Jobs Involving Routine Tasks Aren't Growing
stlouisfed.org
stlouisfed.org
Many backend jobs involve routine work from people who could be doing non-routine work if the CRUD routine job was automated. Within an industry or a profession it seems like there could be a mix of routine and non-routine work.
Similarly, how many people's jobs have changed from being routine jobs to non-routine jobs through either retraining or automation allowing them to take on higher order work? Again, doesn't say.
Also, interesting to note that the actual unemployment rate for routine congnitive, routine manual etc 2009 shock aside doesn't actually look anomalous compared to much of the last two decades of the graph is to be believed.
Talking in terms of "automation of routine tasks" sounds reasonably descriptive of how the spread of technology works. But, technology is a concept that is notoriously elusive in economics. That's because it's hard to clearly define in concrete terms useable in the context of economic theories, whether they're mathematical conceptual.
For example, we've always been imagining "robots" as tin humans that do stuff people where doing. You have science fiction movies, books and such being written right now with this imagery, just like the 1950s sci fi art, Jetsons. Just like the mechanical turk and automatons of the 1700s.
I think robots are a useful mental placeholder. "Technology will be doing task X." But in reality, technology is usually more like "tools." Imagine a mechanic in the future. Maybe the cars come in with better self diagnosis before he sees them. The parts he needs are already known so he has them ready ahead of time. An AR (or whatever) info delivery thingy tells him exactly how to install or remove parts. etc.
What you have is a more useful mechanic. As long as a mechanic is still involved, I think "tool" is a better description. If people are no longer involved, "robot" seems a bit better. Ahead of time, when you are trying to imagine where technology is going it is very hard to discern tools from robots. Is a lawnmwer a robot? Is a a self guiding scalpel a tool?
That's not true. The word "technology" is quite precisely defined in economics, and the definition is generally agreed upon by economists. This definition does indeed lend itself to useful economic theories.
The challenge is that the economic term "technology" does not map 1-1 with the colloquial use of the word "technology", which leads to an abundance of confusion for people unaware of this distinction when they interpret economic analyses.
The endgame here is that I don't need a mechanic.
But endgame, where it's "robots" as we all imagine it is always a moving target. Mechanics weren't needed 200 years ago either. So far, we've needed new professions. Maybe that will end.
My point was just that automation is a fairly fuzzy way of explaining what is going in.
Once you have exact replication steps, automation is easy.
Is the distinction between routine and nonroutine manual occupations simply technical vs service?
> Nonroutine manual occupations, which include service occupations related to assisting or caring for others
> Routine manual, which include construction, transportation, production and repair occupations
For example, why is driving a truck considered more routine than being a nursing assistant or phlebotomist?
Interesting, could you please expand ?
A bookkeeper/accountant might be manually entering invoices into a system, emailing managers to approve payments and such. It takes 3 hours per day. This work is "routine" in the relevant sense (it can be automated) but our young, congenial and soon-to-be-replaced-by-robots accounts payable woman (her name's Lynne) also does and is capable of doing "non-routine" tasks. She can call up suppliers to negotiate payment terms, have lunch with the manager in the next office to clandestinely suss out some questionable expense charges, etc.
Calling Lynne a "routine worker" is kind of problematic. She does routine work because it needs doing. When that work disappears, she's not out of work by default. Generally when technology creates efficiencies, the unemployment rate does not just increase by the number of people that did that work yesterday. People who are capable of doing other useful things do those things, if it's useful to do more of those things. There's still a question of whether the economy can use up these newly available Lynne hours. Often it can.
What happens when some technology automates a lot of "routine" programming work? Programmers make more program(s).
Perhaps not, but it does increase. Because somebody else was calling suppliers and having lunch with the manager yesterday.
And here's the magical thing! Once all of bucket 1 is automated, a lot of my bucket 2 tasks start becoming routine--patterns emerge and I can imagine ways to even chip away at them with automation, allowing me to do even more high-value work, in theory making me more valuable to my employer.
