by Ajay Agrawal
Intelligence on Innovation
An Exploratory Analysis
Iain M. Cockburn, Rebecca Henderson, and Scott Stern
4.1 Introduction
Rapid advances in the fi eld of artifi cial intelligence have profound implica-
tions for the economy as well as society at large. These innovations have the
potential to directly infl uence both the production and the characteristics of
a wide range of products and services, with important implications for pro-
ductivity, employment, and competition. But, as important as these eff ects
are likely to be, artifi cial intelligence also has the potential to change the
innovation process itself, with consequences that may be equally profound,
and which may, over time, come to dominate the direct eff ect.
Consider the case of Atomwise, a start-up fi rm that is developing novel
technology for identifying potential drug candidates (and insecticides) by
using neural networks to predict the bioactivity of candidate molecules. The
company reports that its deep convolutional neural networks “far surpass”
the performance of conventional “docking” algorithms. After appropri-
ate training on vast quantities of data, the company’s AtomNet product
is described as being able to “recognize” foundational building blocks of
Iain M. Cockburn is the Richard C. Shipley Professor of Management at Boston University and a research associate of the National Bureau of Economic Research. Rebecca Henderson is the John and Natty McArthur University Professor at Harvard University, where she has a joint appointment at the Harvard Business School in the General Management and Strategy units, and a research associate of the National Bureau of Economic Research. Scott Stern is the David Sarnoff Professor of Management and chair of the Technological Innovation, Entrepreneurship, and Strategic Management Group at the MIT Sloan School of Management, and a research associate and director of the Innovation Policy Working Group at the National Bureau of Economic Research.
We thank Michael Kearney for extraordinary research assistance. For acknowledgments,
sources of research support, and disclosure of the authors’ material fi nancial relationships, if any, please see http:// www .nber .org/ chapters/ c14006.ack.
115
116 Iain M. Cockburn, Rebecca Henderson, and Scott Stern
organic chemistry, and is capable of generating highly accurate predictions
of the outcomes of real- world physical experiments (Wallach, Dzamba,
and Heifels 2015). Such breakthroughs hold out the prospect of substantial
improvements in the productivity of early stage drug screening. Of course,
Atomwise’s technology (and that of other companies leveraging artifi cial
intelligence to advance drug discovery or medical diagnosis) is still at an
early stage: though their initial results seem to be promising, no new drugs
have actually come to market using these new approaches. But whether or
not Atomwise delivers fully on its promise, its technology is representa-
tive of the ongoing attempt to develop a new innovation “playbook,” one
that leverages large data sets and learning algorithms to engage in precise
prediction of biological phenomena in order to guide design- eff ective inter-
ventions. Atomwise, for example, is now deploying this approach to the
discovery and development of new pesticides and agents for controlling
crop diseases.
Atomwise’s example illustrates two of the ways in which advances in arti-
fi cial intelligence have the potential to impact innovation. First, though the
origins of artifi cial intelligence are broadly in the fi eld of computer science,
and its early commercial applications have been in relatively narrow domains
such as robotics, the learning algorithms that are now being developed sug-
gest that artifi cial intelligence may ultimately have applications across a very
wide range. From the perspective of the economics of innovation (among
others, Bresnahan and Trajtenberg 1995), there is an important distinction
between the problem of providing innovation incentives to develop tech-
nologies with a relatively narrow domain of application, such as robots
purpose- built for narrow tasks, versus technologies with a wide—advocates
might say almost limitless—domain of application, as may be true of the
advances in neural networks and machine learning often referred to as “deep
learning.” As such, a fi rst question to be asked is the degree to which devel-
opments in artifi cial intelligence are not simply examples of new technolo-
gies, but rather may be the kinds of “general purpose technologies” (GPTs)
that have historically been such infl uential drivers of long- term technologi-
cal progress.
Second, while some applications of artifi cial intelligence will surely consti-
tute lower- cost or higher- quality inputs into many existing production pro-
cesses (spurring concerns about the potential for large job displacements),
others, such as deep learning, hold out the prospect of not only productivity
gains across a wide variety of sectors, but also changes in the very nature
of the innovation process within those domains. As articulated famously
by Griliches (1957), by enabling innovation across many applications,
the “invention of a method of invention” has the potential to have much
larger economic impact than development of any single new product. Here
we argue that recent advances in machine learning and neural networks,
through their ability to improve both the performance of end- use technolo-
The Impact of Artifi cial Intelligence on Innovation 117
gies and the nature of the innovation process, are likely to have a particularly
large impact on innovation and growth. Thus the incentives and obstacles
that may shape the development and diff usion of these technologies are an
important topic for economic research, and building an understanding of
the conditions under which diff erent potential innovators are able to gain
access to these tools and to use them in a procompetitive way is a central
concern for policy.
