Questions to be answered:

What is the most likely scenario that will result from greater use of technology?

In what ways might the economy adjust to maximize the use of technology?

Case Study

Introduction

The year 2017 represented a landmark accomplishment for the development of artificial intelligence (AI). It was then that AlphaGo, a project engineered by technology giant Google, beat the European grandmaster of the ancient Chinese board game Go by five games to none. There had been a few earlier displays of the apparent superiority of technology over human intellect that garnered considerable press attention, as when the IBM supercomputer program Deep Blue beat the chess giant Gary Kasparov in 1997 and the IBM computer system Watson flexed its power over all-comers on Jeopardy (though ultimately defeated by a human). The AlphaGo success, however, represented a departure from the previous approaches to mastering a game that were demonstrated in the earlier examples.

So what was it about these earlier instances that pale in comparison to the most recent technology triumph? In a word, it was the method through which the machine managed to learn. The earlier efforts relied on a straightforward strategy of simply facilitating massive recall and simple application of a statistically defined decision tree to be built into the computer programming. In any given scenario on the chessboard, it was determined what the statistically optimal move would be for any given piece in order to increase the odds of capturing the king. The strategy that successfully subdued Kasparov was, appropriately enough, referred to as "brute force." There are a considerable but manageable (for a computer) number of permutations that must be weighed when arriving at an optimal strategy for winning at chess.

Mastering the game Go raised the bar for success considerably, with its 30 million potential moves to consider. The limits to the earlier efforts soon became apparent in that those programs could sharpenbut not move beyond replicatinghuman recall; for Go, the number of permutations and calculations required to arrive at an optimal statistical choice would overwhelm the Deep Blue program. What was necessary to meet the challenge of winning with Go was the development of a learning strategy. Essentially, deep learning computer programs were pit against one another to permit a constant, iterative learning strategy. The result was that a computer program interacting with another computer program perfected its understanding of how to master the game. In many ways the approach works to move AI just that much closer to passing the Turing test, which requires a machine to mimic human behavior in such a convincing manner that humans cannot distinguish between fellow human and machine.

When we come to realize that technology is moving briskly toward sentiencethinking as a human doesthe story moves from one that generates curiosity to one that has been fodder for enduring visions of dystopias. Anxiety results from visions of humankind being forced into subservience borne from the rise of the machines. In Stanley Kubrick's classic film 2001: A Space Odyssey, The Hal 9000 onboard computer famously rejects an override to its system and throws the offending astronaut into space. The implications of the computer winning at the Go board game raise the specter of an existential crisis of less dramatic but nevertheless vital concern. What if we could produce an automated workforce, effectively putting human labor out of business for good? After all, AI is the use of computer programming that replicates human intelligence in terms of learning and decision-making, which includes the analysis and interpretation of data such as visual and speech recognition. Automation is the application, by dint of engineering, of these insights into action. Entrepreneurs are eager to lift technology from the blueprint to the real world of work.

The economic implications that arise from considering such breakthroughs are profound. The potential applications within our lifetimes range from driverless cars supplanting vast segments of workers in the transportation industry all the way to displacing medical professionals through the use of more accurate readings of medical data such as CAT scans and cancer screenings. Many are left behind in today's modernizing economy, with Chinese car manufacturers and Elon Musk (CEO of Tesla) actively working to completely automate their assembly lines with use of AI, while fast food providers such as McDonald's and grocery stores are phasing out check-out staff in favor of automated kiosks.

Pessimism

The predictions founded on these observations are troubling to many. Economist Tyler Cowen's New York Times best seller The Great Stagnation set the trend for the genre. He argues that while technology is generating enormous income, there are not many benefits for the average worker. He draws a contrast between the technological breakthroughs of the combustion engine of Henry Ford, resulting in a massive auto industry, complete with the generation of a top-tier U.S. city (Detroit) and contemporary innovations. That collective effort employed tens of thousands of workers. Meanwhile, more recent technology has not produced the same dividends in terms of employment numbers.

Here are the (approximate) employment figures for some of the top Web companies (Cowen, 2011):

Online Industry Employment levels

• Google20,000
• Facebook1,700
• eBay16,400
• Twitter300

Although these companies generate a greater amount of employment and revenue indirectly, still our major innovations are springing up in sectors where a lot of work is done by machines, not by human beings (Cowen, 2011, p. 50).

While Cowen is bullish on there being an enduring need for human labor, albeit represented in a shrinking percentage, others have taken the projections much further. While The Great Stagnation was skeptical of the idea that technology has actually resulted in much growth of innovation, the much cited work Race Against the Machine (Brynjolfsson & McAfee, 2012) argues the opposite. In fact, by their account the process of innovation begetting yet more gains in innovation is happening. Yet millions are being left behind. The timing of the work, in the immediate aftermath of the Great Recession that began in 2008, meant that the story would be linked to the jobless recovery (Vardi, 2012). The characterization of machines displacing human input sounds to be a plausible explanation of how the United States could demonstrate gains in production efficiency while shedding labor at the same time. In fact, according to the New York Times "Technology" page, the economic downturn may have forced many companies to consider accelerating their investment in AI as a cost-saving measure. Citing Brynjolfsson and McAfee, Lohr (2011) indicates that during the Great Recession 1 in 12 people lost sales jobs. That trend corresponds with a spike of 26% in corporate spending on equipment and software. Companies are achieving the same ends while cutting their payrolls.

