Artificial Intelligence (A.I.), especially its subset Machine Learning (ML), will undoubtedly shape the future of the U.S. labor market, economy, and society. In the future, the U.S. workforce will be more efficient and have the opportunity to allocate more time towards the meaningful aspects of their work, such as social interaction and creatively challenging tasks. Nevertheless, studies confirm that American adults are less optimistic about the current developments in the field of A.I. For example, according to a survey by Zhang and Dafoe (2019), more than 80% of American adults expressed their concern regarding the careful management of A.I. development. Even some of today's economists are convinced that technological progress in the field of A.I. will lead to severe societal outcomes. Brynjolfsson and McAfee (2011), for example, controversially argue that continuous improvement in A.I. is destined to result in the vast unemployment of today's U.S. workers. In support of this pessimistic forecast, an Oxford study by Frey and Osborne (2017) predicts that around 47% of U.S. jobs will fall victim to the substitution of labor by means of A.I. and mobile robotics. In reality, however, these pessimistic forecasts often ignore fundamental laws of labor market theory and, thus, are to be taken with a pinch of salt.
To understand why and how A.I. will impact the U.S. labor market, the first section of the essay will shortly introduce the field of A.I. Secondly, recent advancements in A.I. and their implications for U.S. employment are investigated. Finally, the third section analyzes the influence of A.I. on the U.S. labor market composition.
Recent advancements in A.I.
AI generally refers to "computer systems that perform tasks or make decisions that usually require human intelligence" (Zhang and Dafoe, 2019, p. 5). Particularly important for this essay is the A.I. subset ML, which refers to "computational methods using experience to improve performance or to make accurate predictions" (Mohri, 2018 p.1). It is vital to understand the basic reasoning of ML because it is fundamentally different from traditional software development. Whereas software developers code algorithms and explicitly write instructions for a computer to execute, ML engineers train self-learning algorithms by exposing them to task-specific data. ML has been around since the 1960s, but the "pace and scale of this encroachment into human skills is relatively recent" as Brynjolfsson and McAfee (2011, p.10) state in their popular book Race against Machine. In today's reality, technological breakthrough projects, such as Amazon's personal assistant Alexa, Waymo's approach to autonomous driving, or IBM's medical assistant Watson are powered by A.I. technology and will increasingly assist and substitute human labor.
The recent acceleration of advancements in A.I. is twofold. Firstly, computational power has grown exponentially since the 1960s. More precisely, computing processor power has approximately doubled in 18-month intervals, which allows computers to process tasks of ever-increasing complexity (Brynjolfsson and McAfee, 2011). Secondly, the emergence of Big Data enables businesses to make use of large data sets. Remarkably, the increasing usage of smartphones has allowed companies to collect vast amounts of data, which has fostered the recent developments in A.I. (Frey and Osborn, 2017). The seemingly ever-increasing computational power and accumulated data in today's globally connected world are the two leading forces in the latest progress in A.I. and, thus, will enormously impact the future U.S. labor market.
Implications for U.S. employment
A.I. enhances human productivity and accuracy in many different occupations by either complementing or substituting the functions that a specific job embodies. In some professions, A.I. may even wholly replace human labor by automatizing all of the involved tasks. Thus, a numerical forecast of the impact of A.I. for U.S. employment requires first to analyze how A.I. influences jobs on the micro-scale. For example, A.I. is increasingly popular as a complement to human labor in the healthcare industry. According to Jiang, et al. (2017), A.I. systems can provide state-of-the-art medical information, reduce errors in diagnosis and treatment, and can be equipped with ML algorithms for continuous self-improvement. Notably, many therapeutic areas, such as cancer treatment, neurology, and cardiology, are likely to benefit from the technological aid that A.I. provides to medical experts (ibid). In a few other industries, A.I. may even completely render human labor obsolete. Driving is among these occupations. This view is supported by the thesis of Yaqoob, et al. (2020, p. 1), which states, "in the foreseeable future, millions of autonomous cars will communicate with each other and become prevalent in smart cities." These examples may only represent a small fraction of potentially impacted occupations but undoubtedly show that A.I. has the potential to complement and substitute many of today's job titles.
Unsurprisingly, technological automatization leads to an increase in unemployment, at least in the short term. For example, AI-powered autonomous vehicles are likely to replace many of the 15 million U.S. employees in transportation and related industries (U.S. Department of Transportation, 2020). Regarding a global and industrywide perspective, a recent McKinsey study predicts that A.I. may replace 400 to 800 million jobs by 2030 (Manyika, 2017). On the other hand, through job creation and scale effects, 500 to 890 million jobs are estimated to be newly created within the same timeframe (ibid). In summation, the overall long-term increase in the real demand for employees will certainly exceed short term job losses resulting from A.I. substituting part of human work.