Some technology advances seem to reduce unemployment. When factories were doubling efficiency several times in an average career-span, employment in manufacturing kept going up and so did pay. This was a long period. It started with farmworkers (replaced by automation) urbanising to work in the rapidly advancing, high tech manufacturing sector. The factories kept automating more and more. Efficiency went up. During long periods labour got a good cut of these gains, salaries rose.
At other times and other technological events (including trade which is economically very similar to technology) unemployment went down in the sectors closely linked to the tech.
Maybe it would be doable at a lower salary. Let's pay Lynne half as much as yesterday for this new job.
This immediate nitpick aside - I think it's great that routine task jobs are going away. The bulk of those are mind-numbing, dead-end jobs - do we really need to keep subjecting people to 40h/wk routine boredom?
They are so kind, and helpful, both to residents and their families, and they are just so superb at what are really difficult jobs. They deserve much more than most of them are making. They're some of the higher-paying jobs in the town, but relatively speaking, they're not high-paying.
Back when I was in middle and high school, my other grandmother was in the same home, though not in the Alzheimer's wing, and they seemed to take really good care of her, too. I wasn't as cognizant of it at the time, but talking to my parents and looking back, I can see that it's the case.
I know that I couldn't do their job.
What happens if "crappy jobs" are not available anymore and all the available jobs require 1) non trivial credentials and 2) enough mental energy that after work you are "done" and can't do anything else? And what happens to people that are only capable of dealing with a routine task job?
Until there is something like universal income routine jobs seem to be the lesser evil, because the odds of universal income coming into being in our lifetimes seem pretty low, and having something to fall back on to pay the bills while you figure out what to do with your life can be quite helpful.
Sure, it's great that we are gaining this capability, and the people working towards it should be lauded for their progress. The issue is that many people rely on the ability to exchange 40h/wk of routine boredom for the basic necessities of living.
Our society assumes that most people will contribute enough value from their "work" that they can exchange for the things they need in life. If their work loses all value, then we need to figure out what else we can do for/with these people.
At this point, it seems as if our engineering ability is advancing beyond the pace of our social structures and legal system.
Global warfare would solve this problem on all fronts. You know, if the prospect of war these days didn't stand a good chance of ending the species.
Very very scary if tech that most of us still haven't touched is also in a sense routine.
I hope I'm misquoting him. But my takeaway is that between manually coding processes that could be solved with ML & doing infrastructure profiling, most devs are spending half their time fixing problems that are 'routine' at the big three.
Isn't that more properly regarded as exciting than scary? The long term benefits to automation have rarely failed to outweigh the temporary costs.
I would personally be very happy if my own current job- software developer- were somehow automated out of existence. Not only would it allow me to put my money where my mouth is regarding praising automation but it would mean huge gains for the human race.
This entirely depends on the model the human race is running off of. With the current state of mass-centralization in tech (see, again, the big 3), you will never get to experience such gains. In fact, you would quickly grow dependent on someone else to feed and shelter you, considering you can no longer provide a benefit to society that puts food on the table. In fact, most people wouldn't provide value anymore.
> The long term benefits to automation have rarely failed to outweigh the temporary costs.
For some people, yes. You are correct when it comes to the numbers. GDP increases. A nation has more "wealth" to work with altogether. But the other side of the coin is that wealth disparity becomes more extreme.
Big however - If we ran off a decentralized model - arguably the way the Web was originally intended, I could see how the human race would be placed in an unprecedented position for future growth, and humanity would really thrive.
Granted, regarding the current state as ""exciting"" boils down to what you value, I guess.
When (not if) I automate myself out of existence, I'm sure as hell not telling anyone I did it.
In my experience, this is true over longer (1-2 year) timescales, even if it's not obvious how at the outset. Most people, when made redundant, find new ways to make themselves relevant. The process isn't exactly pleasant, but the outcome often results in a lot more lucrative and fulfilling career than they had before.
Put another way, if you could create a way to automate your job entirely, then there also has to be a guarantee that doing so will not result in you getting fired. Alternatively, if you automate someone else's job, then the company needs to be responsible for training them to do a new job. Even if this is not practical or possible in all cases, it needs to be the case more often. Right now, corporations pay the government to solve this problem (not voluntarily), but the government does a pretty bad job of helping. Ideally corporations interested in any social responsibility need to solve this problem themselves. When the automation revolutions becomes real enough to threaten executive jobs, I suspect they'll solve it pretty quickly.
