Data Science is dead
Disclaimer: my opinions are my own – nobody and no organization is responsible for this post except me.
This week is the Dutch Data Science Week (#DDSW17). Probably the last. Why? Data Science is dead. It is quickly getting replaced by stuff great data scientists built. Every big cloud vendor is doing it. Need to do a churn analysis? Done. Need to predict energy efficiency? Done. Want a recommendation engine? Done. Need to process genders, age and emotion from images? Done. From video? Done. Need to recognize certain logo’s in images? Done. Need to understand sentiment? Done. Need to translate text or summarize text? Done.
What do you need to use these solutions? Just programming skills. Being able to talk to a REST API, that is all you need. Any programmer can do this. Behind this REST API is a algorithm built by the best and brightest of data scientists in the world that made themselves (and their fellow data scientists) obsolete.
Of course, I know that all these pre-canned algorithms will not give you the best possible results ever. Let’s be honest however, does your algorithm really give stellar results? Do you always need a 99,9% certainty or do you even want it? No – so, if I can just call an API and get the prediction with 80% that is good enough (depending on your use case). I would be done quickly, get the results and that would be it.
That leaves us at my original statement: data science is dead. You will not need to have a data scientist in-house or hire one for the majority of your use cases. What you need is someone that can use the APIs of the intelligence vendor of your choice and that would be all. Done deal. Not data scientist required. Luckily data scientist are scarce, so replacing them with pre-canned APIs that perform just as well or better is a good thing. This does call for a different kind of data scientist, or maybe more of an data artist, one that can orchestrate the code that ingests the data, call the APIs and present in easy to use and esthetically pleasing images. That might be the future for the current data scientists, who knows.