At first, it may seem illusional to believe that job creation could replace millions of employees who are no longer needed. However, throughout history, technology has not led to long-term increases in unemployment. For example, according to a report published by Fisk (2003), unemployment decreased from approximately 5% in the year 1900 to 4,2% in 1999. Opponents of this view may argue that these numbers neglect critical figures, such as the average length of unemployment. There is no doubt, however, that advancements in 20th-century technology have not led to a long-term increase in unemployment rates.
According to modern labor market theory, two main reasons explain the phenomenon of stable employment despite technological progress. Firstly, technological and business model innovation leads to the creation of new job titles. Goldin and Katz (2009) support this view naming innovation and education as driving factors in augmenting the demand for human labor. Regarding AI-powered autonomous vehicles, inefficient drivers will undoubtedly be replaced by the easily scalable work of relatively few but efficient machine learning engineers and data scientists. Secondly, scale effects arise from the described gains in productivity and increase the demand for employees. According to renowned economists, such as Aghion and Howitt (1994) or Ehrenberg and Smith (2016), productivity gains lead to decreased prices, and thus, the real income of employees rises. A rise in real income allows employees to consume more, which increases aggregate consumption and incentivizes businesses to either produce more or offer more services, which requires more employees too (ibid). Consequently, despite relatively fewer employees necessary for similar output due to A.I., the total number of employed workers may even rise on aggregate since productivity and total output increases.
Implications for the U.S. labor market composition
Humanity has continuously evolved, and daily activities have changed from hunting and gathering to highly specialized contractual work in a globally interconnected world. Indeed, as Frey and Osborn (2017) suggest, employment in the past has shifted from simple agriculture to manufacturing and finally led humanity to today's service-oriented economy. However, A.I. may be somewhat different from prior technological breakthroughs, such as electricity or the combustion engine. Thus, it is crucial to understand that past assumptions about the relative strength of humans vs. technology have to be adapted to the recent advancements in the field of A.I. For example, Brynjolfsson and McAfee (2011) argue that A.I. is increasingly capable of performing highly cognitively challenging tasks. Consequently, A.I. will not only affect predictable, routine labor but will also substitute professions in which humans have possessed comparative advantages in the recent past.
Due to technological breakthroughs in A.I., it has to be reconsidered how automation influences the labor market composition. For example, Levy and Murnane developed a framework suggesting that straightforward, repetitive tasks may be susceptible to automatization, whereas complex pattern recognition tasks could not be automated (Levy and Murnane, 2004 cited in Brynjolfsson and McAfee, 2011). Driving in traffic was explicitly named as one of the jobs not automatable because of the many arbitrary factors and the challenge of decision-making when facing uncertainty (ibid). In 2017, however, Waymo (formerly a Google project) officially announced that its automotive cars had driven more than 4,000,000 miles on public U.S. roads (Waymo, 2017). This clearly shows that past assumptions about what machines can or cannot do are outdated due to the recent advancements in A.I.
An increasing number of jobs will be replaced by A.I., whereas occupations that require fundamentally human characteristics will be less affected. In accordance, Frey and Osborne (2017) suggest that, in particular, jobs that require human traits, such as social intelligence, creativity, or even perception and manipulation, are not replicable through A.I. On the other hand, occupations in areas such as transportation, logistics, and administrative work could be either replaced or at least complemented and streamlined through A.I. (ibid). In short, applying recent developments to the framework of Osborn and Frey, the U.S. labor market will undoubtedly shift further into the service industry. In particular, occupations that require fundamentally human traits, such as empathy and persuasion, will experience a significant increase in demand.
In conclusion, A.I. will replace many of the tasks that are currently performed by human workers. However, the overall creation of new jobs through innovation, as well as scale effects arising from productivity gains, will increase net employment. Secondly, due to further advancements in A.I. and its many business use cases, today's comparative advantages of human vs. technology will change. This will lead to a gradual shift in the composition of the U.S. labor market towards occupations less susceptible to automation through A.I., such as professions requiring social-emotional intelligence. These developments will enable the U.S. society to benefit from rising productivity (GDP per capita), thus allowing the opportunity to allocate more time towards meaningful tasks and more significant social interaction. After all, this will determine that the world will remain a human-dominated domain for a long time.
Word Count: 1701 words
- Quote paper
- Philipp Reinstorf (Author), 2020, Artificial Intelligence and its Implications for the U.S. Labor Market, Munich, GRIN Verlag, https://www.grin.com/document/995292