The Economics of Artificial Intelligence

Home > Other > The Economics of Artificial Intelligence > Page 63
The Economics of Artificial Intelligence Page 63

by Ajay Agrawal


  wages have to fall or that other complementary factors like capital have to

  adjust enough for labor market equilibrium to be restored at or above the

  historic wage.

  14.5.1 Effi

  ciency Wage Theory and Nonadjustment of Wages

  The fi rst category of technological unemployment arises when wages do

  not adjust for structural reasons. Effi

  ciency wage theory emphasizes that

  productivity depends on wages and so employers may have reasons to pay

  wages above the market- clearing level. The original effi

  ciency wage paper

  (Stiglitz 1969) noted one of the reasons for this: that income disparities can

  weaken worker morale. Akerlof and Yellen (1990) have formalized this into

  the “fair wage hypothesis.”

  If fairness considerations are signifi cant enough, and workers think that a

  decrease in their wages is “unfair” (e.g., because the income of entrepreneurs

  increases so entrepreneurs could easily “aff ord” pay increases), it means that

  the scope of labor- saving progress that shifts the utility possibilities curve

  out without redistributions is very limited. Similar results hold if workers’

  well- being and eff orts are related to relative incomes. The new utility pos-

  sibilities curve may lie outside the old one to the “north” of E , that is, there 0

  is scope for a Pareto improvement in principle; but it may lie inside of the old

  utility possibilities curve near E , that is, the utility possibilities of workers 1

  decrease for a given level of utility of entrepreneurs because workers reduce

  their eff ort so much that the eff ective labor supply declines—any gains from

  technology are more than off set by increased shirking. Shapiro and Stiglitz

  (1984) emphasize that paying a wage above the market- clearing level reduces

  shirking, leading to unemployment.

  An even more daunting example of effi

  ciency wages may arise if automa-

  tion continues and the marginal product of labor for low- skill workers falls

  below their cost of living at what they view as their basic subsistence (even

  if they exert their best eff ort). Unless basic social services are provided to

  such workers, a nutritional effi

  ciency wage model applies in that case, similar

  to what Stiglitz (1976) described for developing countries: employers could

  not pay a market- clearing wage because they know that this would be insuf-

  fi cient for their employees to provide for themselves and remain productive.29

  We will follow up on this theme in the fi nal section of our chapter.

  In traditional effi

  ciency wage models, the unemployment eff ects of effi

  -

  29. Even worse outcomes could emerge in the presence of imperfect capital markets, if

  expenditures on health and nutrition at one date aff ect productivity at later dates.

  AI and Its Implications for Income Distribution and Unemployment 379

  ciency wages are permanent, part of the long- run equilibrium. For example,

  if technological change leads to greater inequality (or better information

  about the existing level of inequality), morale eff ects and the resulting

  effi

  ciency wage responses imply that the equilibrium level of unemploy-

  ment rises.

  However, effi

  ciency wage arguments may also contribute to slowing down

  the transition to a new equilibrium after an innovation, as we will explore

  subsection 14.5.2.

  Minimum Wages and Nonadjustment of Wages

  An alternative reason why wages may not adjust to the market- clearing

  level are minimum wage laws. Basic economics implies that there will be

  unemployment if wages are set to an excessive level. Although this is a

  theoretical possibility, recent experience in the United States has repeatedly

  shown that modest increases in minimum wages from current levels have

  hardly any employment eff ects but raise the income of minimum wage work-

  ers, which may have positive aggregate demand eff ects since low- income

  workers have a high marginal propensity to consume (see, e.g., Schmitt

  2013). From an economic theory perspective, these observations are possible

  because wages are not determined in a purely Walrasian manner—there is

  a signifi cant amount of bargaining involved when prospective employers

  and employees match—and increases in minimum wages substitute for the

  lacking bargaining power of workers (see, e.g., Manning 2011).

  14.5.2 Technological Unemployment as a Transition Phenomenon

  The second category of technological unemployment is as a transition

  phenomenon, that is, when technological change makes workers redundant

  at a faster pace than they can fi nd new jobs or that new jobs are created. This

  phenomenon was already observed by Keynes (1932). It is well understood

  that there is always a certain “natural” or “equilibrium” level of unemploy-

  ment as a result of churning in the labor market. In benchmark models of

  search and matching to characterize this equilibrium level of unemployment

  (see Mortensen and Pissarides 1994, 1998), employment relationships are

  separated at random, and workers and employers need to search for new

  matches to replace them. The random shocks in this framework can be

  viewed as capturing, in reduced form, phenomena such as life cycle transi-

  tions but also technological progress in individual fi rms. In this view, an

  increase in the pace of technological progress corresponds to a higher job

  separation rate and results in a higher equilibrium level of unemployment.

