If your job doesn't require college, then pretty soon it won't require a human either. The more brain required the longer you have until a robot or computer can do it better.
A cleaning lady is going to be able to keep her job longer than a radiologist as the radiologist requires mostly the ability to understand visual patterns where as a cleaning lady requires general purpose AI.
It's way harder to lift the rug, get in between the books etc.
Furthermore you get it wrong.
Specialization requires knowledge yes but mostly requires less of the brain. A cleaning lady have to use all sorts of different parts to do their jobs.
So your rule is right in some way but you apply the wrong people to it. It's the cleaning lady who uses most of her brain not the specialist.
Just folding clothes is only commercially automated in very expensive and limited contexts (can fold shirts, but not blankets), and it won't put up the items for you. There are any number of restaurants that would love a kitchen-cleaning robot that guarantees that part of the restaurant inspector's rounds is always taken care of to a consistent degree. Just a robot modified from a pick and place solution to declutter by putting up out-of-place items and remembering where everything is located would find a large market. We're nowhere near good solutions for these and many other "simple" skills, much less mass market solutions.
I sometimes wonder if a giant "donut" job market will develop: many pattern-based very high-end specialties get automated, and many extremely low-end (Roomba, for example) ultra-repetitive tasks get automated, but lots of low'ish end jobs like custodial work stubbornly remain. Above the income strata of those low'ish end jobs in the middle sits a group of steadily-insecure positions as automation constantly nibbles away at their edges from the higher-end income stratas.
Why would it need to be a single package? Is there a point to making the roomba do laundry?
Initially, we will be forced by circumstance and market to adopt single-purpose robots like the Roomba if we want to avail ourselves of their benefits. But there are so many different tasks that I do not foresee a future of mass adoption of such single-purpose devices; the economics do not support such an outcome unless you imagine a near-future where most of the planet's population is compensated around the current US median household income per year.
But coming back to stock brokers, is there an epidemic of college educated stock brokers ending up unemployed due to automation? I don't think so and that's because they have other skills they gained from their education. Automated trading algorithms don't make stock brokers useless in the same what that automated garbage collection doesn't make programmers useless.
ED: yes, they're not perfect but give it about 20 years. I'm pretty sure we won't have human cleaners by then.
Have you ever had a cleaning job?
A Roomba isn't going to solve your cleaning needs, not even by a longshot.
Also what's your evidence that there's no relation between college degrees and automation? What we know for a fact is that wage growth for people without college degrees has been flat or negative for the past eight or so years, while people with college degrees have seen some wage growth. Note, this can't be explained by illegal immigration because the number of undocumented immigrants has stayed constant or declined since the beginning of the Obama-era.
The reasons why it isn't being used instead of doctors are mostly political, institutional and human. First doctors will use them as assistants then they will be doing all the work.
But when it comes to actually detecting the patterns which is exactly what a radiologist trains to learn computers are not only better at finding the problems they also find much subtler and thus more early stage signs.
Some people in collage with collage degree have wage growth, you go ask your friend who studied english literature when they last had a pay-rise.
The original claim was that collage meant more use of brain thus better job security.
Thats simply wrong on both accounts.
If it were just about reading the films, the tech who does the procedure would read the films. To be effective at a tech job you must know what you are looking at so that you can tell if you have successfully captured images for the doctor's report. If something looks off they may need to take further measurements and images.
Meanwhile, it will take longer for people to build robots for each particular manual job. Not to mention the service industry where a lot of people prefer to deal with other humans, even when it wouldn't be necessary.
Cleaning ladies are being displaced by disposable products that don't need to be cleaned.
Plumbers and electricians are being displaced by labor saving technologies. A new house contains no copper tubing or sweat fittings -- they use flexible hose and swaged connections. Wiring is no longer done with hard conduit. Framing is done with machine made components such as roof trusses.
Inspection is done by bribery. ;-)
And furthermore you need only very few people running a whole radiology business for billions of people.
"The workers own the means of production." ;-)
The point is that you don't need the amount of radiologist you are currently education if pattern recognition gets done by machines.
Why is this? Because many medical jobs consist of looking at information, making a diagnosis, and suggesting the course of action most likely to work. This is something that machines excel at. Many medical professionals never even communicate directly with the patient, leaving that to the doctor, while they assist the doctor with diagnosis behind the scenes. Take a radiologist, for example.
On the other hand, a call center employee must be able to communicate in real time with a human at conversation level. That is doable by AI, but is much more difficult, and requires something closer to AGI.
So the jobs that will be replaced first have nothing to do with education. They have to do with jobs which are reliant upon pattern recognition and information analysis. Those will go first, and some of them are very high paying and require quite a lot of education.
an investment banker typically advises on M&A transactions which is a very human business where your career path is strongly correlated with your social skills and status rather than technical knowledge.
Pretty much everything they do is non-repetitive except at the lowest level.