Is the generalist returning?

It would be supremely ironic if the advance of the knowledge economy had the effect of devaluing knowledge. But that’s what I heard, recurrently, while reporting this story…If that’s the case, I asked John Sullivan, a prominent Silicon Valley talent adviser, why should anyone take the time to master anything at all? ‘You shouldn’t!’ he replied.

That is from a new Atlantic piece by Jerry Useem.  In essence, the division of labor may be running in reverse in some endeavors.  In Adam Smith’s argument, division of labor and specialization increase with the size of the market.  But say a mix of Moore’s Law and globalization means that software (output and operations) expands rapidly, yet companies seek to shed labor costs due to competition.  At the margin the new demand might be for generalists, who can step in whenever unforeseen problems arise which need fixing.  Or in other words, you may not wish to specialize with your truly scarce factor, namely labor.  In contrast, in Smith’s time, demographics were favorable and labor was pouring into cities from the countryside.

As for the Navy:

The LCS was the first class of Navy ship that, because of technological change and the high cost of personnel, turned away from specialists in favor of “hybrid sailors” who have the ability to acquire skills rapidly. It was designed to operate with a mere 40 souls on board—one-fifth the number aboard comparably sized “legacy” ships and a far cry from the 350 aboard a World War II destroyer. The small size of the crew means that each sailor must be like the ship itself: a jack of many trades and not, as 240 years of tradition have prescribed, a master of just one.


Minimal manning—and with it, the replacement of specialized workers with problem-solving generalists—isn’t a particularly nautical concept. Indeed, it will sound familiar to anyone in an organization who’s been asked to “do more with less”—which, these days, seems to be just about everyone. Ten years from now, the Deloitte consultant Erica Volini projects, 70 to 90 percent of workers will be in so-called hybrid jobs or superjobs—that is, positions combining tasks once performed by people in two or more traditional roles.

Via the excellent Kevin Lewis.


One of the reasons to doubt that machines will replace all human workers is that machines can no more have a comparative advantage in everything than humans can. Comparative advantage is a *relative* concept. A machine cannot be better than itself at A than B and simultaneously be better than itself at B than A. The rejoinder is that, sure, given enough time to adjust, a displaced human worker can re-tool for a different job, but the adjustment period is painful and may be practically impossible for older workers. This article would appear to argue against that objection (although the author may not necessarily recognize that). If workers are more generalist, then they face less risk of being locked in to a specialty that becomes obsolete.

It's fun to watch people struggle to explain why miracle technologies that produce stuff for us is actually a bad thing.

If machines are taking over the world, then that doesn't mean we need to be generalists but that we should be specializing in making machines.


Some genuinely superb info, Glad I found this.

"If machines are taking over the world, then that doesn't mean we need to be generalists but that we should be specializing in making machines."

No, the machines will be specialized in making more machines. The operators role will be as a generalist dealing with areas the machines fail at and as an overall director.

Not making machines which is predictable and therefore subject to automation, but in fixing machines, or the subassemblies if the machine is pull and replace modules. Even then, you need someone to figure out which module and whether it is the module or the wires/comms or the sensor or the ground (since corrosion is a dynamic resistor in a system)

"It's fun to watch people struggle to explain why miracle technologies that produce stuff for us is actually a bad thing."

There's no need to explain it, it's enough to observe that it is there. Which it is.

Where does system's thinkining fit into this trade-off between generalist and expertise.

One of the main arguments made by the author is that, increasingly, expertise is being devalued- with a preference towards individuals who are quick to learn, and on the fly. Why bother to go deep when you barely know what new roles are going to be created in five years time?

But is there a case to be made for developing strong expertise in a particular domain, and then constantly relearning new skills within a particular sector. One of the first industries that comes to mind is that of programming- developing a core skill set ( of computational thinking), and rapidly learning new programming languages as they are created ( ruby on rails, the web-development software, was created in the late 2010s).

Is there a case to be made for developing core skills in a particular domain- and then learning and constantly relearning technical expertise? Or do you think that this trade-off is more particular to certain industries?

This is nitpicking, but your choice of example is odd. Ruby on Rails was created ages ago... waaaay back in the mid 2000s... and is pretty much obsolete nowadays.

