Artificial Intelligence and the Future of Work (2025) / Chapter Skim
Currently Skimming:
Pages 65-100
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 65... ...
Manufacturing, agriculture, and mining show lower exposure. In contrast to the results from Acemoglu cited earlier, exposure appears to be uncorrelated with both recent factor productivity growth and labor productivity growth by sector.
|
From page 66... ...
Kalakota, 2019, "The Potential for Artificial Intelligence in Healthcare," Future Healthcare Journal 6(2)
|
From page 67... ...
If these cases generalize, that portends well for aggregate productivity growth. Some of the key effects of generative AI can be observed in a paper about the phased roll-out of an LLM-based system designed to assist thousands of contact center workers.69 The research compared agents who had access to this system with those who did not.
|
From page 68... ...
Other interesting examples of the productivity effects of machine learning sys tems are medical image recognition and machine translation. A convolutional neural network trained on 129,450 medical images and 2,032 different diseases was able to diagnose different types of cancer at a level that matched or exceeded 21 board-certified dermatologists.70 The authors argue that AI could provide low-cost access to diagnostic care to billions of smartphone users, a dramatic increase in dermatology productivity.
|
From page 69... ...
Noy and W Zhang, 2023, "Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence," Science 381:187–192, https://doi.org/10.1126/science.adh2586.
|
From page 70... ...
First, through automa tion and the resulting substitution of human labor, AI can increase labor productivity directly by reducing the worker hours required for a given amount of output. Second, AI can complement workers, making them more productive in their tasks.
|
From page 71... ...
Pierdomenico, 2023, "Global Economics Analyst: The Potentially Large Effects of Artificial Intelligence on Economic Growth," Goldman Sachs, March 26, https://www. gspublishing.com/content/research/en/reports/2023/03/27/d64e052b-0f6e-45d7-967b-d7be35fabd16.html.
|
From page 72... ...
23–57 in The Economics of Artificial Intelligence: An Agenda, National Bureau of Economic Research. 87 The transformative effect of AI is multifaceted, as noted in several recent articles -- for example, E
|
From page 73... ...
For instance, better tools for high-frequency trading or deep fakes for fooling people could be privately lucrative innovations that do not necessarily translate into faster productivity growth. More effort repackaging existing art and literature and less secure property rights for artists, writers, and inventors may also lead to less innovative output.
|
From page 74... ...
The answer is uncertain. Another question with an uncertain answer is whether businesses will deploy generative AI in "so-so" applications that reduce costs by replac ing labor without generating much productivity growth or improvements in the quality of service -- think self-checkout kiosks at stores as an example.97 What is certain, as this example indicates, is that the productivity effects of generative AI will depend on how it is used -- to create revenues, to reduce costs, or to enhance productivity.
|
From page 75... ...
. Although TFP is more difficult to measure at the firm level for other sectors, dispersion in labor productivity is rising across firms within industries in all sectors of the economy.98 Andrews and colleagues provide evidence of rising productivity dispersion within industries in many OECD countries.99 This increase in dispersion has not been well explained but may in part reflect growing digitization and information technology use among firms.100,101 Technological automation in recent decades has been skill or routine replacing, with the biggest impact on workers in routine middle-skill, middle-wage jobs.
|
From page 76... ...
Bureau of Labor Statistics over the next decade. If demographics and consumer demand increase employ ment in jobs in low-productivity sectors with low exposure to AI as a share of total employment, this will limit the economy-wide aggregate productivity effects of AI.
|
From page 77... ...
Artificial Intelligence and Productivity 77
|
From page 78... ...
Vanhaelen, and T.I. Oprea, 2020, "Will Artificial Intelligence for Drug Discovery Impact Clinical Pharmacology?
|
From page 79... ...
Globally, McKinsey Global Institute also estimates large contributions to productivity growth from these technologies. Worldwide, it predicts that they would add 0.2–3.3 percentage points to 2022–2040 and 0.1–0.6 percentage points from generative AI alone.
|
From page 80... ...
The gap between wage growth and productivity growth was larger for the median wage than for the aver age wage, reflecting growing wage inequality. And decoupling contributed to a drop in labor's share of national income to differing degrees in the advanced economies (Figure 3-11)
|
From page 81... ...
Economists have identified many factors behind the decline in the labor share of income and the decoupling of productivity growth and wage growth.122,123 These include both macroeconomic factors -- such as skill or routine-biased technological change, globalization, and the decline in workers covered by collective bargaining -- and micro economic factors -- such as increasing concentration in product and labor markets and growing differences in firms' productivity, profits, and wages.124,125 122 M
|
From page 82... ...
As discussed in the explanations of the productivity slowdown, structural changes imply headwinds to productivity growth. These include declining dynamism, rising concentration, rising market power, and rising markups of price over cost at the largest firms.
|
From page 83... ...
Artificial Intelligence and Productivity 83
|
From page 85... ...
Artificial Intelligence and Productivity 85
|
From page 86... ...
