It is probable that a child born today will never need to learn how to drive a car. At the advent of autonomous driving vehicles, skills such as driving a car could soon be rendered obsolete because of technological advances. While we marvel at the pace of innovation and wonder about its development, the key ingredient in artificial intelligence (“AI”) remains rather mundane – i.e., tiny bits of digital data. Without the ability to quickly collect and analyze mountains of data, AI would not be possible. For this reason, companies have invested billions in pursuit of an information advantage. In certain respects, global economies have entered a digital arms race where data has supplanted oil as the most important commodity.
Purchasing a book over amazon.com or conducting a Google search are wonderfully efficient services that have benefited consumers, students, researchers, and the general public. Technology has fostered greater transparency while disintermediating numerous industries, crimping profit margins for producers but offering discounted prices to consumers. Companies have been forced, in turn, to invest in technology to improve efficiency and preserve margins. The impact of technology can be viewed as a virtuous cycle of continuous investment focused on eliminating inefficiencies.
Behind some of the services we now take for granted, there are powerful engines at work that collect and analyze data at the speed of light. Algorithms (i.e., mathematically based logic sequences, or rules, embedded in computer programs) sift through reams of data to detect patterns and signals. That information is then processed to offer recommendations or to make decisions in less than the time required to blink an eye. In the case of autonomous driving vehicles, sensors placed throughout a car collect data and transmit the information to sophisticated microprocessors programmed to direct various mechanical systems (e.g., brakes). As greater amounts of data are collected, the algorithms become more accurate, and the car’s autonomous driving system becomes more efficient and reliable.
Users of technology (i.e., everybody) are, in effect, producers of data, and data generation often occurs unwittingly. By accepting a trade-off in privacy and security for greater convenience, the typical smart phone user is generating valuable information for the companies that provide the Apps. Each purchase over amazon.com and each search on Google provides those companies with incremental information about you, the user of its service. With each purchase, amazon.com’s future recommendations will become more targeted toward your preferences. With each search, Google’s algorithms will be more efficient in future searches.
Academics describe a “data network effect” where the use of data to improve algorithms for a service ultimately attracts more users. Consequently, the value of a service depends on the growth of its network of users that, in turn, contributes to the value of the data collected (i.e., creating a circular dynamic). For example, the enterprise value of Uber, a pioneer in “ride sharing” and a disruptor to the legacy taxi industry, is not determined solely by the profitability of its service but primarily by the vast treasure trove of data created by its passengers. For the typical Uber customer, this matters little as the service is what is valued. However, as the number of Uber’s customers grows, their use of the service creates ever more data that contributes to the value of Uber as a company.
There are limitations and concerns with the rapid evolution of the data economy. First, the value of each incremental data point diminishes. That reflects the abundance or, perhaps, the infinite nature of data. By way of contrast, oil is a scarce commodity. Second, the efficiency of algorithms can, over time, actually reduce choice. As a long time customer of amazon.com, I find its recommendations remarkably well suited to my preferences. However, I strongly suspect there are products I would enjoy that are not identified by its algorithm. In economics parlance, I believe the efficiency of amazon.com’s algorithm could be creating an opportunity cost for me as a consumer. Third, there are uncertainties as to who owns the data and as to how data can be used. Finally, as the data base of customer interactions grows for large technology companies, there is legitimate concern regarding the development of “data monopolies”. A company with a monopoly over certain types of data can wield enormous power.
To many, the data economy poses a significant threat to economic models as automation aggressively displaces jobs and functions. For certain industries, this prospect is frightening. However, this is not a new development. Historically, job destruction because of automation has been offset by a cycle of new job creation fostered by the same technological advances. Innovation in agricultural equipment reduced the number of farmers. Telephone switchboards used to be operated manually. Robots have assisted in the assembly of automobiles for decades. In the past, technology facilitated the migration of jobs from agriculture to manufacturing and, eventually, to services. Economies and industries adapt to new technologies, and the challenge for many people going forward will be to acquire the skills necessary to remain competitive in a data-driven world.
With each transformative technological advancement, value shifts over time from the producers of an innovation to the users of an innovation. Companies such as Google and Amazon, while they are providers of technology in certain respects, are successful because they constructed business models that took advantage of the Internet. Other companies were responsible for the global build-out of fiber optic cable and internet connectivity. Those companies are now considerably less valuable than Amazon or Google. Today, the most respected, admired, and feared companies are those that shrewdly leveraged the efforts of the early Internet pioneers. That is unlikely to change going forward; companies that incorporate technological advances should thrive.
Automation is also transforming financial services. Most trading of securities occurs electronically (i.e., no human involvement). The floor of the New York Stock Exchange is a ghost town when compared with the throngs of traders, brokers, and specialists of the 1980s. Clearly, on the transactional side of the financial markets, automation has eliminated many functions previously fulfilled by well-trained and well-compensated humans. Yet, I find prognostications of the dominance of algorithmic trading and fully automated markets unconvincing. Trading algorithms analyze patterns in historical data (i.e., ex post, or after the fact) to isolate factors that could prove predictive. The frequency of occurrence in past market cycles greatly influences the estimates of probability associated with future cycles. The challenge of investing is to correctly identify factors ex ante (i.e., beforehand), and in that regard, algorithms offer no apparent advantage. Because of ever changing market conditions, there is a natural limit to the value of algorithms in investing. Prudent judgment exercised by experienced professionals during unfamiliar situations, on the other hand, should remain a valued service. As with other industries, however, successful financial services companies will be defined by the extent to which they incorporate technological advancements.