The US just has a pretty bifurcated economy with a high spread in earnings (inequality).
You can see that here in earnings by undergraduate major:
For most majors, the 75th percentile make almost double the 25th percentile and the difference between the 90th percentile and 75th percentile is almost as big as the difference between the 25th and 75th percentiles.
I see a few notable exceptions:
Allied health majors (nursing and PA) tend to have a high floor but low ceiling.
Engineering majors have a high ceiling but also a high floor.
When I was just starting grad school, all the hype was about massive parallelism. The âvon Neumann bottleneckâ was coming to get us and we would not be able to keep increasing computational power indefinitely. I bought into this myself, and though I never became an expert in hypercube architecture, for instance, I did write papers with an eye to keeping things parallelizable.
Some funny things happened in the meantime. First off, sequential computers continued to get a lot faster for years after the panic, though I guess things have settled down. Second, they got so cheap that there was âparallelismâ inherent in the fact that you werenât time-sharing them with other people. Third, there was a certain degree of parallelism in inexpensive multiprocessors, and with multiple threads running, there was no difficulty keeping utilization high.
Finally, when some form of parallelism is actually necessary, it often suffices to have very course-grained parallelism. Instead of splitting your problem of a billion elements into a million little processes communicating all the time, it often suffices to split it off into 100 or so chunks that can each be processed on their own and recombined later. This kind of parallelism maps well to a cluster of sequential machines with ethernet addresses. Thatâs actually what we do most of the time now and call it âparallelism.â It is not massive parallelism though and nobody should fool themselves.
There are techniques like map-reduce, which while certainly clever and effective is not really more sophisticated than massively parallel algorithms that preceded it by years. It just fits existing hardware better (which I agree is important).
So in short, hypercubes and more theoretical PRAMs (parallel random access machines) never became necessary and certainly are not ubiquitous today. I assume hypercubes still show up in supercomputing applications, but something like the CM-2 (early 90s Connection Machine) solves a problem that isnât as significant as the things done every day by cheaper less powerful processors and GPUs.
Itâs my view that people often jump to a multithreaded solution prematurely anyway, when a sequential one is simpler and more maintainable. But a lot of things do require course-grained parallelism. Very few practical problems require massive parallelism. I suspect that a similar thing may hold for quantum computing.
Sorry. I get carried away. But Iâd repeat the overall point that itâs really impossible to predict what skills are going to be useful, and I think a lot of people pushing hard on âteaching kids to codeâ are selling a false dream. All fields are subject to âelite or bustâ and you should pick something (a) you like, (b) that youâre good at, and (c) that is useful enough to somebody that theyâll pay you to do it. That is really the only secret I can think of to having a happy career.
In theory, machines can do everything and everything can be outsourced to somewhere cheaper.
In practice, even though people were saying the above decades ago, generally the more quantitative folks have done better and many of the less quantitative fields (like marketing) have become very quantitative or are being disrupted by the software revolution (like law).
I mostly agree. However, I think a lack of people skills and communication skills may be more detrimental now when there are fewer specialities. E.g., at some time in the past it might have been a useful skill to learn how to do accounting by hand in a ledger book. Now all of that is easily handled by a spreadsheet. At the same time, if your work is mostly technical or quantitative, you are still going to need to communicate with a product manager or equivalent. If you think you can work in a vacuum or gravitate to jobs where you mostly can, thatâs likely to stifle your career. As for people skills, sadly I think a large part of climbing the âcareer ladderâ in well-established tech companies is determining tactically which projects and teams are perceived as having the most âimpactâ (note: impact is important but it does not always match perception).
So I think that ideal career preparation cannot neglect basic communication skills, though quantitative skills are still likely to bolster your employability. The corporate world is not kind to employees that just want to focus on a specialty and I suspect (though I donât have the data) that it is worse than it used to be, since specialized skills are often the ones most subject to automation.
âCommunication skillsâ does encompass a varied set of skills. For example, communication skills for explaining in-depth material to another expert in the general subject area may differ from communication skills for explaining something to a general audience who has no particular knowledge of the general subject area.
Another example of differences between types of communication would be the difference between communicating to neutrally report the facts and findings, versus communicating to persuade (where those most skilled at it can make opinions appear to be neutral facts and findings).
Communicating goes in the other direction too, like listening and reading specs. I am not as good at this as Iâd like to be I have to admit. It is embarrassing to implement something, hear âThe feature doesnât work the way I expect.â and then realize that the expected behavior was specified and discussed in meetings with product management, frontend, etc. Also, just keeping two-way channels open is very important: not letting uncertainties slip by but going back and resolving.
If you are working far enough behind the scenes, or doing a code port or optimization that just has to keep functioning the same way, you might get by without this, but for application development, the requirements process is going to be much of the work, and it is a two-way process. This is one reason Iâm a little skeptical that âSTEMâ is the key to success. That said, you still need to be able to write software, just not necessarily the way you might do it for an ACM contest or top coder.
