by Ajay Agrawal
Bloom, Nicholas, Charles I. Jones, John Van Reenen, and Michael Webb. 2017. “Are
Ideas Getting Harder to Find?” Unpublished manuscript, Stanford University.
Bloom, Nicholas, Luis Garicano, Raff aella Sadun, and John Van Reenen. 2014. “The
Distinct Eff ects of Information Technology and Communication Technology on
Firm Organization.” Management Science 60 (12): 2859– 85.
Boppart, Timo. 2014. “Structural Change and the Kaldor Facts in a Growth Model
with Relative Price Eff ects and Non???Gorman Preferences.” Econometrica
82:2167– 96.
Chetty, Raj, John N. Friedman, Tore Olsen, and Luigi Pistaferri. 2011. “Adjustment
Costs, Firm Responses, and Micro vs. Macro Labor Supply Elasticities: Evidence
from Danish Tax Records.” Quarterly Journal of Economics 126 (2): 749– 804.
Comin, Diego, Danial Lashkari, and Marti Mestieri. 2015. “Structural Transfor-
mations with Long- Run Income and Price Eff ects.” Unpublished manuscript,
Dartmouth College.
Danaylov, Nikola. 2012. “17 Defi nitions of the Technological Singularity.”
Singularity Weblog. https:// www .singularityweblog .com/ 17-defi nitions- of-the
- technological- singularity/.
Elsby, Michael W. L., Bart Hobijn, and Ays¸egül S¸ahin. 2013. “The Decline of the
U.S. Labor Share.” Brookings Papers on Economic Activity 2013 (2): 1– 63.
Garicano, Luis. 2000. “Hierarchies and the Organization of Knowledge in Produc-
tion.” Journal of Political Economy 108 (5): 874– 904.
Good, I. J. 1965. “Speculations Concerning the First Ultraintelligent Machine.”
Advances in Computers 6: 31– 88.
Gordon, Robert J. 2012. “Is U.S. Economic Growth Over? Faltering Innovation Con-
fronts the Six Headwinds.” NBER Working Paper no. 18315, Cambridge, MA.
———. 2016. The Rise and Fall of American Growth: The US Standard of Living
since the Civil War. Princeton, NJ: Princeton University Press.
Graetz, Georg, and Guy Michaels. 2017. “Robots at Work.” Unpublished manu-
script, London School of Economics.
Grossman, Gene M., Elhanan Helpman, Ezra Oberfi eld, and Thomas Sampson.
Artifi cial Intelligence and Economic Growth 281
2017. “Balanced Growth Despite Uzawa.” American Economic Review 107 (4):
1293– 312.
Hart, Oliver, and Bengt Holmstrom. 2010. “A Theory of Firm Scope.” Quarterly
Journal of Economics 125 (2): 483– 513.
Helpman, Elhanan, and Manuel Trajtenberg. 1998. “A Time to Sow and a Time to
Reap: Growth Based on General Purpose Technologies.” In General Purpose Tech-
nologies and Economic Growth, edited by E. Helpman. Cambridge, MA: MIT Press.
Hemous, David, and Morten Olsen. 2016. “The Rise of the Machines: Automa-
tion, Horizontal Innovation and Income Inequality.” Unpublished manuscript,
University of Zurich.
Herrendorf, Berthold, Richard Rogerson, and Akos Valentinyi. 2014. “Growth and
Structural Transformation.” In Handbook of Economic Growth, vol. 2, 855– 941.
Amsterdam: Elsevier.
Jägger, Kirsten. 2016. “EU KLEMS Growth and Productivity Accounts 2016
release- Description of Methodology and General Notes.” The Conference Board
Europe.
Jones, Benjamin F. 2009. “The Burden of Knowledge and the Death of the Renais-
sance Man: Is Innovation Getting Harder?” Review of Economic Studies 76 (1):
283– 317.