There's nothing that says that jobs are sacred. Indeed, most folks born since 1980 believe that they'll have to switch jobs every few years to a.) stay relevant and b.) get paid what they're worth.
a) do your job, same as always b) automate your job, and either end up unemployed, or find a new job, still working xx hours/week. c) don't tell anyone you automated your job, don't work, collect the money.
If you are capable of automating your job (you have the skills, the know-how etc.) then all of these options have different moral and economic tradeoffs.
a) if you like your job, you get to keep doing it, but, by not automating it, you are maybe not doing your job as well as it could be done, and/or are costing the company money which could be reallocated. As an employee, you maybe have a responsibility to automate your job and by not doing it, you are shirking your duties (its a stretch) b)You automate your job, get a new job, and take everyone else who was doing the same job as you and automate their jobs as well. You end up ok, because you're talented enough to automate your job. the other people who weren't talented enough just get fired and have no shiny new credential. c)You are definitely dodging the obligation to your employer, but at the same time, they don't know the difference. If the work is unchanged, then you are free to use your time for an alternative economic benefit. It seems unfair, since you get to double-spend your time.
In my mind, B is the choice most people worry about. There are fewer people who can automate a job than there are people currently doing that job, and that doesn't even touch the problem that few low-skill jobs are filled by the people that can self-automate their jobs. So, more worrying is that someone invents the roomba of floor waxers, or the self cleaning toilet stall and then we don't need janitors anymore.
Not sure about that, that's not at all my experience coming from a poor area which has gone through several mass-scale outsourcing. ( i.e. a whole vertical slice of the economy just moves away )
You need an set of positive circumstances in order to make yourself relevant again. Access to appropriate education/resources, capability to identify a new opportunity, capacity to actually take it ( i.e. money, flexibility, ... ) If you miss any of that, you just enter a vicious circle that drags you down.
Over the course of a generation, what you say is true. Most of the youth has moved away from that poor area I came from, including me.
The real problem is that you don't know what is the rational decision at the time you need to take it, only in hindsight. It is not in the business owners to share their 10 year strategy with their employees. Everything is positive, all green, until the very day you have 2 weeks to pack. You don't necessarily know the big picture. Is your company having trouble or is the whole region is going to be in ruin. You can make the rational decision to start a local business and have your market vanish way before you have benefitted. Same thing with the local politician, we are always at the end of the tunnel.
Same problem with training and moving. How much do you leverage your current competence. As a DevOps, do you learn development, or plumbery ? How far do you need to move. It is even worse here as you have even less information about the target sector or region.
So what happened to people that made the right decision, they did alright of course. But there have been a lot of people that trained in using newer better technology of their field and that did not help them when the job moved to India. Some moved to Spain and France, but the crisis started a few years later there. A lot opened businesses, and failed as new businesses do, but also because the whole region went under. Same outcome for people that moved to different field in the same region, they were caught in the ripple effects. None of those decision were irrational, or necessarily easy. They were just not the right decisions when you have perfect information.
So how do these people "will" their way past illiteracy, plummeting home values (Makes it quite difficult to sell and start somewhere new), and the need to suddenly compete with all of their former co-workers? Economic data suggests these people simply become "Disassociated" from the Labor Force and are no longer counted as unemployed whether they found employment or not.
History has shown this not to be true. The labor force has evolved several times in the relatively short history of the US. Markets are able to adjust these changes pretty efficiently: people whose skills become obsolete learn new ones, and higher unemployment winds up encouraging companies to find uses for cheaper labor.
Sure, it's not in your best interest to tell your boss you found a way to automate your job, but it's definitely in everyone else's.
Considering that my own personal theory is that software developer (also my profession) is the last job that will ever be automated (there will have to be one last person that writes software to run "the robots" before "the robots" write all their software), I can't say that I share your excitement at this milestone.
But then again, maybe it's simply evolution and nothing to fear, since we are as temporary in the universe as the last species we evolved from.
