In 2018 Fast Company declared the Data Scientist the best job for the third year in a row, which I wholeheartedly agree with (besides the Director of Fun at the York National Railway Museum), however the role of data scientist, as we know it, will soon have the same fate as the bowling pinsetters, chariot racers, and human alarm clocks.

In 2000-2010 data science was dominated by masters of herculean subjects, with PhDs in linear algebra and statistics, combined with expertise in the uncelebrated (at the time) field of coding.  Data science truly had an emphasis on the science of manipulating data, focusing on how to mathematically validate significance and trends.  This was a great first step in helping society gain insights from the massive influx of big data, however it now has its drawbacks.

Tipping the balance too far towards degrees of freedom and vectors is great in the ivory towers of academics, but when it comes to practical and timely results for businesses, is not ideal.  I recently heard a story about a team of PhD data scientists at a Fortune 500 company having trouble improving their built from scratch multi-layered neural network model’s accuracy. They spent hours meticulously tuning cryptic hyper-parameters and adding layers to their model with no success.  The data then ended up falling into the hands of an employee fresh out of his undergraduate degree. After quickly looking at the data, his first step was to create a simple regression model and remove all zero values, immediately skyrocketing accuracy, and creating a cluster of self-conscious PhDs.  Despite his lack of experience with scalar multiplication or multi-threading programming, his domain and practical knowledge made all of the difference.

With the increasing power of user-friendly tools and GUIs, and a data science course seemingly available on everywebsite, being able to perform data science will eventually be like being competent in Excel.  Just knowing the ins and outs of data science as a skill will not be enough. The tools will be powerful enough to handle the data “sciencey” aspects, and the fundamental concepts will be taught throughout school, evolving data science into a skill integral to every job role, not a title.  There will be no more data scientist roles, just roles that use data science.

For now, before the data scientist role goes into retirement, these forces of user-friendly tools and democratization of knowledge is increasing the potential of beginner data scientists to get powerful results with the right training.  Beginner data scientists are spearheading advanced AI across Fortune 500 companies developing deep learning computer vision and natural language processing models for predictive maintenance of assets, facial recognition, and generating valuable insights from social media and news.  Data science managers should be raising their expectations of what their teams can achieve, and be willing to invest in training their teams to get them confident with advanced techniques.

Ultimately, although the role of data scientist may be in its golden years, it still currently has amazing opportunities to create transformational changes across businesses, and should be leading the odds to fourpeat Fast Company’s best job award in 2019.

Learn how to get your team of data scientists at any level jump-started to computer vision and natural language processing at https://www.textvox.ai/

Views: 64688


You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by Eduardo Maria T Morelli on July 17, 2018 at 2:47am

I worked as a DBA for 20 years and since 2014 I have been around dealing  with Data Science (creating models and  experiments). Well, for 20 years, whenever  Oracle or SQL Server launched a new version, I heard the same story: The DBA role was going to die.

In order to survive, I simply had to keep my studies up to date. Simple like that. Keep learning and you will never die

I suspect the same will happen with Data Scientist role ;)

Comment by Robert de Graaf on July 14, 2018 at 2:24am

"I do believe the current job role/function at a typical corporation of "data scientist", whose expertise is solely how to work with and model data, will become obsolete."

I suppose on that point of I have a different kind of scepticism - I'm not sure that individual exists, or if they do exist, they're already a failed experiment. The need for subject matter expertise has been recongised from the outset - even the horrendously over-simplified Drew Conway diagram puts it on an equal footing with statistcal and computing knowledge. 

'However, that is only for typical corporations looking to launch practical projects for dollars and cents'

Corporations is interesting - the only point in having a corporation is it contains multiple people. So a data specialist, however meant, working with a software engineer and someone representing the business understanding seems quite rational. I'd say it's even becoming popular.

The trend I do see which became apparent after re-reading parts of Donoho's '50 Years of Data Science' to fact check an element of this response is that there is a skill which used to be considered core to the notion of a Data Scientist which seems to be fading out, and that is facility with 'Big Data' in terms of ability to interact with Hadoop or other distributed databases at the metal (i.e. writing the queries directly in MapReduce rather than using PigLatin or SparkSQL as the case may be) combined with deep statistical skills (Masters level at least) in a single person. The Big Data stuff seems to have been more hived off to Data Engineers now while others analyse the data. So the 'Big Data' Data Scientist seems to have become obsolete, or perhaps it could be more accurate to say that employers have called off the search to find one just as the search for El Dorado eventually petered out.

Comment by Matt Tucker on July 14, 2018 at 1:32am

Appreciate all the thoughts, very interesting!