This chapter begins to unpack the potential impact of advances in arti-
fi cial intelligence on innovation, and to identify the role that policy and
institutions might play in providing eff ective incentives for innovation, dif-
fusion, and competition in this area. We begin in section 4.2 by highlighting
the distinctive economics of research tools, of which deep learning applied
to research and development (R&D) problems is such an intriguing example.
We focus on the interplay between the degree of generality of application
of a new research tool and the role of research tools not simply in enhanc-
ing the effi
ciency of research activity, but in creating a new “playbook” for
innovation itself. We then turn in section 4.3 to briefl y contrast three key
technological trajectories within artifi cial intelligence (AI)—robotics, sym-
bolic systems, and deep learning. We propose that these often confl ated fi elds
will likely play very diff erent roles in the future of innovation and techni-
cal change. Work in symbolic systems appears to have stalled and is likely
to have relatively little impact going forward. And while developments in
robotics have the potential t
o further displace human labor in the production
of many goods and services, innovation in robotics technologies per se has
relatively low potential to change the nature of innovation itself. By contrast,
deep learning seems to be an area of research that is highly general purpose
and has the potential to change the innovation process itself.
We explore whether this might indeed be the case through an examina-
tion of some quantitative empirical evidence on the evolution of diff erent
areas of artifi cial intelligence in terms of scientifi c and technical outputs
of AI researchers as measured (imperfectly) by the publication of papers
and patents from 1990 through 2015. In particular, we develop what we
believe is the fi rst systematic database that captures the corpus of scientifi c
paper and patenting activity in artifi cial intelligence, broadly defi ned, and
divides these outputs into those associated with robotics, symbolic systems,
and deep learning. Though preliminary in nature (and inherently imperfect
given that key elements of research activity in artifi cial intelligence may
not be observable using these traditional innovation metrics), we fi nd strik-
ing evidence for a rapid and meaningful shift in the application orientation
of learning- oriented publications, particularly after 2009. The timing of
this shift is informative, since it accords with qualitative evidence about the
surprisingly strong performance of so-called “deep learning” multilayered
neural networks in a range of tasks including computer vision and other
prediction tasks.
118 Iain M. Cockburn, Rebecca Henderson, and Scott Stern
Supplementary evidence (not reported here) based on the citation pat-
terns to authors such as Geoff rey Hinton, who are leading fi gures in deep
learning, suggests a striking acceleration of work in just the last few years
that builds on a small number of algorithmic breakthroughs related to multi-
layered neural networks.
Though not a central aspect of the analysis for this chapter, we further fi nd
that, whereas research on learning- oriented algorithms has had a slow and
steady upward swing outside of the United States, US researchers have had
a less sustained commitment to learning- oriented research prior to 2009,
and have been in a “catch-up” mode ever since.
Finally, we begin to explore some of the organizational, institutional,
and policy consequences of our analysis. We see machine learning as the
“invention of a method of invention” whose application depends, in each
case, on having access not just to the underlying algorithms, but also to
large, granular data sets on physical and social behavior. Developments in
neural networks and machine learning thus raise the question of, even if the
underlying scientifi c approaches (i.e., the basic multilayered neural networks
algorithms) are open, prospects for continued progress in this fi eld—and
commercial applications thereof—are likely to be signifi cantly impacted by
terms of access to complementary data. Specifi cally, if there are increasing
returns to scale or scope in data acquisition (there is more learning to be
had from the larger data set), it is possible that early or aggressive entrants
into a particular application area may be able to create a substantial and
long- lasting competitive advantage over potential rivals merely through
the control over data rather than through formal intellectual property or
demand- side network eff ects. Strong incentives to maintain data privately
has the additional potential downside that data is not being shared across
researchers, thus reducing the ability of all researchers to access an even
larger set of data that would arise from public aggregation. As the competi-
tive advantage of incumbents is reinforced, the power of new entrants to
drive technological change may be weakened. Though this is an important
possibility, it is also the case that, at least so far, there seems to be a signifi cant amount of entry and experimentation across most key application sectors.