Tech Talk with Jeff Dean at Campus Seoul
'deep learning for building intelligent computer systems'
Think about the last time you heard how a developer tells you that you don't need to learn the ten finger system to program well, or that he is capable of using 10-20 programming languages. How can a person think about usability of his software and a debuggable architecture, if he has to think about how to put an "i" and an "f" in the text editor? How much time is left for genius ideas when he has to look up all the time if the current language requires him to write a try-catch or a try-except block?
Typing is only a small part of programming, and touch typing isn't a huge benefit even for that part, considering all the numbers and symbols used, auto-complete for identifiers and keywords, and macros for common idioms.
And it painfully overworks the right pinky.
I disagree. Being able to think through my fingers has helped a lot. It's hard to explain so let's use a driving analogy. Think back to how good your driving was when shifting wasn't muscle memory, when you had to think about the clutch every time, or when you had to think about just how much brake to apply and how much throttle in various situations. Or even when you had to think about how much to turn the wheel.
When all those basic tasks required higher-level brain function involvement. How much did you actually focus on the road? How much could you think about choosing the most optimal lane? Etc.
Even something as simple as keeping appropriate safety distance for the speed you're driving. When it becomes intuitive, you're free to think about other stuff.
Same for typing. When I think "string template" and `foo ${bar}` flies out on its own. That helps. It means I can think about more important things.
If you don't like my driving analogy. Pick up a skateboard. Or rollerblades. Or a hoverboard. And try to hold a conversation while you learn how to ride.
Let me give you another example that maybe drives the point home better: Have you seen the videos of how Minecraft was made? A very short amount of time was spent on getting the I/O to the filesystem to work, of setting up a window and filling it with colors, of creating the basic 3D world. I believe that is a huge part why not everybody can write a successful game as Minecraft. Because I don't have trained the routines, I need to spend a considerable amount of time getting all these basic things done before I can start with an actual 3D game. Thus an experienced game developer is already done with his first demo in the time I am done with the basics. And at this point he has a demo to show for, I have nothing. So he even gains more motivation in the same time to continue, and he may even gain some first user feedback. This will make him code longer (huge benefit, I give up at this point) and he will also develop more in the direction that is fun for players (huge benefit two).
It is hard to connect directly the geometrical transformation math to making a game that is more for the players. It is in fact not directly connected. But if you don't sit down to learn the "boring math" you can't get there. That's what typing faster is for. It's not yielding benefits directly. But if you learn that, learn your text editor, learn your language, learn some design concepts, you will reach a level of competence that is not reachable without. Of course you could also do other things to get the ideas from your head into the computer faster, like voice input, or flow programming with touchpads. But then you need to get into these very well to get them out of your way as well.
I think Minecraft's success is mostly luck and doesn't really prove anything. A lot of very experienced game developers write games that fail, all the time.
You're right that as a programmer, when you start working on something you're not yet familiar with then there will be a period in which things that are routine and automated for someone experienced in that area are not yet routine for you and you haven't yet automated them. I just don't see that as particularly important. You learn how to do these things, then you automate them, and you've only needed a small one-off extra effort compared to the experienced programmer doing the same thing. 80% or more of the work of writing a program isn't solving business problems directly, it's writing the tools that make it straightforward to do the actual business problem solving. Most of the skill of doing this is transferable, and learning 10-20 languages really doesn't involve that much overhead.
Sure, he did get lucky that Penny Arcade (and Kotaku, and a bunch of other sites) covered his minecart rollercoaster video to get that initial burst of attention to his project, and maybe you could argue that it was lucky that he stuck with the project's development for two years in relative obscurity for it to get developed enough to help reach its overnight success.
However, I definitely feel it was a novel enough of an experience that it would have eventually found a way to hit the mainstream and been a big hit regardless.
I agree a mountain (of trivia, routine tasks, and communication barriers between man and machine) stands between our ideas and working code, and it's essential to deal with all that in some way.
But there's more than one way to do it.
Wut?