Whatever skill set you choose, youâd better be on top of it, in which case youâll do just fine. The world of the future needs fewer people, but of greater skills, in each of the skilled categories. Even electricians and plumbers are no exception.
This is more true for fields where technology allows the best to outcompete everyone else.
When it comes to electricians and plumbers, I just donât see that happening any time soon. I donât see some technology that allows the best plumbers to take on 10 times more jobs given the same amount of working hours. Maybe technology would expedite the diffusion of knowledge across organizations; maaaaaybe it would allow each plumbing problem to be addressed/solved more quickly, but I just donât see a 10-fold increase in plumbing problem-solving efficiency happening any time soon.
Smart home technologies are already appearing in homes. Electricians already have to deal with integrated smart lighting, applicances, etc. Smart plumbing (self-sealing/isolating pipes, temperature-regulated and on demand water flows, smart toilets/faucets, etc) is also on the horizon.
Sure, and plumbers will have to adjust, but until a plumbing system manages to somehow fix itself if it is broken, I donât see plumbing becoming winner-takes-all.
Cars are now essentially computers on wheels, so the job of a car mechanic is very different from a generation ago, but I donât see a decrease in car maintenance or fewer problems or car mechanics out of a job. If anything, the added complexity has meant more breakdowns and more need for car mechanics.
I thought about that yesterday, though I wanted to avoid being even more argumentative. I had a few emergency plumbing tasks come up this year and tackled them myself. But just getting the right fixture to attach a kitchen sink faucet was a process of trial and error with several trips back to the hardware store. Climbing into that space was pretty tough too. It would have been an easy job for a real plumber, but it would have cost me⊠and more to the point I would have had to choose one and in a case like that Iâd rather just do it myself.
But this is a legacy task. My plumbing wasnât designed for robotic maintenance. It wouldnât have made any sense to do that. But if that was your goal, youâd at the very least have a higher level of standardization. You could put optical coding on the pipes. You could make the space more readily navigable even while keeping it compact. Nobodyâs going to bother to do that as long as they can pay a plumber or occasionally do their own repairs. So we may have human plumbers for a very long time to come. However, there is no reason in principle much of this cannot be self-maintaining. In hazardous environments or where reliability is critical, it makes more sense to automate the task, and it can probably be done.
Electric vehicles will eliminate the most common car service requirements. No more oil changes. Brake pad replacements will be rare, although tire changes might become more frequent due to weight.
In fact, cars are insanely reliable compared to my childhood memories (or maybe it was just my fatherâs penchant for used Peugeots). Itâs a rare thing for me to be able to say âMy carâs in the shop.â I had a 2002 Toyota Corolla and I did bring it in for a little maintenance, like brake work, but I never imagined it was going to stall. I eventually just bit the bullet after 13 years and got rid of it because of a problem with the trunk latch and that my kids when toddlers had ripped up the ceiling upholstery. But honestly, if I had wanted that car to die, I would have had to go to some effort. It wasnât happening naturally. This is absolutely nothing like my recollection of cars in the 70s, even Toyotas, but definitely not American or European cars.
Whatever a human creates, a human will have to operate, maintain, fix and upgrade, creating more jobs.
If there comes a time when a machine can completely replace a human then those machines can then overthrow us like in those 80s horror movies.
I think that there will always be a need for the human touch. While say the supermarket cashier jobs that I had as a teenager have now been replaced by automatic checkouts, young people are now doing home deliveries ordered by people shopping online with a variant software of those checkouts.
While these may require electricians and plumbers in general to increase their skills, they do not cause the most elite ones to win all or most of the business (or have the capacity to do so if somehow every customer came to the elite few).
Actually, cars have been getting more reliable over time, and maintenance requirements have decreased. However, when something does break, it can be difficult and/or time consuming to fix. The computerization of cars also means that some fixes are not within the realm of the mechanic, but mainly dependent on the manufacturer delivering a new version of the software with bug fixes.
I donât anticipate new technology that will greatly enhance maintenance of my 1950s era home plumbing (or even plumbing installed in 2021). The point would be to construct a home that was ready for automation in the first place. I share your skepticism that itâs coming any time soon, but I think itâs backwards to look at plumbing as installed now and ask how a robot can do repairs. The whole system would have to be integrated. The cost vs. benefit still probably means we will have plumbers for a long time to come, but thatâs only because buildings are long term investments. There are many repair services that have gone by the wayside, like electronics repair. Itâs cheaper to replace, and the replacement unit was at least soldered together automatically, even if some of the assembly was done manually.
About twenty-some years ago, I wanted to future-proof my house by installing ethernet cables and a few other types of wiring to every room in the house. Few regular electicians were able to do it. The ones who specialize in this type of wiring wanted to charge many times what a regular electrician would charge. Something similar still exists today. The bifurcation is real and will become more pronounced as homes become more high tech.
Servicing high tech gears may create some new jobs in new categories, but their number will likely be more limited. Smart homes of the future probably self-diagnoses problems and reports directly (because your regular electricians/plumbers wonât know what to do) to the manufacturer who sends a technician over to replace some component that it manufactures.