Book Read Free

The Economics of Artificial Intelligence

Page 32

by Ajay Agrawal


  1. education: search for ways to provide for the changing nature of skills

  required for the AI era;

  2. personal services: these are the fastest- growing occupations, but as

  defi ned at present cannot benefi t from AI; and

  3. direction of technical change: strive to human- enhancing innovations,

  not human replacing.

  6.3.1 Education: The Upcoming Revolution

  As already mentioned, the expectation is that AI will become the domi-

  nant GPT of the coming era, spreading throughout the economy, and

  180 Manuel Trajtenberg

  displacing in the process a great many occupations. At the same time, the

  remaining occupations and new ones that may spring up as complementary

  to AI will require a new set of skills that are not quite those currently pro-

  vided by the education system, at all levels.

  This is not new: the fi rst and second industrial revolutions in the course

  of the nineteenth century required, and were accompanied by, correspond-

  ing revolutions in education. The need to rely on a more skilled, educated

  workforce, as well as a more disciplined one, fed educational reforms fi rst in

  Prussia (already in the late eighteenth century), then in the United Kingdom

  and in the United States, that led gradually to the institutionalization of

  free and universal education, with highly structured, government- set cur-

  riculums.

  From the late nineteenth century to this day, this “factory model” of edu-

  cation spread widely, expanding quantitatively in all dimensions: more hours

  spent at school, more subjects covered, and more years of study. Thus, for

  example, the average years of schooling in the UK adult population was

  less than 1 in 1870, whereas at present it stands at over 13. Universal educa-

  tion now starts at age three to four in many countries, high school became

  compulsory in the second half of the twentieth century, and in the past three

  decades some form of tertiary education has become commonplace.

  It is now widely accepted that this “factory model” needs to be revised and

  perhaps totally revamped in view of twin pervasive developments: fi rst, the

  internet revolution, which in this context means the availability of informa-

  tion/ knowledge on any subject, at all times and virtually at no cost; second,

  the rapidly changing requirements for meaningful employment.

  In particular, the advent of AI as the new GPT, with its expected perva-

  sive impact on employment, may call for a new education revolution, very

  much like the industrial revolutions of the nineteenth century. The key to it

  appears to be the shift away from imparting knowledge per se, to developing

  skills relevant for an AI- based economy. Likewise, such educational revo-

  lution will in all likelihood aim toward “personalized education,” departing

  from the quest for uniformity that has characterized education systems ever

  since Prussian reforms.

  What are likely to be the top skills required for employment in the upcom-

  ing AI era? There is a great deal of heated discussion in this area, but some

  agreement is emerging around a core set of skills, such as those listed in

  table 6.1.

  There is a great deal of similarity between these three lists of skills, and

  in fact they can be classifi ed into the following (nonexhaustive) main types:

  • Type I: analytical, creative, adaptive

  • critical and creative thinking

  • analytical and research

  • sense- making

  AI as the Next GPT: A Political- Economy Perspective 181

  Table 6.1

  Skills sought for employment (from websites)

  UNICEF 10 life skills

  MyStartJob .com

  Top10onlinecolleges .org

  1. Problem- solving

  1. Communication skills

  1. Sense- making

  2. Critical thinking

  2. Analytical and research

  2. Social intelligence

  3. Eff ective communication

  3. Flexibility- adaptability

  3. Novel adaptive thinking

  4. Decision- making

  4. Interpersonal abilities

  4. Cross- cultural competency

  5. Creative thinking

  5. Decision- making

  5. Computational thinking

  6. Interpersonal relationships

  6. Plan, organize, prioritize

  6. New media literacy

  7. Self- awareness

  7. Wear multiple hats

  7. Transdisciplinary

  8. Empathy

  8. Leadership/ management

  8. Design mind- set

  9. Coping with stress

  9. Attention to detail

  9. Manage cognitive load

  10. Coping with emotions

  10. Self- confi dence

  10. Virtual collaboration

  • novel adaptive thinking

  • design mind- set

  • Type II: interpersonal, communication

  • eff ective communication

  • interpersonal relationships/ abilities

  • social intelligence

  • virtual collaboration

  • Type III: emotional, self- confi dence

  • self- awareness

  • empathy

  • coping with stress

  • manage cognitive load

  • coping with emotions

  The important point to notice is that most of these skills are neither

  imparted in the current K– 12 system, nor in academia. The whole system is

  still geared primarily toward the transmission of knowledge, highly struc-

  tured and uniform, and not toward skills, let alone those skills. Pupils of

  all ages are now very aware of the fact that school- like information is avail-

  able at the tip of their fi ngers, they are less receptive to frontal classes, their

  attention span is much shorter, and the sort of stimuli that makes them tick

  is diff erent. This is also true at the tertiary level, and in addition, we are

  witnessing there the rise of the massive open online courses (MOOCS) and

  of other such online- based teaching tools.

