we have always done. As you mentioned, many other working roles have disappeared over time, but these have not caused mass unemployment. The economy, and society more generally, have adapted.
I think that the current discussions around AI are possibly portrayed a little differently because those writing about AI may be affected by AI. Lots of blue collar workers were affected by robots in manufacturing plants. But the journalist reporting on that was very unlikely to directly affected. Now that AI may affect the journalist or blogger directly, I wonder if the urgency level in the reporting is being cranked up a little?
I’ve been a programmer for 50+ years, now 70 and still working. Finding a way to get rid of programmers has been the wet dream of businesses since I started. AI may make programmers more efficient and thereby reduce the need for a small percentage, but if Zuckerberg thinks he’s going to replace midlevel programmers by the end of the year, he’s talking about pretty stupid programmers. Writing code is just one small element of programming. The biggest element is thinking, something that AI does not do.
Spot on. Design work and debugging won’t be easily replaced. But it will make them more efficient.
I’m not a great coder, but what I have that AI doesn’t is that I understand the nuances of the images that I’m analyzing. AI has brought us new impressive tools to classify pixels (the older jargon is to segment features), but it doesn’t understand the images. It would have to pull metadata from the raw files (which would be easy to program for AI, but hasn’t been done yet) and would need to know about the biology downstream (good luck getting researchers to link to their electronic notebooks or tag the images). This could be done, and cross-modality omics could drive science way ahead, but this remains science fiction and will for years to come. At least in this corner of the world, we’re still in control of AI, but this dynamic may not last forever. In real science, someday human-AI may become a collaboration.
I enjoyed reading your perspective on this great and informative article about AI’s impact. I share your optimism. Historically, technological advancements, from typewriters to factory robots, have reshaped the workforce rather than causing widespread, sustained unemployment. They often create new industries and roles we can’t even imagine yet.
The idea that AI will enable us to tackle problems previously unfeasible is particularly compelling. I’m excited to see how these developments plan out over the next ten years, and we live in exciting times. This sentiment is echoed by the old Chinese proverb, “May you live in interesting times,” which, despite its apparent blessing, often implies periods of upheaval and challenge. However, I believe we can navigate these changes positively.