While most of my writing regards pilots, this one is for everyone. That said, please forgive the aviation examples I use! Many people worry about their job security as companies shift manufacturing to other regions of the world while simultaneously implementing automation squeezing people out of jobs. Is your job at risk?
Free trade is in the news these days. Populists argue that we need to raise tariffs to protect jobs. Advocates of free trade argue that the jobs are being replaced more due to automation than being shipped overseas. To be sure, these are complex issues. Companies will shift jobs to manufacturing in a particular country to avoid tariffs as well as gain direct access to markets and in in so doing, protect their market share. Arguably, higher tariffs accelerate this process as it becomes more fiscally prudent to manufacture in the local market. Of course, much of this is not fully understood, and things are not always what they seem when it comes to trade, as this article in The Economist points out.
All the talk about trade deficits misses an important aspect. While a country might be “negative” on trade, all that means is that it is buying more goods then it is selling. However, the primary source of income for most (all?) of the leading economies is not manufacturing but comes from the service industry (for the U.S. services account for around 80% of the GDP, see chart above). Transportation, financial services, information technology and the like are where we find the bulk of the highest paying jobs. As the service sector and the income generated in that sector grows the country tends to see rising wages. This, in turn, creates an incentive where the cost of shipping and importing the product from another location starts making economic sense. It begins at the most unskilled of manufacturing jobs, such as garment workers, and then slowly moves up.
In all of this turmoil, is your job safe from being eliminated? Let us first look at out-sourcing.
In order to shift a job to another country several requirements are needed. If your job is working as a plumber it is pretty difficult to imagine a scenario where someone in a developing country could perform that job (unless they were brought to your country, of course). Similarly, jobs involved in the medical profession are fairly immune. We will continue to need nurses, doctors will still need to see patients, etc. The jobs that are more likely to be moved involve very low skill positions in manufacturing. Of course, these are also the most susceptible to automation. The job of peeling nuts for a store, sewing together a hat or a repetitive task such as putting cherries in a fruit jar are very easy to move and the differences in wages so stark between the rich and developing countries that it become almost silly not to shift production. It leaves us with the option of paying a lot more for the product, reducing the wage so much that virtually nobody will take the job (except perhaps an undocumented worker), or automating it. As it is unlikely that anyone reading this article is working in one of these fields, I will leave it at that.
The other side is the ability to automate the job. The key here is whether the problems that the person faces in the job can be well defined. The more gray areas that require judgment are involved the more challenging defining the task is. In my article on the topic of “Single-Piloted Commercial Aircraft”, and also in my article “Are pilots going to be eliminated” I touched on some thoughts on this topic as they pertain to pilots. It is important to understand that to automate the task one must not only define the task itself well, but also the environment. The more the environment can be constrained the easier it is to define. Weather is a prime example, where the ability to predict what will occur at a micro-level is much more difficult than at a macro level. We can generate a forecast for an area that can be fairly good (and even those can be horribly wrong as everyone knows!), but to define what will occur at a point in time at a small geographical area (like over your home) is far outside anything we can do at present. The problem is not just computer power. I read once that even if we could capture the data of the atmosphere in every cubic meter of the world there still would be enough data missing to result in forecast errors. It is that complicated. So how do we define the job of a pilot that has to deal with the weather directly at a micro-level through using judgment that is hard to define? To be sure, this has also led to a lack of training and evaluating for pilots on this topic, as it is hard to teach and hard to test. Pilots must rely on their own experience, and that can take many years (we can do better here!).
In an attempt to address this, the manufacturers of onboard weather radar systems have made some efforts to automate this. Unfortunately, we once again see limitations. One manufacturer, for example, limited the upper weather display to 60,000 feet. This seems reasonable as only a few aircraft get above that, and there is not much weather at those altitudes. Unfortunately, as I found passing over India one day, some storms get much higher than 60,000 feet. Not a problem? What about the potential for hail blow-off? Hail gets blown far downwind of many storms but it can be possible to get some idea of the extent by examining the expanse of the “wet core” near the top of a storm. Unfortunately, the designers did not anticipate this. In an older radar unit I could have examined the area manually.
In another case, a manufacturer has included algorithms to adjust for the differences in storms around various parts of the world. This is very good, however, it is then not optimized for a storm that might be atypical for a particular region. Pilot knowledge and experience with weather can make all the difference here. All of this make automating the pilot job more difficult than some would like to believe, absent accepting more risk. This is not to say that pilot jobs will not become vulnerable at some point, of course. Interestingly, flight attendants are much more difficult to remove from the aircraft than pilots due to the requirement of human interaction and the need for human variability in an emergency such as a crash.
Absent a major breakthrough, computers remain limited to those problems that we can well define. Even with the hype of machine learning we still see these limitations unless we can create boundaries to the environment and eliminate any surprises. Humans are still uniquely able to manage the unexpected, poorly defined problem. We are also still much better at solving problems that are completely novel.
So, is your job safe from being eliminated due to automation or out-sourcing? Think about what you do and whether you could write a script to solve any problem you encounter. Are there problems you encounter that are outside of anyone’s imagination? If not, does your job require physical aspects that would be very difficult to replicate with a machine (such as crawling through an attic or sub-floor of a house and encountering a different set of conditions and designs each day? Does it require direct human interaction? Through asking these questions we can fairy quickly find where on the scale your job or choice of career might be in terms of risk in the future.