4.2 The Economics of New Research Tools: The Interplay between
New Methods of Invention and the Generality of Innovation
At least since Arrow (1962) and Nelson (1959), economists have appreci-
ated the potential for signifi cant underinvestment in research, particularly
basic research or domains of invention with low appropriability for the
inventor. Considerable insight has been gained into the conditions under
which the incentives for innovation may be more or less distorted, both in
terms of their overall level and in terms of the direction of that research.
As we consider the potential impact of advances in AI on innovation, two
The Impact of Artifi cial Intelligence on Innovation 119
ideas from this literature seem particularly important—the potential for
contracting problems associated with the development of a new broadly
applicable research tool, and the potential for coordination problems aris-
ing from adoption and diff usion of a new “general purpose technology.”
In contrast to technological progress in relatively narrow domains, such as
traditional automation and industrial robots, we argue that those areas of
artifi cial intelligence evolving most rapidly—such as deep learning—are
likely to raise serious challenges in both dimensions.
First, consider the challenge in providing appropriate innovation incen-
tives when an innovation has potential to drive technological and organiza-
tional change across a wide number of distinct applications. Such general
purpose technologies (David 1990; Bresnahan and Trajtenberg 1995) often
take the form of core inventions that have the potential to signifi cantly
enhance productivity or quality across a wide number of fi elds or sectors.
David’s (1990) foundational study of the electric motor showed that this
invention brought about enormous technological and organizational change
across sectors as diverse as manufacturing, agriculture, retail, and residential
construction. Such GPTs are usually understood to meet three criteria that
distinguish them from other innovations: they have pervasive application
across many sectors, they spawn further innovation in application sectors,
and they themselves are rapidly improving.
As emphasized by Bresnahan and Trajtenberg (1995), the presence of a
general purpose technology gives rise to both vertical and horizontal exter-
nalities in the innovation process that can lead not just to underinvestment
but also to distortions in the direction of investment, depending on the
degree to which private and social returns diverge across diff erent appli-
cation sectors. Most notably, if there are “innovation complementarities”
between the general purpose technology and each of the application sectors,
lack of incentives in one sector can create an indirect externality that results
in a system- wide reduction in innovative investment itself. While the private
incentives for innovative investment in each application sector depend on
its the market structure and appropriability conditions, that sector’s innova-
tion enhances innovation in the
GPT itself, which then induces subsequent
demand (and further innovation) in other downstream application sectors.
These gains can rarely be appropriated within the originating sector. Lack
of coordination between the GPT and application sectors, as well as across
application sectors, is therefore likely to signifi cantly reduce investment
in innovation. Despite these challenges, a reinforcing cycle of innovation
between the GPT and a myriad of application sectors can generate a more
systemic economy- wide transformation as the rate of innovation increases
across all sectors. A rich empirical literature examining the productivity
impacts of information technology (IT) point to the role of the microproces-
sor as a GPT as a way of understanding the impact of IT on the economy as
a whole (among many others, Bresnahan and Greenstein 1999; Brynjolfsson
120 Iain M. Cockburn, Rebecca Henderson, and Scott Stern
and Hitt 2000; Bresnahan, Brynjolfsson, and Hitt 2002). Various aspects
of artifi cial intelligence can certainly be understood as a GPT, and learning
from examples such as the microprocessor are likely to be a useful founda-
tion for thinking about both the magnitude of their impact on the economy
and associated policy challenges.
A second conceptual framework for thinking about AI is the economics
of research tools. Within the research sectors some innovations open up new
avenues of inquiry, or simply improve productivity “within the lab.” Some of
these advances appear to have great potential across a broad set of domains
beyond their initial application: as highlighted by Griliches (1957) in his clas-
sic studies of hybrid corn, some new research tools are inventions that do
not just create or improve a specifi c product—instead, they constitute a new
way of creating new products with much broader application. In Griliches’s
famous construction, the discovery of double- cross hybridization “was the
invention of a method of inventing.” (IMI) Rather than being a means of
creating a single new corn variety, hybrid corn represented a widely appli-
cable method for breeding many diff erent new varieties. When applied to
the challenge of creating new varieties optimized for many diff erent locali-
ties (and even more broadly, to other crops), the invention of double- cross