I've been touch typing for years, and none of my fingers have experienced any strain, or ever been overworked. In fact, the reason I learned how to touch type was to _reduce_ the strain on all my fingers.
The right pinky is used for the most common symbols, and for the ENTER key. It's really bad in C/C++, with {} and [] and ; and = and _ everywhere.
Or maybe it's just different between individuals.
Programming outside the web/enterprise bubble is far less about connecting existing black boxes.
That's 99% of all programming.
There's a question of: If you know 20 programming languages, how well do you know any of them individually?
I don't think that's necessarily a fair statement, however it rings true when you talk about actual development. Library knowledge, beyond a certain point, is far more important than programming skill.
The real question is how well should you know each individual language to be productive?
Is knowing dark corners of some obscure language bears more value (in terms of productivity) than an ability to chose the most fitting language for your particular task you're doing at the moment?
And when I mentioned dozens of languages, I was talking about the DSLs primarily. A well-designed DSL does not need much learning effort at all.
> Library knowledge, beyond a certain point, is far more important than programming skill.
This must be very dependent on an industry you're in. Enterprise Java coding - maybe. HPC - partially true. Embedded - what libraries?!? Can't find any!
I know for a fact that work I have done has contributed to the demise of many positions. Could you imagine the industry for "Internet Cataloging" if we didn't have search engines? Email if we didn't have Gmail?
Technology is a job killer. That idea is something that's been a part of society and literature for hundreds of years, sometimes in violent fashion. Our economy needs to evolve in such a way that the destruction of jobs is a net positive for society.
What we should accept is that the time is approaching where a) it is possible for skill and lucky people to make many millions of times more money than those who are neither and that b) the cost communication has decreased so much that in many cases it no longer makes sense to employ people, instead it makes more sense to contract the work out.
This means we should do two things: stop talking about income differences and focus on making sure the poorest have tolerable lives and b) make it much, much, much easier to start your own small business. This means tax cuts but above all simplifications of the relevant law.
Imagine a world (or the US) where 75% of the people are at the poverty level and 25% are "rich". How long do you think that's going to last?
You can rent for less than $1000 and should be able to do utilities for ~$300.
That leaves about $200 a week, which is certainly not the lap of luxury, but it's plenty for food and necessities (I've done 2 people eating well enough for ~$100 a week so I really don't feel full of shit saying that).
There's no planning for the future in that, but part of the idea of a basic income is to de-risk things like that no?
(At least, if we means the US and we keep providing Medicare and similar)
Only as long as it takes for those 75% to make it to the pitchforks.
> This means tax cuts but above all simplifications of the relevant law.
How do you pay for the poorest to have tolerable lives with tax cuts? Or are you talking about cutting out regressive taxes on the working poor?
There's literally no evidence whatsoever for that, but evidence isn't what causes people to advocate for tax cuts for people making "many millions of times more money" than those who aren't lucky or talented.
Large companies can afford the complexity but small companies typically can't. More importantly a neighborhood printshop isn't a complex business (it doesn't, for example, have overseas income on intellectual property taxes).
I love the idealism, but that's not how it works. Small or large, you will hire a company that specializes in customs to handle all the paperwork for you. It's about as complex as paying a bill for a small business.
The same applies to accounting, taxes, legal issues, etc. all businesses outsource this work to specialists.
The problem is that all facets of business have become so complex and specialized that you now need a specialist for everything. This isn't a burden on the business, since all businesses have to pay it and handle it in the same way.
It's a burden on the employees and consumers. They pay for all this complexity through lower salaries and higher prices.
It's a quote, but I forget where from.
If your work is just following instructions, then it's probably routine.
As for me, I'd like to see most middle management go away, since I largely see it as a waste (basically, if people know how to manage themselves, you can get rid of most middle-managers).
Most middle managers end up achieving neither, but a layer of management is the default solution that companies most end up with.
By the way, what is the largest organization you can think that follows that swaps 'designated communicators' roles with no management?
>By the way, what is the largest organization you can think that follows that swaps 'designated communicators' roles with no management?
It's rare in any organization that the people who have power are the same ones who get things done.
