Man versus machine: Who will prevail? It’s popular fodder for Hollywood films, and also of concern to many in the workforce, especially in light of Gartner’s claim that more than 40 percent of data science tasks will be automated by 2020. So, should data scientists be worried?
The short answer is no, but it’s still a question worth exploring. As analytics techniques grow more advanced, many data science tasks will become automated, including:
Data integration is the process of combining data from disparate sources into a unified, cohesive view. This is essential for businesses to make informed decisions. Currently, best practices around this are divided: many data scientists use automated tools, while others do manual coding. However, given the massive amounts of data that are being analyzed, and the lack of error risk associated with machines, all signs point to this process growing increasingly automated across organizations.
Model building involves collecting data, analyzing for patterns, and making predictions based on observations. There are already tools that are automating elements of model building. Given the increased intelligence and sophistication of machines, and the massive amounts of data being analyzed, it is critical that there remain ample automated resources for this process so that data scientists have the time and bandwidth to focus on other initiatives.
The Bigger Picture
While it’s apparent that we need machines to do cutting-edge data science analysis, we also need people who can understand, interpret, and implement solutions based on this data analysis as well. Here are three indispensable qualities that data scientists bring to the table that machines do not. At least, not yet.
Behind every dataset is a question. What is the ideal pricing strategy to target customers during a summer clearance sale? If I replace one ad with another on a B2C E-commerce site, will it generate more clicks? How should I customize the marketing outreach plan for Group A, compared to Group B? Ultimately, every algorithm is powering an experiment designed to solve a problem that only a human has the power to articulate. We need data scientists’ curious minds to ask the questions before they work alongside machines to answer them.
Ultimately, once a “data experiment” is underway, a data scientist’s critical thinking and judgement are essential in monitoring the parameters and ensuring that the customized needs for their specific business are being met. Machine-generated algorithms, therefore, can only work under the supervision and judgement of a discerning data scientist. As Joel Shapiro, executive director of the data analytics program at Northwestern University’s Kellogg School of Management says, “Analytics still rests fundamentally on good critical thinking skills —how to ask good questions and rigorously assess evidence that can lead to action.”
What happens after you ask a question, gather data, run necessary algorithms, and review findings? A whole new phase of work begins. The data scientist will need to figure out what the larger business implications are and present takeaways to C-suite executives. Collaborating alongside marketing, sales, and engineering, solutions will need to be implemented and deployed based on these findings. Ultimately, the interpersonal conversations driving these initiatives, which are fueled by abstract, creative thinking, surpass the capabilities of any modern-day machine. And it looks as if, for the time being at least, it will stay that way.
The takeaway here? There is no reason for data scientists to be afraid of machines’ capabilities. Rather, they should view the advancement of technology as an exciting opportunity to expand what is possible within the scope of their roles.
Want to learn how to incorporate automated machine learning into your data science workflow? Enroll in our webinar on predicting loan defaults in partnership with H20.ai.