Jones, Charles I. 1995. “R&D-Based Models of Economic Growth.” Journal of
Political Economy 103 (4): 759– 84.
———. 2016. “The Facts of Economic Growth.” In Handbook of Macroeconomics,
vol. 2, 3– 69 Amsterdam: Elselvier.
Jorgenson, Dale W., Mun S. Ho, and Jon D. Samuels. Forthcoming. “Educational
Attainment and the Revival of U.S. Economic Growth.” Education, Skills, and
Technical Change: Implications for Future US GDP Growth, edited by Charles
Hulten and Valerie Ramey. Chicago: University of Chicago Press.
Kaldor, Nicholas. 1961. “Capital Accumulation and Economic Growth.” In The
Theory of Capital, edited by F. A. Lutz and D. C. Hague, 177– 222. New York:
St. Martins Press.
Karabarbounis, Loukas, and Brent Neiman. 2013. “The Global Decline of the Labor
Share.” Quarterly Journal of Economics 129 (1): 61– 103.
Kehrig, Matthias, and Nicolas Vincent. 2017. “Growing Productivity without Grow-
ing Wages: The Micro- Level Anatomy of the Aggregate Labor Share Decline.”
Unpublished manuscript, Duke University.
Kleven, Henrik J., and Mazhar Waseem. 2013. “Using Notches to Uncover Optimi-
zation Frictions and Structural Elasticities: Theory and Evidence from Pakistan.”
Quarterly Journal of Economics 128 (2): 669– 723.
Kongsamut, Piyabha, Sergio Rebelo, and Danyang Xie. 2001. “Beyond Balanced
Growth.” Review of Economic Studies 68 (4): 869– 82.
Kortum, Samuel S. 1997. “Research, Patenting, and Technological Change.” Econo-
metrica 65 (6): 1389– 419.
Krusell, Per, Lee E. Ohanian, José- Víctor Ríos- Rull, and Giovanni L. Violante. 2000.
“Capital- Skill Complementarity and Inequality: A Macroeconomic Analysis.”
Econometrica 68 (5): 1029– 53.
Kurzweil, Ray. 2005. The Singularity is Near. New York: Penguin.
Legg, Shane, and Marcus Hutter. 2007. “A Collection of Defi nitions of Intelligence.”
Frontiers in Artifi cial Intelligence and Application 157 (2007): 17– 24.
Manuelli, Rodolfo E., and Ananth Seshadri. 2014. “Frictionless Technology Diff u-
sion: The Case of Tractors.” American Economic Review 104 (4): 1368– 91.
Ngai, L. Rachel, and Christopher A. Pissarides. 2007. “Structural Change in a Multi-
sector Model of Growth.” American Economic Review 97 (1): 429– 43.
282 Patrick Francois
Nordhaus, William D. 2015. “Are We Approaching an Economic Singularity? Infor-
mation Technology and the Future of Economic Growth.” NBER Working Paper
no. 21547, Cambridge, MA.
Peretto, Pietro F., and John J. Seater. 2013. “Factor- Eliminating Technical Change.”
Journal of Monetary Economics 60 (4): 459– 73.
Romer, Paul M. 1990. “Endogenous Technological Change.” Journal of Political
Economy 98 (5): S71– 102.
Saez, Emmanuel. 2010. “Do Taxpayers Bunch at Kink Points?” American Economic
Journal: Economic Policy 2 (3): 180– 212.
Solomonoff , R. J. 1985. “The Time Scale of Artifi cial Intelligence: Refl ections on
Social Eff ects.” Human Systems Management 5:149– 53.
Stole, Lars, and Jeff rey Zwiebel. 1996. “Organizational Design and Technology
Choice under Intrafi rm Bargaining.” 86 (1): 195– 222.
Tirole, Jean. 2017. Economics for the Common Good. Princeton, NJ: Princeton University Press.
