Today we’re back talking about why I save such a high percentage of my income.
One of the reasons that we are being paid less for our labor is the increasing reliance by companies on automation. This increased automation is a serious enough concern that it gets its own article today.
In technology, there is a rule of thumb called Moore’s Law. What Moore’s Law says is that the number of transistors that can be fit onto circuits doubles every 12-18 months.
In and of itself, that sentence means absolutely nothing to me.
What it signifies, though, is that technological growth is exponential.
Technology advances, which allows technology to advance even faster, which allows technology to advance even faster.
As we are money nerds over here, think of it like compound interest. When you invest money, your money earns interest. Then your interest earns interest. Then your interest’s interest earns interest.
The pace of growth gets faster and faster.
This exponential growth is what makes it possible for you to carry around a phone in your pocket with more processing power than all of the technology that we used to land a man on the moon. Millions of times more power.
Unexpected Advances in Automation
What Moore’s Law means for automation is that it will happen faster than we expect.
There were a number of studies, papers, and books in the early 2000s that discussed automation. The belief at that time was that automation would only be able to replace routine, repeatable tasks.
Driving is one example of work that supposedly could not be automated. Surely there was no way to program a car to take a left turn in traffic. There is far too much human perception that needs to go into making those kinds of decisions.
Another example, from a 2003 study, was “deciphering the scrawled handwriting on a personal check.” This is non-routine because everyone’s handwriting is different. While it is easy for a human to read different writing, it was not easy for a computer.
Except that now we deposit checks at ATMs all the time. We can even deposit checks by taking a picture of them with our phones. (Think about how that last sentence would have sounded in the 1990s.)
We’ve also got driverless cars on the road successfully making left turns in traffic every day.
These are tasks that specifically were mentioned as long-shots for automation that are automated a decade later.
If you feel like your job is exempt from automation, it might be time to do a little more research.
(And if you think you are safe because you are involved in a job that requires creativity, then I’d like you to meet the robot composers that have been writing music since 2010.)
How Automation Expands
There are two main ways that automation has been able to break into so many fields. The first is splitting up tasks. The second is machine learning.
(I also apologize in advance to any tech experts in the audience. I am new to this area, and so my explanations are overly simplified. If you believe any corrections or clarifications are necessary, let me know in the comments.)
First, splitting tasks. If one person is designing and building a piece of furniture, it requires a very high degree of skill. If instead, you split this task among ten people who each have one small part of the process, it suddenly becomes more routine and repeatable.
This was the idea behind the assembly line. If each person had a smaller and more specialized task, the team as a whole could work more efficiently and create more products.
But tasks that are routine and repeatable are also susceptible to automation. Once something becomes routine, it becomes possible for an algorithm to take over. This is why we are seeing a lot of automation right now in manufacturing.
The next method (and the method that was not predicted by the early-2000s prognosticators) is machine learning.
The simple explanation of machine learning is that humans write algorithms that then improve on their own as they are exposed to new data. This is how Netflix learns what you like to watch and how Facebook learns what to put in your newsfeed. The predictions get more accurate as you provide the algorithms with more data about yourself.
The implications of machine learning are really fascinating. For an understanding of how machine learning could cure cancer in our lifetimes, check out The Master Algorithm. For an understanding of how machine learning could lead to the extermination of all humanity, check out Wait But Why’s Road to Superintelligence part 1 and part 2.
For our purposes here today, machine learning means that the tasks that can be automated in the near future may be tasks that we cannot predict. Big data can be absorbed to solve lots of engineering problems that seemed insolvable, like teaching a car how to drive itself safely and learning how to read messy handwriting.
The Oxford Study
The most commonly-cited study on automation is a 2013 Oxford study.
(Normally I read the studies so you don’t have to, but this one is actually an interesting read. It has a well-written discussion on the relevant background information and what to expect going forward if the topic interests you.)
Most mentions of the study cite the top line number and ignore everything else. And, to be fair, the top line number is a huge deal that everyone should know.
The study found that 47% of jobs that people are working in today are at high risk of being fully automated in the next 10-20 years.
47%! That’s almost half of all jobs! In the next 10-20 years!
And I’m here to tell you that that may be underselling the total changes.
The 47% number that the Oxford study lands on is for jobs that can be fully automated. This means that it excludes all jobs that will lose workers due to partial automation.
The example the Oxford study uses is that of construction. Robots cannot really take over the job of a construction worker building a house. They can’t traverse an ever-changing job site in unpredictable weather to do a wide variety of different tasks.
But they can make parts of houses in a warehouse and then ship those parts out.
And this is what we see happening more and more. Prefab homes are built in a warehouse piece by piece. The sections are then shipped to the construction site and assembled there.
This process won’t eliminate the need for homebuilders, but it will certainly reduce the need for them. You need a much smaller crew to assemble a home prebuilt by robots than you do to build a house from scratch.
The same type of trend is (and will be) true in my field. As a lawyer, I do a number of different tasks. For example, I interact with clients, review and analyze documents, research case law, write memos and motions, and argue cases in front of judges and juries. As a government lawyer, I also do a lot of work that at private firms would be done by legal assistants and paralegals.
As noted in the Oxford study, legal assistants and paralegals are at high risk to be fully automated out of existence in the next 10-20 years. On top of that, document review and legal writing are already being automated and the quality of those processes will continue to improve. Researching case law has also become easier and easier with the advancement of technology.
While interacting with clients and arguing cases are at a very low risk of automation (at least in the near future) a significant amount of work could be automated off of my plate. And those of my coworkers. At that point, we may only need a third of the current lawyers on staff to do the same amount of work. Maybe fewer.
Add to that the online companies that are starting to computerize a lot of legal tasks (think Legal Zoom and Novo) and you’re looking at the same pool of lawyers competing for a smaller and smaller list of jobs. Not only will this make jobs harder to come by, it will also drive down salaries as the supply exceeds demand by more and more.
All of this probably makes it sound like I am predicted a permanent 60% unemployment rate and the collapse of the American economy. I’m not.
We’ve gone through economic revolutions before and the economy as a whole has come out the other side just fine.
The Industrial Revolution was a massive economic upheaval. It displaced entire industries and left workers unemployed. But it also created new jobs that nobody could have predicted. The entire idea of a factory job was nonsensical and unimaginable before the industrial revolution.
This is the pattern of economic revolutions. Creative destruction as a process creates new jobs and industries while destroying old ones.
It’s possible that this time is different. Maybe this time we only get the destruction. Maybe we end up with a permanent 60% unemployment rate.
But I am always skeptical of anyone who says “this time is different” to pretty much anything. So I am moving forward under the assumption that the economy will be fine in the long run. I am anticipating that our kids will choose from a number of jobs that we haven’t even imagined at this point.
Automation and the Individual
The biggest problem with creative destruction on a massive scale is not to the economy as a whole. Instead, it is to the individual.
The solar power industry has created far more jobs than have been lost in the coal industry. This is great news for the economy.
It doesn’t help the unemployed coal miner who has spent his whole life in that one industry.
So, yes, the economy will probably be fine. But individuals are going to be hurt.
I don’t want to be one of those individuals. If my field is automated out of existence when I am 50, I don’t want to need to learn a whole new field and start at the bottom of a whole new career ladder.
Maybe I will want to. Maybe I’ll see a field that interests me and will want to dive in and learn new skills.
But I don’t want to rely on that path as the only way to pay my bills. I want options.
So what about you? Have you thought about automation in your field? Do you think my analysis is overboard or unwarranted? Let us know in the comments!