  In view of these trends, educational strategies may need to undergo

  equally signifi cant changes away from the “factory system,” and the fact that

  the incipient GPT may render many existing occupations obsolete, provides

  it with renewed urgency. These are some of the issues to tackle:

  • Invert the pyramid: it is now widely recognized that critical skills, hard

  and soft, cognitive and social, are acquired very early on. Furthermore,

  182 Manuel Trajtenberg

  failure to do so at the earliest stages may be hard (even impossible) to

  remedy later on (see, e.g., Heckman et al. 2014). Thus, we may have to

  consider investing much more in early childhood education, from birth

  to age six.

  • Find ways to incorporate the development of skills (of the three types

  sketched above) as an integral part of teaching in every discipline and

  at all stages, including in academia.

  • Eff ective educational methods are hard to come by, thus it is important

  to engage in bottom-up experimentation in pedagogy, school design,

  and social skills development in the context of fl exible, creative, teach-

  ing environments.

  • Reconsider the prevalent norm of u
niform (typically government-

  mandated) curriculums and educational models, vis à vis diversity and

  open- innovation communities built around educational institutions.

  • Foster research on the eff ectiveness of new educational models, their

  adequacy to shifting needs, the extent to which they promote equal

  opportunity, and so forth. This type of research will be crucial given the

  move away from “top- down” models and the emphasis on widespread

  experimentation.

  6.3.2 Upgrading Personal Services

  A Bureau of Labor Statistics (BLS) study4 projects that virtually all of the

  employment gains in the decade to 2024 will be in services, and within the

  service sector particularly in health care and social assistance (see table 6.2).

  Many of these occupations as performed today require little training and

  minimal educational attainment. Not surprisingly, most confer low wages,

  low status, and are supported by very little complementary technology. As

  the projections suggest, those occupations are at present not seriously threat-

  ened by AI—on the contrary, they will grow signifi cantly. Thus, the overall

  prospects look rather gloomy when not only employment is considered by

  also wages: major upscale occupations are projected to remain stagnant or

  decline, whereas low- scale occupations are expected to grow.

  Is this a deterministic outcome? Not necessarily, and the case of nursing

  may be quite instructive. After World War II, nursing was one of the lowest-

  ranking occupations in the United States: in 1946 the average wage of a

  nurse was just one- third that of female workers in the garment industry.5

  In 1964 Congress passed the Nurse Training Act, which essentially rede-

  fi ned the occupation and turned it into a profession requiring an academic

  4. See: Occupational Employment Projections to 2024, Monthly Labor Review, US Bureau

  of Labor Statistics, Dec. 2015. Also in https:// www .bls .gov/ opub/ mlr/ 2015/ article/ occupational

  - employment- projections- to-2024 .htm.

  5. In 1946, the average registered nurse (RN) earned about one dollar an hour—or $175 a month.

  AI as the Next GPT: A Political- Economy Perspective 183

  Table 6.2

  US employment by major sector (millions)

  Percentage

  Sector

  2014 2024* Change*

  growth*

  Goods producing

  19

  19

  ~

  ~ 0

  Services

  121

  130

  + 9.3

  + 7

  Of which: health care and social assistance

  18

  22

  + 3.8

  + 20

  Other

  10

  11

  + 0.5

  + 1

  Total

  151

  160

  + 9.8

  + 6

  *Forecast

  degree, with an upgraded curriculum. Since then the nursing profession has

  risen in every dimension—salaries, status, academic requirements, range

  of responsibilities, and so forth. These days, the nursing profession spans

  a range of specializations, whereby the upper echelon commands annual

  wages as high as $100,000. Moreover, nurses now use advanced technolo-

  gies, and these in turn contribute to upgrade the profession.

  It could have been otherwise had it not been for the legislation of 1964,

  and so it is for other occupations in personal services. Thus, we need to

  consider proactive strategies for the professionalization of personal services, particularly in health care and education, setting standards and academic

  requirements.

  Take for example early childhood education: in most countries there are

  virtually no such standards for caregivers of children age one to three, pre-

  cisely the ages that are crucial for their development. Suppose now that

  they were required to have specialized academic degrees, with a curriculum

  that would include psychology, brain development, testing for learning dis-

  abilities, and so forth. Not only would the status and wages of these workers

  increase, but they would be much more likely to benefi t from complementary

  advanced technologies.

