How Machine Learning is Changing the Way We Think

David Weinberger, a senior researcher at Harvard's Berkman Klein Center for Internet & Society, has written a new book, titled Everyday Chaos: Technology, Complexity, and How We’re Thriving in a New World of Possibility.

Weinberger will showcase Everyday Chaos (Harvard Business Review Press), available from Amazon on May 14, at DBTA’s upcoming Data Summit conference at the Hyatt Regency Boston (May 21-22, 2019).

According to Weinberger, machine learning’s importance is not only in the benefits its use brings, but also in how it is transforming our understanding of how the world works and our most basic strategies for dealing with the future.

How Machine Learning is Changing Our Perception of the World

“The internet and machine learning are having an important and remarkable effect on our understanding and perception of how the world works,” said Weinberger. “The book points out ways in which our basic understanding of how the future occurs has shifted toward the model machine learning uses.”

At its most basic, Weinberger said, “We've traditionally thought about the future as consisting of possibilities that narrow down as they come closer to us in time." He says our basic approach for thousands of years has been to try to make sure that the single possibility that we want is the one that survives the cut, primarily by anticipating what will happen and preparing for it.

But, he said, there are important deviations from that anticipate-and-prepare approach that are evident in the way businesses have been using the internet. For example, he said, companies use A/B testing that lets them try out small variations in ads rather than try to predict which ones will work best. “Companies launch MVPs—minimum viable products—that contain as few features as possible so the company can learn from how their customers actually use them, rather than trying to anticipate those uses." He also points to the creation of open platforms by organizations as large as Facebook, Google, Dropbox, and entire national governments. These platforms let users use the organization's data and services to do things the organization could never have anticipated. "Even billion-dollar game companies give tools to users to let them modify, extend, and even transform their games in ways the game company couldn't anticipate and may not entirely approve of," he points out. "In these and many other ways, organizations are learning the value of holding back from anticipating and preparing.”

An Epochal Change 

The result is that this is changing the way people think about the future.  In essence, Weinberger said, “We're seeing value in opening the future up, rather than trying to narrow it down. That is an epochal change."

He turns to the role of machine learning, which, he claims  "is providing us with a different way of understanding how this opened-up future works."  "Traditionally, we've tried to predict and manage the future by discovering the general principles that govern it. We use this knowledge to build conceptual models that identify the controlling factors in a domain and how those factors are interrelated. But machine learning doesn't work that way.

Clearly, machine learning has made the world more predictable,  but also more confusing. “We use it because it enables more accurate prediction," he says, pointing to advances in how far out, and how accurately, we can predict the weather thanks to this new technology. And, he says, we can now even make predictions about things which we previously thought were closed to predictability such as human behavior. “For example, it can be humbling and embarrassing to learn that humans, at least en masse, are more predictable than we would have thought.”

However, machine learning sometimes uses models that are far more detailed, intricate, and complex than humans can comprehend. There are some data scientists who say that we are on our way to solving that problem while others disagree, he observed. “The price of all this predictability may be human understanding,” said Weinberger.

At Data Summit 2019, Weinberger will consider the implications of these facets of machine learning and explore what it means for society and our moral sense of the world, and be available to sign the new book.

Weinberger will present his Data Summit 2019 talk, “How Machine Learning Is Changing the Future as a Fact and as an Idea” on Tuesday, May 21, 2019, from 4:15 pm–5 pm.