But is there a case to be made for developing strong expertise in a particular domain, and then constantly relearning new skills within a particular sector. One of the first industries that comes to mind is that of programming-

Thanks for the tip boss. I'll give up my personal life so that I can learn the next gen web development stack. That way you won't have to fire me so you can hire someone who learned the latest stack in school (you can wait til I'm 38 and fire me for lack of cultural fit instead).

Managers and marketing people are of course quintessentially generalist.

That isn't being a "generalist", that's a "sequential expert".

A friend of mine was unable to get their PhD in climate science and is now having difficulty getting a job. He thinks a lot of people that pursue scientific studies in depth will share his fate. It’s easy to spend years becoming more and more specialized in esoteric or narrow studies. He is now trying to figure out how to use some of his skills in other fields but it’s an uphill struggle.

Is it too late for your friend to get a Ph.D. in a real science?

(There are Ph.D. programs in climate science where little physics, chemistry or biology is required.)

Such as? Where?

Look at Colombia University.

Columbia University

I don't know I took a brief look at it, seems like you need to have taken undergraduate chemistry courses to get in. The graduate courses seem to include a good deal of chemistry and physics. Not sure if you meant something else.

Those graduate courses are special "climate courses" within the Earth and Environment department. The Ph.D. requires Chem I and II, physics I and II and Calc I and II. That is almost nothing for entering a science Ph.D. program.

That sounds like first year math, physics, and chemistry - which was required of everyone taking any engineering or science track where I was an undergrad. Not nearly enough to get any undergraduate engineering or science degree, let alone get into grad school.

I first noticed this when a climate scientist with a Ph.D. from Columbia University, Radley Horton, was interviewed on PBS' The News Hour last fall and saw his undergrad degree was in Environmental Policy Studies and his Ph.D. was in Earth and Environmental Sciences from Columbia. He went against IPCC conclusions three times so wondered what his Ph.D. required. He teaches in a "Development" department, not a science department.

"He went against IPCC conclusions" The actual science parts, or the summaries? They often conflict.

The science parts, which is what matters to a climate scientist, right?

Although the key courses for being an engineer are definetly 400-level Cuckold Courses where you learn how to be a knob-polishing Beta-Cuck. You can't be a real engineer like me without extensive bimbofication training.

Climate science has more math and science in it than economics.

Not even close to true with respect to math.

Econ has a lot of math to no purpose other than to obfuscate and make their simple, common-sense, conclusions look as if they're something worth publishing in an academic journal.

As if you know.

To enter a good Ph.D. program in economics you need Calc I,II and III, Differential Equations, Econometrics/Statistics, Linear Algebra and Real Analysis. Seven or eight courses as opposed to two for some climate science Ph.D. programs

No, I didn't say they don't use math in economics. They have quite a bit of math in Economics PhD programs. I said the math is used largely for obscurantist purposes.

Basically everything in economics is either untrue or rather trivial but the mathematics is used to obscure this and make the simple look complicate and hide the details of how results are deduced from lay people.

To model the weather, you also need a ton of calculus, differential equations, linear algebra, dynamical systems, chaos theory, etc. I like you compare "good" programs in econ vs "some" climate science programs. Nice hedge. Also GP mentioned both math and science, but you only criticized the math but avoided the science. In other words, by omission, you tacitly admit climate science has more science than economics. Hah!

To cite with evidence, here's UCLA's program (note the math requirements):

MATH 134: Linear and Nonlinear Systems of Differential Equations (4)
MATH 135: Ordinary Differential Equations (4)
MATH 136: Partial Differential Equations (4)
STATS 101A: Introduction to Data Analysis and Regression (4)
STATS 102A: Introduction to Computational Statistics with R (4)
STATS 170: Introduction to Time-Series Analysis;
STATS M171: Introduction to Spatial Statistics. (Same as Geography M171) (4)
STATS C173: Applied Geostatistics (4)

At the end of the day, climate science is a real science. Economics is fake science for fake news.

That is for undergrad. At UCLA where Sam Harris got his Ph.D. in "neuroscience" it is an interdisciplinary program with little science requirements but the undergrad major in neuroscience requires much more science and math.

At Columbia, they have set up an "Earth and Environmental Science" Ph.D. with little science and math requirements so that activists can also be "climate scientists."

By the way, only a very small percent of climate scientists try to model the climate.

