The Future of the Data Scientist's Role

Data scientists are the new "hotness"; everybody wants to be one and every company is trying to hire one. The barrier to entry is big, but once you are "in", you have joined an elite group of professionals that are in high demand. But will there be a time when data science is so easy and common that everyone is doing it? Will it go through the same transformation that web design went through, going from a highly skilled, in-demand job to something that your grandma did last week? Or will it separate itself by employing the scientific method as a foundation?

First, data science has been around for decades and some of the hottest topics like deep learning have been around 20+ years in papers and research. That said, the boom in data science and the speed at which new tools have been released is enough to take a step back and evaluate the field. So what is a data scientist? Is it that bearded guy with a PhD from CMU that talks about algorithms and draws formulas on whiteboards like an artist? Is it the woman in the office that talks about technologies that you've never heard before like Tensorflow, CNTK, scipy, numpy, SVMs, etc., etc.?

I say that a data scientist is someone that knows they can solve certain problems with data, knows 80% of the math and 80% of the coding to solve those problems, knows the various tools that are available, and how to use some of those tools to solve those problems. Why so vague of a definition? I could take any decent software developer, show them how to use Tensorflow APIs, give them a curated dataset, and they could get results. That person is not a data scientist, even though they could do that over and over again. I consider that person a data engineer, someone that takes the work of a data scientist and makes it reusable, production ready. Could the job of a data engineer disappear in the future? Possibly! As new tools, faster & cheaper resources are made available in the cloud, I can definitely see a future where data engineering is automated.

So what about the data scientist, will we automate that job with new tools? Make no mistake new tools are coming out regularly that make the task of doing data science easier and faster. Nobody writes the code to do a linear regression from scratch, they use a prebuilt function in Python, R, or even Excel. Even though it is easier and faster to perform a linear regression, that doesn't mean it is completely automated. A human still has to understand the data, understand that a linear regression is the correct model to build, and understand the results. And in most cases, a lot of data cleansing and feature engineering has to be performed before they can actually get meaningful results. Can those tasks be automated or made so simple that an untrained human could do them? Possibly! Teams are working on AI right now to automate data cleansing, fundamentally neural networks and deep learning automate feature engineering, and there are tools available to help me figure out what and if I chose the correct model.

There is no doubt that we are on the precipice of a revolution in machine learning. Tools have made it easier for data analyst to "do data science" and more are being created every day. I have no doubt that my kids kids will learn data science in middle school. And while tools will make the job faster and building models easier, the human element will never go away. The barrier to entry will become smaller and more approachable, it will still require a scientific mind to be a data scientist. And that is what differentiates data science from other "computer jobs", where those jobs are creative in nature like web design or building apps and we can create tools to make the creative process easier, data science has more in common with physics or chemistry than with software development.

Now to the original question, will there be a time when data science is so easy and common that everyone is doing it? No. There will be more data scientists, performing various degrees of work, but they will remain in demand because of their scientific approach to data. Some of the work that data scientists do now will be done by data analysts because it will be easier and more approachable. However you will still have the demand for that scientific mind in every company.

Agree/disagree, please comment.