I agree that the actual field of data science will continue to evolve and provide value beyond what we can currently imagine, and those who go beyond pre-built frameworks will be leading the way.  

I do believe the current job role/function at a typical corporation of "data scientist", whose expertise is solely how to work with and model data, will become obsolete.  With all job roles needing to be competent in how to interact with data, those with deep domain knowledge will bring the most value, not those with deep insights into neural network architecture.

However, that is only for typical corporations looking to launch practical projects for dollars and cents.  For society as a whole, we need people pushing the boundaries on the theoretical concepts and applications of data science and AI.

Comment by Robert de Graaf on July 14, 2018 at 1:16am

I'm sceptical, partly, because I've been hearing this for a while - more than a few years - without seeing any trend in that direction.

Less trivially, my observation is that most data scientists have skills gaps due to the need to develop deep skills in disparate areas, and those skill gaps lessen efficacy - as illustrated by the proferred andecdote. To my mind software taking up any of the skills needed will first and foremost increase the efficacy of data scientists - and as data scientists (as alluded to in the article) are disillusioned partly due to a lack of efficacy this will improve both data scientist morale and also increase data scientist demand. That is, rather than cause the end of data science, these tools look to me to possibly be what's needed to save it.

Robert de Graaf

The Lazy Data Scientist

Comment by JC Rio on July 13, 2018 at 10:39pm

There is not a month when I interview someone boasting to be a data scientist only because he knows how to make an OLS regression (but don’t even know the assumptions). Not a month when I see a company that says performing deep learning when they only apply a PCA.

I feel like there are more and more jokes in data science and less serious scientists. When I question candidate on state-space model, some has no idea of what I’m talking about, some even struggling to calculate a max likelihood function.

It’s amazing to see all those tools available for the general public, but pure science still needs to be respected.

Comment by Tom Osborn on July 13, 2018 at 8:33pm

If the old dogs aren't achieving, but a young pup does using simple regressions, there are problems with the old dogs: Not being curious; not being scientific; addicted specific tools so they don't explore; ignoring basic principles like looking for information content. If a simple model isn't tried up front (to set a baseline), they are lazy, too.

BTW, what do you mean by "remove all zero values"?


Comment by Harshendu Desai on July 13, 2018 at 8:39am

Just  remove the Scientist  after the Data for a moment. Data is a art of science.. We need art to recognize the data which means to  understand the subject  domain  of a data.  The science gives just  methods and formulas to manipulate the data. There are many statistical  and machine learning  methods but  which  one to choose for a specific data pattern where "Art" plays the important  role.  In this regard , we may  call  Data Artist.  But  lonely artist also  without paint brush  and static colors.  So  to  complete two sides of coin , we may  say Data Science Artist which  will  never die.  The reason is Data is "Universe".  

Comment by Ashok Chilakapati on July 13, 2018 at 8:35am

Excellent comment by Mike Richmond. Frameworks and tutorials to build the latest and greatest analysis are coming out of the woodwork. Because there is $$ to be made on this wave. Just because you can use a framework or write some Python code, does not make you a data scientist. Few of these end users and new data scientists have the passion understand and develop deep insights to make a contribution to the "science" they profess to practice. Take the $$ away, they will move to the next big thing with $ signs attached. Not blaming them, but that is how it is. Back in 1990's ability to write html turned one into a pioneer, discoverer, and what not. Guess a lot of them are now doing doing data science. True contributions result from passion & dedication to the subject and they stand the test of time. At least some pf these new data scientists will hopefully stick around for some time to make such contributions.

Comment by Alex Constantin on July 13, 2018 at 8:03am

Hi Matt,

The context is not totally clear: if Data Scientists die, all of them do, including ‘Beginner ones’ & the ones managed by the “Data Science managers” you refer to.
If you instead refer only to the titles of core experts that helped popularize the algorithms that now everybody takes for granted, then the argument still doesn’t really stand because it is implying that Data Science stops after this round — that there can exist no more new algorithmics that will be useful and that the breadth of what the field can do will be exhausted. 
This is far from the truth, and we still need Data Scientists to exist to conceptualize & code new functionality into the ‘tools that handle data “sciencey” aspects’ of tomorrow.

Comment by Ziyad Nazem on July 13, 2018 at 7:57am

I just addressed this in a blog post -- my TL;DR of that post is that we have been calling data science things such as the sexiest job yet we have data scientists that lack domain knowledge and are disillusioned with the profession.

The profession is more than just a science and is truly an art of how we use the science aspects, with our domain knowledge and strategic thinking to answer questions and solve problems.

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service