Oh, that doesn't mean there aren't climate scientists out of Columbia who have a strong math and science background as well. Judith Curry has pointed out that in the past decade or more, many more students in climate science have been taking what she considers the much easier path that requires much less science.

They can never even get their stats right, never mind real math.

And similar predictive power. Complex systems are a bitch.

It's Todd, and his daily flight of climate fantasy!

A textbook case of how to abuse a comments section, rejecting human consensus, rather than informing on it:

Be the Information Age, not Mis- or Dis-

Were you a lit. major or a history major?

You should warn your friend that he should choose his pronouns forthwith: He can have "they" or "he" but not both, as of today anyway.

Ph.Ds have always been risky unless you had an in at a university or you caught the front wave of the trend. Way back in the late '70s when I was entering college, the news was filled with astrophysics Ph.Ds washing dishes since the space program was on the wane. Nuclear physics Ph.Ds had similar issues later as the Cold War ended and nuclear power waned. Obviously, the underlying skillset offers a good foundation, but the specialization gets your job application shunted if you apply outside your specialty.

I guess coders better learn how to use a mop.

Coders will program the machines which I imagine pays more than mopping.

LOL, those high-tech jobs of the future: a generalist, mopping floors. As for coding, you know we've reached Peak Coding when there's a kiddie boot camp for coding that elementary school kids learn in... I myself hobby code (C# is the base language, has garbage collection of pointers, good stuff) but I'd never recommend a career in it, as you're competing against a bunch of Indian minimum wage teens who are faster than you are. Better to be a gate keeper (owner of capital, in me and my family's case over $10M in net worth). Biggest challenge I have is simply on occasion having to take an ad in Craigslist and find a new tenant--sigh, such a hassle--takes 2-8 weeks depending on the property and time of year.

For you purists--and I knew this when I typed it but was too lazy--C# technically does not use pointers that much but reference types, that's what's garbage collected.

I know the guy that developed C#. Bright guy, and very nice too. Not a hint of arrogance or ego.

We have kiddie coding camps, some open only to girls. While I would not be able to say why, seeing on my street the "bandit signs" advertising them gives me the distinct feeling that those parents who sign their kids up are probably behind the eight ball.

Last year, among all the various signs, a new one popped up: handwriting camp. That one made my heart sink for about a dozen reasons. Maybe it's how you build a generalist, one wretched "camp" at a time.

This is well known in software practice. At its worst, the industry had a whole lot of specialization: Teams of 10+ people, with 6+ roles, to write run of the mill business software. Once you added infrastructure teams, management and all the matrix-shaped overhead that companies loved to have, we are talking dozens of roles and hundreds of people to do relatively simple things. Communication costs grow exponentially and kill productivity.

Top software organizations have shrunk the list of job descriptions, but everyone gets to do more. This makes hiring far harder, but we are better off with mostly generalists than specialists. Cutting people cuts communication lines that have to be maintained, making teams more efficient, and as a friend of yours keeps telling everyone, efficiency is leverage.

Instead of hiring specialists, nowadays we hire generalists that have a bit more experience in the areas where we might be lacking expertise, but chances are that part of their role will be to transfer this experience to generalist with less experience in the subject. This is also why job changes are great for learning in the industry: Chances are that you will be able to pick up new tricks from people who have different experience than anyone you worked with at the previous job. This wouldn't happen if we had very well defined roles with clear practices.

This sounds very familiar. I run a company of 20, around half of whom either build or deploy what we make (hardware-enabled software solutions). At our size we can't afford to hire any complete specialists in anything, or at least we can't afford to fully utilize that specialization so they'll get paid as a generalist who's missing a few skills. So we hire generalists with a balanced range of experiences and they figure out how to solve problems and build things. It's not cutting-edge R&D stuff, but it solves the customers' problems, the company gets paid, and we can then pay our scrappy generalists. Most of them like working this way - if they don't, they don't stay long. But you can tell they start to get restless if they end up doing the same shit too many times in a row. Luckily we're constantly forced to try new things (some might say overstretch ourselves...) to win business and stay ahead of the competition, so there's always some new headache waiting just around the corner for these guys/gals.

The danger of not doing cutting edge stuff and instead doing average stuff is that you now have more potential competitors. Companies will bump you for cheaper workers from Romania. As the other Tyler has warned, the average is over.

