Sunday, September 23, 2018

Agrawal et al., Prediction Machines

Prediction Machines: The Simple Economics of Artificial Intelligence (Harvard Business Review Press, 2018) by Ajay Agrawal, Joshua Gans, and Avi Goldfarb, all chaired professors at the University of Toronto’s Rotman School of Management, is an excellent introduction to the opportunities and limitations of AI. Their thesis is that, as things now stand, AI is becoming an ever cheaper community that lowers the cost of prediction and will change how businesses operate. But it cannot replace judgment, so human beings will not be sitting around as unproductive idlers.

One of the goals of enhanced prediction is to replace satisficing with more optimal solutions. For instance, airport lounges are “imperfect solutions to uncertainty” and “will be undermined by better prediction.” With traffic apps and apps tracking flight delays, the traveler has new options such as “unless there is a traffic problem, leave later and go directly to the gate” or “if there is flight delay, leave later.”

The authors explore various ways in which artificial intelligence enhances decision-making, especially in the business context, but in the end (at least so far) it cannot replace human beings. We have data that machines don’t, from our senses to data that we opt to keep private. Moreover, prediction machines are often stymied by rare events that are difficult to predict because of a lack of data, “including presidential elections and earthquakes.” (Not that human beings and their statistical models did such a great job with the last presidential election.)

The authors speculate about which countries might have an advantage in developing AI. So far the United States is the world leader in terms of both research and commercial application. But “the trend lines are changing.” The future of AI may be “made in China,” as the New York Times suggested. First, China is spending billions on AI. Second, it has more people—and therefore more data (the new oil). Third, China doesn’t regulate privacy, so easy data access is another advantage. And, I would add, a trade war focused on twentieth-century manufacturing is unlikely to derail China’s efforts to gain an advantage in the twenty-first century.

No comments:

Post a Comment