Vinge, Vernor. 1993. “The Coming Technological Singularity: How to Survive in
the Post- Human Era.” In Vision- 21: Interdisciplinary Science and Engineering in
the Era of Cyberspace, 11– 22. Proceedings of a Symposium Coauthored by the
NASA Lewis Research Center and the Ohio Aerospace Institute Held in Westlake,
Ohio, Mar. 30– 31.
Webb, Michael, Greg Thornton, Sean Legassick, and Mustafa Suleyman. 2017.
“What Does Artifi cial Intelligence Do?” Unpublished manuscript, Stanford Uni-
versit
y.
Weitzman, Martin L. 1998. “Recombinant Growth.” Quarterly Journal of Econom-
ics 113:331– 60.
Yudkowsky, Eliezer. 2013. “Intelligence Explosion Microeconomics.” Technical
Report no. 2013– 1, Machine Intelligence Research Institution.
Zeira, Joseph. 1998. “Workers, Machines, and Economic Growth.” Quarterly Journal
of Economics 113 (4): 1091– 117.
Comment Patrick Francois
The political economy of artifi cial intelligence (AI) was not included as a
topic in this conference, but political economy arose in a number of conver-
sations, including my discussion of this immensely thought- provoking chap-
ter. So I want to discuss it further here. It is important for two reasons. One,
if the scientists’ predictions pan out, we are on the cusp of a world where
humans will be largely redundant as an economic input. How we manage the
relationship between the haves (who own the key inputs) and the have- nots
(who only own labor) is going to be a key aspect of societal health. Successful
ones will be inclusive in the sense of sharing rents owned by the haves with
the have- nots. This is quite obvious. Less obviously, I am going to argue that
Patrick Francois is a professor at the Vancouver School of Economics of the University of British Columbia and a senior fellow at the Canadian Institute for Advanced Research.
For acknowledgments, sources of research support, and disclosure of the author’s material fi nancial relationships, if any, please see http:// www .nber .org/ chapters/ c14028.ack.
Comment 283
managing the relationship between high- level human decision- making and
our machines servants will involve humans at many levels, no matter how
productive machines become. So, even in the limit where machines become
better at doing all human production, there will still be work for humans in
what could be broadly referred to as the political realm.
The chapter of Philippe Aghion, Benjamin Jones, and Charles Jones is
a great starting point for the less structured discussion that I am about to
set off on here. The chapter explores the growth implications of AI, where
the aspect focused on is the increasing automation of production. That is,
machines replacing labor at a continually increasing range of production,
service, and creative tasks. Automation in this form is not new and has been
going on since at least the Industrial Revolution. So any model written down
projecting what will/ might happen should not run afoul of the basic Kal-
dor facts. Accordingly, they build a model able to deliver a relatively stable
labor share despite the continual displacement of labor from an increasing
number of sectors.
In a nutshell this works as follows: with multiple sectors and low enough
substitutability across the goods produced in them, consumers spend pro-
gressively more of real wealth on sectors not subject to automation. This
leads to a protracted relative price increase of nonautomated goods’ sectors.
So two counteracting forces generate a force toward relative stability of
the labor share in their model: (a), labor is usefully employed in fewer sec-
tors—lowering its factor share; but (b), in the sectors where labor continues
to work, relative prices are increasing—tending to raise the factor share.
Essentially, though progressively fewer things remain useful for humans
to do, these things become relatively well remunerated, and this can con-
tinue provided there remain some things that humans can do better than
machines.
But it is when we turn to thinking about what are the products or ser-
vices where humans will remain essential in production that we start to run
into problems. What if humans cannot do anything better than machines?
Many discussions at the conference centered around this very possibility.
And I must admit that I found the scientists’ views compelling on this.
Though it has been the case that new services, which have been relatively
labor intensive, have emerged as technology has mechanized the production
of goods and services, and this has been demonstrated by others (Acemo-
glu and Restrepo 2016) to be another force that could stabilize the labor
share. Even with this, the complete displacement of labor from production
of goods and service will arise if machines dominate humans in the perfor-
mance of all tasks.