When I started hiring (mostly software) engineers decades ago, one of the standard questions was something along the lines of "will this person be an asset in 5 years", i.e. are they going to be valuable after the current project is done. Because the next project might be quite different. We always hired with the hope that the new hire would be a career employee.

Among other things, at least in that sort of environment, this tends to means IQ gets weighed more heavily than specific skills.

A really strong focus on specific skills, a guy who thinks of himself as "I'm an expert in X" can be a problem. One guy was like this, in a "can't see the forest for the trees" way - this guy was an expert in bark, and thought about himself that way. Very, very smart, but if your problem wasn't bark, not very useful.

"When I started hiring (mostly software) engineers decades ago,one of the standard questions was something along the lines of "will this person be an asset in 5 years"."

Would most new software employees now stay 5 years?

For the engineering heavy type of software this firm does (i.e. not relatively generic business process stuff), I'd say yes, although I've been away from it for some years. Domain knowledge is valuable, and takes time to develop; you don't want that investment to walk out the door.

They hired almost entirely MS and PhD people; for the software they sell to clients, that's the level of people who sit on the help desk, because that's the level of the clients.

Sounds interesting already. What field of engineering.

If you avoid CRUD, mobile, web, or any kind of glue coding then software development becomes real engineering. CAD, EDA, medical devices, avionics, high frequency trading, compilers, operating systems, or heck, even the innards of a web browser all require deep domain knowledge. You can't bootcamp this stuff.

Most engineers leave after 2-3 years because raises never keep up with the employee's market value.

As in the natural world, whether it is better to specialize or generalize depends a lot of how fast the environment is changing (or the range of environments you have access to). Or in the case of Mixed Martial Arts, who your opponents are going to be (e.g., what their skill-sets are and how good they are at deploying them).

The article in the Atlantic is actually pretty funny. Useem is rightly skeptical about the Navy's insistence on staffing a ship with a handful of John Von Neumann-like enlisted personnel who can master every job. It sounds like Donald Rumsfeld was reading Heinlein's rant about "A human being should be able to change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, cooperate, act alone, solve equations, analyze a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly. Specialization is for insects" and decided that's what the Navy needs.

In a related development, Navy ships keep running into one another.

But their affirmative action training was up to date. Priorities.

Its interesting - and disheartening - to read the accident investigation reports on those incidents. Major personal selection and training problems. Significant operational overload (too few ships for the accepted mission footprint).

We haven't had a real naval war in 75 years. It's hard for the organization to remember what's actually important, and the officers who do don't tend to get promoted.

"set a bone"? How about use a bone saw to hack up murdered journalists for corrupt and authoritarian crown princes who buy off US presidential sons-in-law?

Generalists don't get paid. Specialists do. This applies to medicine, law, engineering, any professional endeavor really. Specialization is for insects they say. Sure, insects are evolution's most successful phylum while others with no niche become extinct.

Also, jobs require credentials (diplomas, licenses). Those credentials are often entirely superfluous but the trend is only accelerating. Specialists have credentials. But how to evaluate the skill set of a generalist?

Most likely they will simply be forced to get multiple semi-useless credentials in various specializations...

Credentials are the easiest but flawed way to evaluate skill. Short of that, the internet has Yelp and Google so you can see what other people think. The funny thing is that dentists, doctors, and lawyers are trying to get bad reviews off the internet. Those professions get paid big money but can't handle a little criticism.

Bingo. as a generalist consultant, I have learned to hate this.

Because the main underlying implication is that a specialist has to pretend to be an expert, whereby a generalists is based on the premise that they do not know what they are about to do, but promise they will be good at it anyway.

Of course, we soon learn that most specialists are faking-it-till-they-make-it, and following one-size-fits-all templates that they learned at some consulting mill or weekend training course.


Except that they often are the ones who get promoted to running the whole show, where they get paid more than anybody else. Sorry, but you are wrong, ooooops!

Or they promote a specialist right above his peter principle level

Who then gets replaced by a more competent generalist when this gets realized.

There's no clear cut winner between specialists and generalists when it comes to business. But when it comes to a specific task, like you need someone to perform surgery, or properly write out a legal contract, you bet your ass specialists win bigly.

