Machine learning offers a wealth of opportunity to companies in terms of faster response to opportunity, risk avoidance, sentiment analysis, fraud detection, and the ability to make targeted offers to customers.
But along with the many answers it provides, it also presents a variety of thorny questions, according to David Weinberger, senior researcher with Harvard’s Berkman Center for Internet & Society, who will give a keynote talk at Data Summit 2018 titled “Once We Know Everything.”
Data Summit will take place May 22-23, at the Hyatt Regency Boston, with pre-conference workshops on Monday, May 21. Cognitive Computing Summit will be co-located at the event.
“It used to be that to get information into a computer we had to reduce it to the essentials,” said Weinberger. “Computers when they were introduced were taken that way by our culture. They were symbols in the 1950s and 1960s of conformity and reduction of people to numbers. That was the general cultural critique of them—and it was right.”
However, in the past 15 years, three developments have occurred that have vastly expanded the possibility of what can be accomplished with computers.
One is big data, and the way that machines have become “much more capacious and powerful,” said Weinberger. The second is the internet which connects people, ideas, machines and devices in as many ways as possible. And the third, he said, is the rise of machine learning which automates the development of models out of data as opposed to absorbing data into models that humans have created.
In some ways, Weinberger said, the most disturbing aspect of machine learning may be that it can provide good answers and classify information very accurately and at a scale that humans simply cannot, but yet be unable to provide an explanation for the results that can be understood by people. “This is pretty new in the world; we are relying on a model that we cannot understand,” said Weinberger.
Machine learning poses a range of issues that must be grappled with, said Weinberger. How important is clean data to fuel machine learning? Are there reasons why data that has not been vetted should be included? What are the advantages of pulling in more data? What are the implications of including historical data? How do we deal with computer-generated results that we can't rationalize? Ultimately, he noted, the combination of big data, the internet, and machine learning changes people’s fundamental idea of how the world works and their strategy for addressing it.
For more information about Data Summit 2018, and to register, go here.
To review the Data Summit program, go to www.dbta.com/DataSummit/2018/Program.aspx.