A recent study by Oxford University scholars put a numeric figure on how many industries will be adversely affected in the years to come. Frey and Osborne (2017) survey the prospects of computerization crowding out job demands in 702 occupations. Their estimates were startling; nearly one-half (47%) of U.S. occupations are "at risk" of becoming computerized. They state further that, "We provide further evidence that wages and educational attainment exhibit a strong negative relationship with an occupation's probability of computerisation" (Osborne and Frey, 2017; p. 254). Abstracting from the statistical jargon, they are indicating that having greater educational credentials and being in a higher wage occupation means you are less likely to become technologically unemployed.

Ambivalence

The middle ground between arguments forecasting gloom and those buying into optimism are those arguments that explain the inherent limitations of simply replacing humans with AI. The Washington Post, in an article entitled "The Robots Are Winning," concedes that there are a number of unskilled jobs for which innovation has yet to master the basics.

Surprisingly, the jobs that have proved the most resilient to this, so far, are those that rely on fine motor skills: humanoid robots haven't proven very successful at subtle movements and have an unfortunate 'habit' of falling down stairs. (Kliff, 2011)

Ezra Klein at Vox.com criticizes the negative forecasts by highlighting an innately human characteristic that consumers desire and that AI technology has been unable to replicate, sociability.

People want to interact with other people, even when they'd be almost as well off interacting with a computer interface. And so even when IT makes it cost-effective to replace people with computers, that often leads companies to plow the savings into more people to work alongside the computers (which is largely what happened with banking). (Klein, 2017)

The above point highlights another limitation of the argument that all labor will eventually be displaced. Beniger's (1986) work on the interaction between man and machine indicates that there are strong Luddite tendencies embedded in human nature. The Luddites (1811-1816) famously organized protests and sabotaged the equipment they blamed for their poverty and unemployment resulting from the automation of textile manufacturing. In The Control Revolution (Beniger, 1986), the author's thesis is that the hostility toward innovation eventually wears away as we come to appreciate how best to utilize it. The dominant trend is for humans to find clever ways to subject technology to our collective will. Just as the unfortunate death of a bicyclist by a self-driving car recently resulted in a panic (Baig, 2017), so did deaths from train accidents in an earlier era (Beniger, 1986). These accidents remind us that we are submitting ourselves to the operations of machines beyond our immediate control. Through a combination of regulation, product improvement, and user education we can manage to reap the rewards to be gained by massive reductions in roadway fatalities. Of course, even optimists should not be peddling myths that those gains will be acquired without real costs.

On this last point, Forbes magazine cautions against the use of gloom-and-doom rhetoric, arguing that the introduction of AI into the workplace will take place in phasesthree to be exact (Lieberman, 2018). In the meantime, there will be plenty of time to adjust and soften the blow. The first stage is already underway, with technology performing simple computational tasks and analyzing structured data.

The second wave, which is just beginning, is that AI performs repeatable tasks (such as filling in forms), exchanges information automatically and analyzes unstructured data in semi-controlled environments. The third wave of AI will extract and analyze real-time data from multiple sources, then make decisions and take physical actions with little or no human input. (Lieberman, 2018)

The author cautions that the last wave will not take place for 10 to 20 years. Furthermore, projections that far in the future are difficult to make with much accuracy.

In fact, many people are questioning whether we are even in the first stage. There are three primary data points that indicate technology has not demonstrated much in the way of negative impact on jobs (Surowiecki, 2017). The first is that automation should be driving much higher productivity growth (as seen in producing more goods with less input required). However, by historical standards, productivity growth in the United States is rather low.

Second, many states are now concerned not with labor surpluses but with labor shortages. Third, if machines are displacing human input the numbers on occupational churn (i.e., employees leaving their jobs) would not be where they are now, which is fairly low.

Optimism

The third perspective is one that genuinely embraces the forecasts of AI replacing a number of jobs. There are gains to be had in terms of the greater human satisfaction that can be reaped from the advance of technology. A perfect example to illustrate this point is the automated teller machine, or ATM for short. There were great fears that its widespread use and round-the-clock availability would bring about the demise of human bank tellers. In the provocatively titled article critical of AI wrecking industry fears, Robopocalypse Not, the leading technology publication Wired summarizes the story told in the data.

First introduced around 1970, ATMs hit widespread adoption in the late 1990s. Today there are more than 400,000 ATMs in the US. But, as economist James Bessen has shown, the number of bank tellers actually rose between 2000 and 2010. (Surowiecki, 2017, emphasis added)

The common explanation of the finding is that the use of tellers' time was put to better use attending to tasks of a higher cognitive order than handling elementary banking functions like dispensing cash.

The same article goes on to detail additional advantages.

A rigorous study of the impact of robots in manufacturing, agriculture, and utilities across 17 countries, for instance, found that robots did reduce the hours of lower-skilled workersbut they didn't decrease the total hours worked by humans, and they actually boosted wages. (Surowiecki, 2017)

Robert Fogel, Nobel Laureate in economics, explains that developed countries will soon be faced with the challenge of the paradox outlined above. In The Fourth Great Awakening and the Future of Egalitarianism (Fogel, 2000), he suggests that economic gains will be seen instead the glut of leisure time afforded by such enormous gains in efficiency. Society, and especially those who are not well trained, will face the difficulty of gaining the skills required to address the new (welcome) problem of having excess free time. The gains to be had from this situation will allow for economic benefits that can be spread more evenly throughout society. There will be renewed search for meaning in life, to be found in the cultivation of hobbies, education, religious, and familial pursuits. Just as the dramatic disappearance in farm labor (now only about 1.5% of the population) resulted in further specialization of labor coupled with leisure time (e.g., retirement), we should similarly expect to see the same in the aftermath of AI disrupting labor in the coming years.