This is precisely why repatriating manufacturing back from China will be far more difficult than people realize.

Revived domestic manufacturing will supposedly create jobs for unskilled workers left behind in the newfangled knowledge economy. Maybe they'll use a wrench to tighten the same nut in the same spot on a thousand identical widgets traveling down a Henry-Ford-style assembly line... except of course robots already do such repetitive identical tasks much better.

Anything that's left over — anything that robots can't do cost-effectively, which is actually a lot of things, as Elon Musk found out to his chagrin — requires problem-solving generalists (who nevertheless still need a high tolerance for a lot of unavoidable drudgery). Things like computer motherboards, for instance, get made in small batches with endless customization options.

With what starting materials could you create such a workforce in the USA? Not from the deplorable bitter-clingers, nor from the lumpen-intelligentsia churned out by the tuition-industrial complex. China wins the trade war.

unskilled workers left behind in the newfangled knowledge economy.

Apparently the Marginal Revolution commentariat views software development as the occupational field of the majority of contemporary Americans and the pinnacle of the skilled trades. Oddly, software developers don't bake their own bread; butcher their own hogs and smoke their own bacon; design, assemble and repair their own cars; install and maintain the plumbing in their homes; or brew the expensive fruit-flavored beer that they've come to love. A certain number, actually a lot, of "unskilled" specialists will always be needed to service the genius class. The business world, however, dislikes paying a baker just to bake bread and wants that employee to additionally shovel the snow off the parking lot and make sure the drains are open. One of the main problems with union labor contracts is specialization of duties. Flight attendants, for instance, should be willing to clean aircraft cabins and the pilots should fuel the plane. It costs to much to have these specialists sitting around in the boarding area watching television.

I'd rather be loathe to accuse software developers of being the genius class.

Robots can come to China too, hard to keep them away in fact.

My point was that there are a lot of things robots can't make. Robots are specialists and a lot of small-batch custom manufacturing still needs human generalists.

If AI ever gives us robot generalists, maybe that'll be a different story.

The benefits of specialized divisions of labour exist in relation to total inputs and total outputs that go into final product. Does specialization contribute to more output, as in mass production? Or does it mostly ratchet up input instead?

The LCS is a strange example to cite - its a no-ruddy-use ship, a good-for-nothing ship, a silly-bloody-idea ship. Mind you all naval vessels that are not submarines will perish almost instantly in a war against a non-negligible opponent.

Most merchant ships are (and always were) low manned. Warships needed large crews to be able to continue to function with casualties. US Nuclear submarines have larger crews than UK subs due to different missions. The USN always wanted every crew member to be able to do multiple jobs in case of emergencies. Automation also has reduced the need for many crew members in all types of vessels. Remember stokers used to be a very important (and skilled) part of any ships crew. There were several hundred on the Titanic.

Interesting question with high stakes for students - should they go deep in one area or not? I'd advise that unless the job legally requires a certified human to do it (or sign off on the work done) which serves as a barrier to entry (medical, engineering, chartered accountant for example) it's very risky.

The specialists who should be skating on thin ice are the unnecessary pharmacists, who often make six figures for taking filled bottles of medication off the shelf and handing it to a customer. Why doesn't the MR Commentariat ever bring up the extra expense these people bring to the medical labyrinth?

it will sound familiar to anyone in an organization who’s been asked to “do more with less”—which, these days, seems to be just about everyone.

So when were the days when private sector workers were not asked to do more with less?

Hmm, 40 generalists on board.

And no mention of crashes or collisions on this page.

Maybe wait until Navy "self-driving" gets better, or add a few more watch crew. Bonus, they only need to work binoculars. Maybe an intercom.

How surprising, Tyler pushing the return/rise of polymaths.

The article is moderately interesting, but isn’t it also just the latest retread of the “you’ll have 10 careers” stuff we’ve been hearing for the last 20+ years? There’s no such thing as a free lunch, and that includes specialization. The tradeoff is that specialized people and systems may be more vulnerable to change, but that doesn’t negate the advantages. Just as with everything, you have to make bets on which investments will pay off for the long run.

In my manufacturing business, we value flexibility over specialty. Nearly everyone in Production is trained to be proficient in every task, and many staff members from one department learn tasks from the others. The results are happier workers and better productivity.

Comments for this post are closed