AUSTIN, Texas – In today's big data era, the U.S. is starving for data scientists, according to panelists at Dell World here. There are plenty of opportunities in data analytics, they said; the big challenge is in finding people who can pluck useful insights from the mountains of information.
(From left) Moderator Geoff Colvin, Michael Chui, Thomas Hill, Tom Reilly, and Shyam Sankar discuss big data.
“Actually finding people who can extract insight, wisdom maybe, from increasingly diverse real-time sorts of data is truly the bottleneck,” Michael Chui, partner at research firm McKinsey Global Institute, said during a panel discussion. “The set of skills partly around statistics, partly around machine learning, around visualization, around being able to design experiments…these are the scarce resources.”
Hill said the United States’ economy is currently lacking 140,000 to 180,000 data scientist positions to handle the demand expected in 2018.
Panel moderator Geoff Colvin, Fortune magazine’s senior editor, said the major challenge in data analytics is no longer “getting your hands on data.” But Colvin questioned whether panelists expected too much from a population that hasn’t been trained.
“The data scientist is the unicorn of our industry today... It’s a very hard job. You can take business analysts today and retrain them on different concepts about how you think about data,” suggested Chui. “Data science will become less of a unicorn and will become business analysts using tools they’re comfortable with.”
Data scientists should be deployed into the business to “steep and marinate” in order to find problems suited for data analysis, said Shyam Sankar, president of data analytics software company Palantir. Such creative collaborations will be key to a company’s data future, he said.
“This is not an IT problem or a business problem -- it’s one where IT and business have to collaborate. Teams that don’t have those two members don’t succeed,” said Cloudera CEO Tom Reilly.
At the same time, analytics systems need to be more easily accessible.
“If you have 10,000 modules, you can’t hire enough data scientists to manage them all," Reilly said. "You need the ability to build, calibrate, test, and detect modules with a degree of automation. It’s a challenge that’s upon us and may actually ease the personnel bottleneck,” he said.
While industries will be charged with creating tools to help business leaders and analysts become more comfortable with data science, Chui suggested something he’s thought about for years.
“All of us will have to raise our skills. We should teach less calculus in the U.S. and teach mores statistics,” he said.
— Jessica Lipsky, Associate Editor, EE Times