For instance this https://en.wikipedia.org/wiki/Time-based_currency
As for "productive societies", I don't know what that will actually mean in the near future. Automation is 100% productive. It needs no people for that. These terms have meaning for industrial society where people are treated like robots. It has little meaning for society where robots produce more of the goods. And everything else is about simply organizing society so that people feel some sense of meaning and contentment.
At the end of the day in motivated people with money, and engineering society in that direction, we actually get a lot of ills like burnout, alcoholism, workaholics, etc. Yes people can be coerced by money, but often it's not for their own good. It for the good of the owners of money so they can make more profit. But at the same time, this sort of thing ruins the experience of life for the worker.
> Automation is 100% productive
Because perfect (i.e. 100%) productivity is technically unobtainable because there will always be efficiencies to be made. Yes, machines are 'always on' and will complete a task in the most efficient way we tell them to, or that they learn to do themselves, but I think that humans will always play a role in raising productivity of even fully-automated systems by making efficiency improvements.
That is for the foreseeable future anyway, until AI starts making its own exponentially big efficiency improvements ;)
The goal seems to be making the rich even richer.
You are right in that the goal is key, because much time and effort will be spent towards the goal. When Kennedy said we should go to the moon, we did. Our leaders today have other goals, and they too will reach their goals. Choose wisely.
So yes, I am looking forward to get even more richer in time. Like most of people. And this is way easier than it ever was before.
The Luddites were basically correct - they were trading "good jobs" for fundamental unemployment. Demographically speaking the families involved did not on average recover from the damage for a few generations.
So while from a global perspective the overall change may be positive over time, you can't discount people out of hand for saying "wait a minute, we're going to get screwed hard here". They may well be right. It may still be the right thing to do.
What the policy implications of such are or should be is a separable issue.
Another potentially deep issue: the industrial revolution creating a bunch of new jobs and job categories does not demonstrate that the same will be true of a putative automation revolution that we are entering....
Which is the point, really. This is exactly the sort of assertion that is often made ... and often the Luddites are trotted out as an example. But they are a better counterexample to what you assert than an example.
The problem the Luddites had is that they were skilled labour facing the possibility of destitution with no welfare state to provide time and money for retraining into new occupations. They could see that there was no future for their occupation, but they were arguing for better terms with their employers so that their short-medium term future wasn't so bleak (e.g. their family being sent to the [workhouse](https://en.wikipedia.org/wiki/Workhouse), and likely a drawn out and dehumanising death).
We have a welfare state now in all developed countries (by varying degrees), so the problems that the Luddites faced have been greatly mitigated from those past extremes. However, all welfare states assume that any state of unemployment is merely temporary, and provides on that basis.
The past is not a wholly reliable indicator for the future, and present indications are that with increasing automation there won't be enough good (interpret as you will) new jobs. Further, many of the new good jobs will be out of reach for a significant portion of our populations without the luck of latent ability and/or the luck being born into the right (well-off) family with access to high quality education and social connections.
The institutions that solved the Luddite problem can't solve a problem of truly long-term/permanent unemployment or very irregular employment.
I started my career as a journalist/editor when paper magazines were still a thing. Soon, the publishing house I was working in went out of business, so I have to learn the new trade: software development. Which also may become obsolete soon, at least for simpler tasks, so I'll be ready to learn something new.
Capitalism works because it needs rich enough customers on mass scale, the middle class. If there will be not enough jobs for the middle class to exist, there will be no capitalism anymore.
Unfortunately I can't show you what life was like before the industrial revolution, because we don't have images or movies from that time, suffice to say that life then "was short, brutish and hard". Child mortality was rampant, according to Adam Smith it was not unusual that a woman who had had 10 children didn't see any of them grew up to be adults.
There is no goal, but progress is moving as far away from that condition as humanely possible, as fast as possible. This is best achieved with capitalism.
If you want full scale employment just create another world war, so that we can draft everybody and piss all our wealth away.
The point wasn't that automation is bad (it clearly is not), but that if we continue to automate without a plan for how we deal with mass unemployment, then automation will lead to mass human suffering of a different kind.
http://www.nytimes.com/2014/04/28/business/economy/recovery-...