Scientists disagree on how imminent this eventuality is, but few doubt that
it will eventually occur. Though it may well be a limiting case reached only
many generations down the track, from now on I will try to imagine what
will happen in that limiting case. The one where machines can do everything
284 Patrick Francois
better than humans. The point I wish to make is that even in such a world
where machines are better at all tasks, there will still be an important role
for human “work.” And that work will involve what will become the almost
political task of managing the machines.
The Political Economic Challenges That
Machine- Superior Societies Will Face
But before I turn to that, a fi rst challenge societies will face in a completely
machine- superior world is: Who owns the machines? Capitalist societies
succeed when they create incentives for investment. They reward innova-
tors who come up with and implement good ideas, and thus encourage
those ideas. Societies with the features that are well suited to pioneering the
advance of machines today are also the economically successful societies,
and generally the most healthy societies socially. Incentives for technological
advance are greatest where property rights are best protected, and where the
taxes on the successful are the lowest. So we predictably see the vanguard of
this new world of machine superiority emerging from the most successful
capitalist economies like the United States of America.
But everything changes when the machines reach the point of displacing
human inputs in the task of innovation, what Aghion, Jones, and Jones
term “AI in the idea production function.” Here I’m again talking about
the extreme case where machines do all of their own innovation much bet-
ter than people, and without requiring any human input. At this point, the
decisions on how to best improve the current technology, the risks to take,
the directions to follow, and the implementation are all done by machines.
Machines then improve themselves and enter in to a process of creating new
and better machines without the need for human intervention.
Aghion, Jones, and Jones developed a fantastically interesting analysis
of the almost science fi ction- like possibility of singularities and productive
extremes that can arise in that stage. I am going to, alternatively, focus on
the political economic implications.
Presumably, at least at the start of this period, the human owners of
these machines made improvements (and the stream of rents that those
improvements generate)that are well identifi able. These are the owners of
the machines that did the previous round of inventing. Similarly, as the next
generation of improvements emerge, the machines that were earlier invented
>
by the previous machines can be traced back to a primal machine inventor(s)
with well- identifi ed human inventor/ owners, and so on. In a sense then,
this last generation of human inventor/ owners will have a claim to the rents
generated by the machines from then on.
Should we, as a society, recognize that claim? The answer to that depends
on where individuals, the political elites, and the economic elites in that
society stand on the issue of inviolability of private property. At the point
Comment 285
where machines become self- inventing, redistributing the ownership rents
to all individuals in society will come without cost in terms of future growth
because human incentives no longer play a role. This won’t be easy for many
of today’s successful societies to do.
The social cost of not doing this will be human unrest on a massive scale.
The degree of inequality in a society where the owners of the machines
are the last generation of human/ inventor/ investors and the rest of society
earns their incomes from labor will be extreme. Nationalizing ownership
of the machines will be costless in terms of future growth, but the elite who
own the machines may be (and if history is any guide, will be) extremely
reluctant to give up their “hard- earned” rents, and their power, to the pas-
sive majority who did not have the foresight, hard work, and luck, to come
up with these machines. The societies that will be most functional in this
future will be those most willing to tax this last generation of productive
inventor/ investors to support the unlucky, less able, and perhaps even will-
ingly slothful, who do not own a machine. Countries that, for the very
reason of not heavily taxing innovation today will be in the vanguard of
creating our technofuture, may have social values that will tend to make
them somewhat poorly placed to manage it.
If the elite of such countries succeed in managing to control the political
channels whereby rival elites may come to threaten them, or where the
excluded masses who do not share ownership of the machines would be
able to coordinate against them, they will be able to enjoy machine rents and
become almost infi nitely richer than the excluded. The autocratic elites of
the Soviet Union employed just such methods of exclusion and disruption