  The transition may be especially prolonged if technology implies that the

  old skills of workers become obsolete and they need to acquire new skills

  and/or fi nd out what new jobs match their skills (see, e.g., Restrepo 2015).

  Even if in the long run workers adjusted to AI, the transition may be

  diffi

  cult. Artifi cial intelligence will impact some sectors more than others,

  380 Anton Korinek and Joseph E. Stiglitz

  and there will be signifi cant job dislocation. As a general lesson, markets on

  their own are not good at structural transformation. Often, the pace of job

  destruction is greater than the pace of job creation, especially as a result of

  imperfections in capital markets, inhibiting the ability of entrepreneurs to

  exploit quickly new opportunities as they are opened up.

  The Great Depression as an Example of Transitional Unemployment

  The Great Depression can be viewed as being caused by rapid pace of

  innovation in agriculture (see Delli Gatti et al. 2012a). Fewer workers were

  needed to produce the food that individuals demanded, resulting in marked

  decline in agriculture prices and income, leading to a decline in demand for

  urban products. In the late 1920s, these eff ects became so large that long-

  standing migration patterns were reversed.

  What might have been a Pareto improvement turned out to be an immis-

  erizing technological change, as both those in the urban and rural sector

  suff ered.

  The general result is that noted earlier: with mobility frictions and rigidi-

  ties (themselves partly caused be capital market imperfections, as workers in

  t
he rural sector couldn’t obtain funds to obtain the human capital required

  in the urban sector and to relocate) technological change can be welfare

  decreasing. The economy can be caught, for an extensive period of time,

  in a low- level equilibrium trap, with high unemployment and low output.

  In the case of the Great Depression, government intervention (as a

  by-product of World War II) eventually enabled a successful structural

  transformation: the intervention was not only a Keynesian stimulus, but

  facilitated the move from rural farming areas to the cities where manu-

  facturing was occurring at the time and facilitated the retraining of the

  labor force, helping workers acquire the skills necessary for success in an

  urban manufacturing environment, which were quite diff erent from those

  that ensured success in a rural, farming environment. It was, in this sense,

  an example of a successful industrial policy.

  There are clear parallels to the situation today in that a signifi cant fraction

  of the workforce may not have the skills required to succeed in the age of AI.

  Transitional Effi

  ciency Wage Theory

  Effi

  ciency wage arguments may also slow down the transition to a new

  equilibrium after technological progress. For example, if worker morale

  depends on last period’s wages, it may be diffi

  cult to reduce wages to the

  market- clearing level after a labor- saving innovation, and unemployment

  may persist for a long time.30

  30. In the limiting case, employers may simply keep wages fi xed to avoid negative morale eff ects, and unemployment would persist forever—or until some off setting shock occurs.

  AI and Its Implications for Income Distribution and Unemployment 381

  14.5.3 Jobs and Meaning

  The potentially widespread destruction of jobs can have large human

  consequences that go beyond just economics because jobs provide not only

  income but also other mental services such as meaning, dignity, and fulfi ll-

  ment to humans. Whether this is a legacy of our past, and whether individu-

  als could fi nd meaning in other forms of activities, mental or physical, is a

  matter of philosophical debate.

  If workers derive a separate benefi t from work in the form of meaning,

  then job subsidies are a better way of ensuring that technological advances

  are welfare enhancing than simply providing lump sum grants (e.g., through

  the provision of a universal basic income), as some are suggesting in

  response to the inequalities created by AI.

  This discussion is, of course, a departure from the usual neoclassical for-

  mulation, where work only enters negatively into individual’s well- being.

  There are some that claim that individuals’ deriving dignity and meaning

  from work is an artifact of a world with labor scarcity. In a workerless AI

  world, individuals will have to get their identity and dignity elsewhere, for

  example, through spiritual or cultural values. The fact that most humans can

  fi nd a meaningful life after retirement perhaps suggests that there are good

  substitutes for jobs in providing meaning.

  14.6 Longer- Term Perspectives: AI and the Return of Malthus?

  There is a fi nal point that is worth discussing in a chapter on the implica-

  tions of artifi cial intelligence for inequality. This point relates to a somewhat

  longer- term perspective. Currently, artifi cial intelligence is at the stage where

  it strictly dominates human intelligence in a number of specifi c areas, for

  instance playing chess or Go, identifying patterns in x-rays, driving, and so

  forth. This is commonly termed narrow artifi cial intelligence. By contrast,

  humans are able to apply their intelligence across a wide range of domains.

  This capacity is termed general intelligence.