  The advent of AI would probably not threaten these growing occupations,

  and furthermore, if they were upgraded in the way just described, AI could

  bestow signifi cant benefi ts to them as well. For that to happen smart inter-

  faces between the practitioners of these occupations and the AI machines

  will have to be developed. Thus, imagine, for example, professional care-

  givers using AI to test very young children for learning disabilities, and then

  for treating them with specially tailored AI- based games.

  To sum up: BLS projections indicate that the bulk of job creation in the

  decade to 2024 will be in personal services, particularly in personal care.

  As currently practiced, most of these occupations are at the low end of the

  scale and rather impervious to technological advances. However, there are

  viable options to upgrade these occupations, particularly by setting aca-

  demic standards and advanced curriculums. If that were to happen, then

  184 Manuel Trajtenberg

  the changing composition of employment (i.e., more personal care, less of

  many others) would not adversely aff ect income distribution but perhaps

  to the contrary; furthermore, and more importantly here, AI may play a

  complementary role vis à vis these occupations, thus raising productivity in

  services and triggering a virtuous cycle.

  6.3.3 The Direction of Technical Change: H- Enhancing or H- Replacing?

  Although one of the seminal volumes in the economics of technological

  change is titled “The Rate and Direction of Inventive Activity,” in fact the

  economic discipline has traditionally dealt much more with the “rate” than

  with the “direction.” That may come as no surprise, since discussing direc-

  tion requires getting into the guts of technology itself, and there is no reason

  to believe that economists have a comparative advantage in that regard.

  Nevertheless, the extent and scope of technological advances that engulf

  us may require us to look more closely into the “black box” and try to under-

  stand, at the very least, what types of innovations we are facing and how they

  impact the economy. Furthermore, we would like to know whether there is

  room to aff ect the relative prevalence of the various types, in view of their

  diff erential economic eff ects.

  Here is such an attempt: consider on the one hand innovations that mostly

  magnify, enhance, and extend sensory, motoric, analytical, and other human

  capabilities such as:

  • In medicine: AI for diagnostics, for example, for reading and inter-

  preting x-rays, CT scans and other imaging modalities; AI for robotic

  surgery (e.g., the da Vinci robot for prostate surgery); AI data mining

  of electronic medical records for follow-up evaluations of drug effi

  cacy

  post- Food and Drug Administriation (FDA) approval, and so forth.

  • In education: AI- based methods for “personalized teaching”; AI for

  online testing in MOOCS; (see also the above- mentioned applications


  for early childhood education), and so forth.

  We label these “human- enhancing innovations” (HEI)—in medicine they

  do not replace doctors, but rather augment their human- bound capabilities

  (think of the precision and consistency of robot surgery), thus making better

  doctors. Similarly for teachers, eventually for judges (ruling with the aid of

  AI- based analysis), and so forth.

  On the other hand, consider “human- replacing innovations” (HRI), that

  is, technical advances that replace human intervention, and furthermore that

  often leave for humans mostly “dumb” jobs that are not worth yet replacing

  given the very low wages that they command (and often are indeed diffi

  cult

  to replicate by machines, the proverbial one being janitors).

  Some HRIs lead to cutting- edge, virtually human- free factories (best

  exemplifi ed by Tesla’s new facilities to produce batteries for its e-cars) that

  AI as the Next GPT: A Political- Economy Perspective 185

  greatly improve productivity, even if reducing employment. Consider, how-

  ever, the polar case of Walmart, the world’s largest private employer (with

  over two million employees), having deployed advanced technologies along

  its whole chain of operations from logistics to retailing; it has turned a large

  proportion of its workers into “unthinking automatons,” commanding very

  low wages with no prospect for improvement.

  These then are two types of innovations (HEI and HRI) that have very

  diff erent eff ects on key economic and social variables. It would seem that

  AI- based HEIs have the potential to unleash a new wave of human creativity

  and productivity, particularly in services (which to repeat are expected to be

  the fastest- growing occupations), whereas HRIs either decrease employ-

  ment (e.g., Tesla), or create unworthy jobs.

  Is it possible to design strategies to aff ect the direction of technical change

  in the sense of stimulating HEIs versus HRIs? It is hard to say, but it is cer-

  tainly worthwhile investigating such possibility given the large impact that

  a change in direction may have on the economy. Incidentally, it would seem

  that in any case the traditional emphasis of economic policy on the “rate”

  of innovation, that is, on how much resources we devote to research and

  development (R&D), is misplaced—worldwide competition may be pushing

  us into too much investment in R&D, not too little (too many patents, too

 

‹ Prev