As someone who considers themselves something of a data scientist, this is an important question. Unfortunately, the answer is: it depends. It is true that some data scientists will be automated out of their usefulness. I’m not a fan of the term ‘data scientist’ for a whole bunch of reasons, not the least of which is its variable definition. For the purposes of this piece, we will define data science using Wikipedia’s definitions: “[Data science] is an interdisciplinary field about scientific methods, processes, and systems to extractknowledge or insights fromdata in various forms, either structured or unstructured.” In short, data scientists are people who practice data science (mind blown, I know).
Data science defined
Data science can be broadly defined to consist of three categories or processes: data acquisition or mining, data management, and data analysis. At first blush, data scientists don’t seem to be all that automatable. For one thing, data scientists already use automation to great effect, but are still involved in the process because of the creativity that is required for success.
Producing creativity
In my opinion, creativity is the greatest defense against automation. Although computer technology will get there eventually, producing true creativity is pretty far down the line toward complete artificial intelligence. By the time we get there, we should probably be worried about SkyNet and not job security. At present, automation is best applied to predictable repeated task. If we look at the three elements of data science mentioned earlier and try to broadly apply these criterion for likelihood of automation, we might be able to partially answer the title question.
Data mining
Data mining is simultaneously ripe for the picking for automators and is also a likely no-automation stronghold, at least in part. Data mining consists of a variety of processes that are often quite tedious. There is also a lot of redundancy or repetition in the performance of data mining. Let’s say that there is a government agency collecting metadata on every phone call placed inside a country. Using any number of data mining techniques, a data scientist could use this metadata to pick out all kinds of interesting things, such as relationships between where calls are made, and who the calls are made to.
Most of this mining would be performed by algorithms that repeatedly comb new data and old data, connecting points and creating usable information for a seemingly infinite pile of individually useless phone call metadata. So much of this process is already automated, but somebody is still there to create and implement the algorithms that are at the core of the automation process. These algorithms might be specifically or generally focused. They may need to be changed as the needs of the government agency changes. So, even if the process is highly automated, data scientists will still have to be involved in the short to medium term.
Data analysis
Data analysis sits in a similar place as data mining in terms of likelihood of automation. Data analysis requires a lot of creativity up front and at the end. Data analysts need to come up with a plan to analyze data, implement the plan, and then interpret the results in a meaningful way. Right now, the easiest part of this process to automate is the implementation. Eventually, artificial intelligence will advance enough that AIs can plan, implement, and interpret data analysis completely with no human involvement. I think this is still a long way off (decades even), and again, keep SkyNet in mind (one can never be too careful).
Data management
Data management seems like it should already be automated. The design of databases does take plenty of creativity, but it’s the creative implementation of fairly rigid standards. This is a level of creativity that automation can currently handle. Data storage, queries, and other elements of data management are already well within the grasp of automation routines. So, if there is one area of data science that is going to go before the rest, it is definitely data management.
Victims of automation
So the answer is yes, data scientists will most likely become victims of automation, but when this happens depends on their specialty or at least their current work responsibilities. This is really true of almost all jobs, so it’s not necessarily a very illuminating answer. I would say, however, that data science is a pretty safe bet if you’re worried about automation. Many other professions will be completely gone—I’m looking at you, automated car developers—by the time data scientists even begin to come under fire.
Data scientists will become unemployed around the same time lower skilled computer programmers and system administrators are heading to the unemployment line. A few data scientists will continue to be employed until the bitter end. I believe this is true of any profession that involves the creative use of technology.
Data scientists should not post their resumes yet. I believe the data science industry will grow for at least the next two decades before automation begins to take its toll. Many data scientists, due to their contributions to automation, will actually be the cause of other people, and perhaps themselves, losing their jobs. But, to end on a happy note, suffice to say, data science is certainly safe for the near future.
About the Author
Erik Kappelman wears many hats including blogger, developer, data consultant, economist, and transportation planner. He lives in Helena, Montana and works for theDepartment of Transportation as a transportation demand modeler.