  If AI reaches and surpasses human levels of general intelligence, a set of

  radically diff erent considerations apply. Some techno- optimists predict the

  advent of general artifi cial intelligence for as early as 2029 (see Kurzweil

  2005), although the median estimate in the AI expert community is around

  2040 to 2050, with most AI experts assigning a 90 percent probability to

  human- level general artifi cial intelligence arising within the current century

  (see Bostrom 2014). A minority believes that general artifi cial intelligence

  will never arrive. However, if human- level artifi cial general intelligence is

  reached, there is broad agreement that AI would soon after become super-

  intelligent, that is, more intelligent than humans, since technological pro-

  gress would likely accelerate, aided by the intelligent machines. Given these

  382 Anton Korinek and Joseph E. Stiglitz

  predictions, we have to think seriously about the implications of artifi cial

  general intelligence for humanity and, in the context of this chapter, for what

  it implies for our economy as well as for inequality.

  Assuming that our social and economic system will be maintained upon

  the advent of artifi cial general intelligence and superintelligence,31 there

  are two main scenarios. One scenario is that man and machine will merge,

  that is, that humans will “enhance” themselves with ever more advanced

  technology so that their physical and mental capabilities are increasingly

  determined by the state of the art in technology and AI rather than by

  traditional human biology (see, e.g., Kurzweil, 2005). The second scenario

  is that artifi cially intelligent entities will develop separately from humans,

  with their own objectives and behavior (see, e.g., Bostrom 2014; Tegmark

  2017). As we will argue below, it is plausible that the two scenarios might

  diff er only in the short run.

  First Scenario: Human Enhancement and Inequality

  The scenario that humans will enhance themselves with machines may

  lead to massive increases in human inequality, unless policymakers recog-

  nize the threat and take steps to equalize access to human enhancement tech-

  nologies.32 Human intelligence is currently distributed within a fairly narrow

  range compared to the distance between the intelligence of humans and that

  of the next- closest species. If intelligence becomes a matter of ability to pay,

  it is conceivable that the wealthiest (enhanced) humans will become orders

  of magnitude more productive—“more intelligent”—than the unenhanced,

  leaving the majority of the population further and further behind. In fact,

  if intelligence enhancement becomes possible, then—unless preemptive

  actions are taken—it is diffi

  cult to imagine how to avoid such a dynamic.

  For those who can aff ord it, the incentive to purchase enhancements is great,

  especially since they are in competition with other wealthy humans who may

  otherwise leapfrog them. This is even more so in an economy which is, or

  is perceived to be, a winner- take- all economy and/or in which well- being

  is based on relative income. Those who cannot aff ord the latest technology

  will have to rely on what is in the public domain, and if the pace of innova-

  tion increases, the gap between the best technology and what is publicly

  available will increase.

  A useful analogy is to compare human enhancement tech
nology to health

  care—technology to maintain rather than enhance the human body. Dif-

  31. Researchers who work on the topic of AI safety point out that there is also a risk of doomsday scenarios in which a suffi

  ciently advanced artifi cial intelligence eradicates humanity

  because humans stand in the way of its goals. See, for example, Bostrom (2014) who elaborates on this using the example of a “paperclip maximizer”—an AI that has been programmed to produce as many paperclips as possible, without regard for other human goals, and who realizes that humans contain valuable raw materials that should better be transformed into paperclips.

  32. In many respects, the issues are parallel to those associated with performance- enhancing drugs. In sports, these have been strictly regulated, but in other arenas, they have not.

  AI and Its Implications for Income Distribution and Unemployment 383

  ferent countries have chosen signifi cantly diff erent models for how to provide

  access to health care, with some regarding it as a basic human right and

  others allocating it more according to ability to pay. In the United States, for

  example, the expected life spans of the poor and the wealthy have diverged

  signifi cantly in recent decades, in part because of unequal access to health

  care and ever more costly new technologies that are only available to those

  who can pay. The diff erences are even starker if we look at humanity across

  nations, with the expected life span in the richest countries being two- thirds

  longer than in the least developed countries (see, e.g., UN 2015). Like with

  health care, it is conceivable that diff erent societies will make signifi cantly

  diff erent choices about access to human enhancement technologies.

  Once the wealthiest enhanced humans have separated suffi

  ciently far from

  the unenhanced, they can eff ectively be considered as a separate species of

  artifi cially intelligent agents. To emphasize the diff erence in productivities,

  Yuval Harari (2017) has dubbed the two classes that may result “the gods”

  and “the useless.” In that case, the long- run implications of our fi rst scenario

  coincide with the second scenario.

  Second Scenario: Artifi cially Intelligent Agents and the Return of Malthus

  We thus turn to the scenario that artifi cially intelligent entities develop

  separately from regular (or unenhanced) humans. One of the likely char-

 

‹ Prev