-wide employment and labor force participation rates were at their highest recorded levels, with half of all OECD countries exceeding previous high-water marks on both metrics.3 It is difficult to predict how unemployment rates may change in the years to come. The Congressional Budget Office projects that the U.S.
|
From page 87... ...
g=1oMKB. Artificial Intelligence and the Workforce 87
|
From page 88... ...
For example, see https://www.merriam-webster.com/dictionary/expertise, where expertise is defined as "the skill of an expert." 88 ARTIFICIAL INTELLIGENCE AND THE FUTURE OF WORK
|
From page 89... ...
. • Routine tasks, as used in economics literature, are physically or cognitively repetitive tasks that follow tightly scripted codifiable procedures.d For the purposes of this report, one can equate "mass expertise" with the ability to execute routine tasks.
|
From page 90... ...
If an unexpectedly urgent need arose for crossing guards, most air traffic controllers could presumably fill these roles. If an urgent need for air traffic controllers arose, the reverse would not be true.9 Jobs for which mass expertise suffices -- such as table-waiting, cleaning and janitorial services, manual labor, and (even)
|
From page 91... ...
Consider the example of software developers: Despite decades of sustained growth in investment in computer technology, slightly less than 1 percent of the U.S. workforce is currently employed in software development.13 If this fraction were to double to 2 percent, this would still comprise a smaller share of the workforce than is currently employed as fast food and counter workers.14 Before applying the expertise framework to assess the potential labor market impacts of AI, it is first used to interpret the labor market impacts of the two preceding technological revolutions: the Industrial Revolution and the computer revolution.
|
From page 92... ...
For example, the market value of taxi drivers' exhaustive and painstakingly acquired knowledge of the streets and alleys of London was diminished when GPS enabled ride-hailing apps made that expertise widely available through smartphones. Although there are currently as many London cabbies as ever, their earnings dropped by about 10 percent when Uber entered the market.15 Similarly, the roll-out of an AI-based taxi-routing program in Yokohama, Japan, erased the routing advantage of expert versus novice drivers, largely eliminating the value of expertise.16 In the foreseeable future, the job of air traffic control may be handled primarily by AI, potentially eroding the earn ings potential of air traffic controllers.
|
From page 93... ...
Many of the most highly paid jobs in industrialized economies -- oncologists, software engineers, patent lawyers, therapists, movie stars -- did not exist until specific technological or social innovations created a need for them. Prior to the era of air transport, for example, there was neither a market demand for nor supply of air traffic controller skills.
|
From page 95... ...
Dorn, and G.H. Hanson, 2013, "The China Syndrome: Local Labor Market Effects of Import Competition in the United States," American Economic Review 103(6)
|
From page 96... ...
Before turning to AI, it is instructive to consider two prior technological revolutions: the Industrial Revolution and the computer revolution. Table 4-1 provides an overview of the impact of each of these eras on the demand for expertise, and Table 4-2 provides a rubric for different types of expertise.
|
From page 97... ...
Elite expertise. Often requires a college era Expertise in learning Combining expert knowledge degree or significant post rules and mastering with acquired judgment to secondary education plus tools (i.e., carrying make high-stakes decisions in years of hands-on supervised out routine tasks)
|
From page 98... ...
This changed in the 18th and 19th centuries as industries mastered a new form of work organization that became known as mass production.30 Mass production involved breaking the complex work of artisans into discreet, self-contained, and often quite simple steps that could be carried out mechanistically by a team of production workers, often abetted by machinery and overseen by managers.31 As a case in point, the Ford Motor Company's River Rouge production plant, an archetype of mass production, employed more than 100,000 workers at its peak.32 The transition from artisanal to mass production profoundly changed the demand for worker expertise -- what expertise was needed, who supplied it, and what wages it commanded. Most directly, mass production reduced demand for artisanal labor by providing a faster, cheaper production system that combined high-tech machinery, managerial expertise, and vast numbers of comparatively unskilled workers.33 Although the skilled British weavers and textile workers who rose up in protest against mechaniza tion in the 19th century -- the eponymous Luddites -- are frequently derided for their naive fear of technology, these fears were not misplaced.
|
From page 99... ...
The demand for educated and highly trained workers rose across the board in maintenance, engineering, production infrastructure, product design, logistics, accounting, communications, sales, and management to coordinate these many sophisticated parts. Whereas mass production was initially expertise-displacing, relying primarily on cadres of untrained workers accomplishing rote tasks under brutal conditions, it ultimately generated demand for mass expertise.
|
From page 100... ...
But the transformative economic growth stemming from the Industrial Revolution allowed countries to focus their resources on these service activi ties, many of which would be considered nonessential in a poorer society. For example, while it would be a stretch to claim that the advent of mass production created the movie industry, neither the technology for producing and projecting movies nor the mass market consumer audience that was willing and able to pay for them would have been conceivable absent the rise in living standards that mass production afforded.39 This section thus far has addressed two of the three questions posed at the begin ning of this chapter: What expertise was replaced (artisanal expertise)
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.