Science and Innovation at Scale
Successful applied research that draws on scientific invention requires the integration of multiple technological breakthroughs organized around use-inspired research goals. A team-based, interdisciplinary approach that cuts across fields of inquiry and institutional frontiers is needed to achieve this.
The great industrial laboratories of the mid-twentieth century provided just such an institutional setting. The decline of the industrial lab and the reliance on research universities is an institutional development that has slowed the pace of technology transition. While the federal government has experimented over many years with new institutional designs none of these experiments have yet proved to be effective substitutes for the industrial lab. What is still needed is an institutional setting that can reliably repeat and deploy successful technology breakthroughs at scale. The industrial labs provided a unique and complete mix of capabilities: Problem-driven research where researchers were exposed to market signals, multi-disciplinary teams that drew on many disciplines, simplified information flows up and down development pathways, large teams mobilized over many years, and invention and technology breakthroughs refined through access to integrated production facilities. Policymaker need to design and fund institutions with these attributes.
Productivity and Innovation
The discussion that follows is designed to isolate the principles to be kept in mind when designing new institutional architectures around substantial science and technology initiatives. It is grounded in a brief survey of key elements in the U.S. R&D ecosystem and of some of the ways in which the system has changed over time. Nothing that follows will be very new to anyone who has given much thought to these questions, and readers may not agree with the balance struck in specific parts of the argument. However, the overall purpose is aligned with the views of many who are concerned with the institutions, programs, and overall effectiveness of the U.S. innovation system.
If we suspect that there is weakness in innovation across the U.S. economy as a whole this is likely to be associated with a general decline in productivity since the late 1990s. There are many plausible explanations for this decline, including an aging workforce and a narrowing STEM workforce pipeline. Some explanations, however, relate directly to the ability of U.S. business to adopt and scale new technologies. We may be in a technological lull while we transition from productivity gains that stem from breakthroughs in single domains (for example, electricity) to those that require the combination of advances across multiple domains (for example, autonomous machines).[1] From this point of view, successful applied research now requires the integration of multiple technological breakthroughs. Success in such a case may require a new kind of institutional setting. In fact, the existing disciplinary setting for research at universities may be an obstacle to success in applied research and technology-based innovation.
The operational problems associated with applied research (technology transition/translation) are outlined by Joel Mokyr as follows: feasible techniques or applications mediate the relationship between propositional knowledge/basic science and prescriptive knowledge/applied science. This set of techniques or applications is determined in part by formal and informal institutions and activities. “Institutions set the incentive and penalty structure for people who suggest new techniques … [they are] … the key to explanations of invention and technological creativity.”[2] In short, the design of the institutional setting can determine the level of innovation success.
A report that recognized this challenge is the National Academy report on convergent science.[3] It emphasized a team-based, interdisciplinary approach to tackle scientific and societal challenges that cut across fields of inquiry and institutional frontiers. Simply stated, it championed a problem-driven approach to science.[4] A focus on real-world problems, rather than on cumulative, disciplinary research, almost invariably fosters a team-based, interdisciplinary approach. A focus on real-world problems also highlights the importance of innovation and enterprise in achieving technology transition. This approach is laudable, but of course requires an institutional setting that overcomes disciplinary barriers. Leading research universities increasingly have made strides in this direction, for example, integrating information science and genetics. But such convergence still has a strong basic science orientation, reflecting the professional incentives faced by faculty.
Institutions and Technology Transition
The actual transition of a technology to the marketplace occurs over a long arc of activities, some associated with technology to product development (moving a technology along from proof of principle to prototype to product in the market), some with technology to market and business model development (shepherding a discovery from a startup or through licensing toward a sustainable market presence). This trajectory must overcome a series of well-known hurdles. Hurdles are hardest for the type of science-based innovation being discussed here. In other words, there are many “valleys of death,” each with specific characteristics.
Examples of these valleys vary. Some technologies are targeted at heavily regulated spaces. Success requires specific legal knowledge and additional resources. Some applications of new discoveries may be in technically challenging areas, for example nanoscale materials, which will require access to a variety of capabilities over a long period of time (including long-term funding, specialized facilities, and technical support services). Some may target highly specialized markets only capable of absorbing limited production runs, in which it may take many months, even years, for an enterprise to become viable or an investment in a license by a firm to earn a return. A wide range of supportive activities are helpful, and often indispensable, in order to successfully commercialize technology. These depend upon the institutional context and its suitability for any particular technology domain, regulatory arena, and—above all—end user market.
There is evidence that fundamental changes over the post-war period in the institutional and business setting for the research, development, and adoption of new technologies by the U.S. economy have contributed to sluggishness in growth in productivity and are associated with widespread failure to harvest the fruits of U.S. R&D. A recent paper suggests that the documented decline of the centralized corporate research lab (as evidenced by scientific publication as well as much else) has yielded a new division of labor, with universities now the primary home of basic research, and corporations focused on close-to market product development. This new division of labor has had consequences for technology transition that mirrors the decline in productivity:
“The translation of scienti?c knowledge generated in universities to productivity enhancing technical progress has proved to be more difficult to accomplish in practice than expected. Spinffs, startups, and university licensing offices have not fully ?lled the gap left by the decline of the corporate lab. Corporate research has a number of characteristics that make it very valuable for science-based innovation and growth. Large corporations have access to signi?cant resources, can more easily integrate multiple knowledge streams, and direct their research toward solving speci?c practical problems, which makes it more likely for them to produce commercial applications. University research has tended to be curiosity-driven rather than mission-focused. It has favored insight rather than solutions to speci?c problems, and partly as a consequence, university research has required additional integration and transformation to become economically useful.”[5]
It is important to distill, to the extent possible, the benefits (now lost) of the industrial research lab. The authors noted above share similar views to those summarized by Suzanne Berger in her work on innovation.[6] The industrial lab was distinguished by:
These characteristics not only produced breakthroughs for the firms in question, but also yielded significant spillovers. They produced general purpose technologies that other enterprises would share in sooner or later, and they fueled an ecosystem of skilled workers, complementary suppliers, and downstream partners. It is this loss that the U.S. economy continues to experience.
U.S. business gave up centralized research labs for a variety of reasons. It became easier to buy innovation through M&A activity. The output from the labs was not valued by activist shareholders with short time horizons. The firms themselves became more specialized and thus unlikely to benefit from broad innovative activity. While it is true that we presently observe very significant investments by Silicon Valley firms in machine learning and artificial intelligence, these are unusual cases. The more general belief among Fortune 500 companies is that the corporate lab no longer brings value to the bottom line.
New Institutional Designs
Policymakers and federal agencies have been well aware for decades that the United States has trouble moving publicly funded technologies out of the lab and into use. There have been many policy initiatives designed to address different aspects of the problem. These changes are wide ranging, from the Bayh-Dole Act which liberated publicly funded IP, changes in anti-trust law that facilitated the formation of industrial consortia, and the creation of a wide range of new programs within NSF and other mission agencies, such as the Department of Defense and the Department of Energy. These programs have focused on different aspects of the problem, and include, among many others, Industry-University Cooperative Research Centers (IUCRCs), the I-Corps program, the SBIR/STTR program, ARPA-E, and the advanced manufacturing institutes that constitute the Manufacturing USA network. Research universities and the national labs participate in these programs and have also built out their own technology transfer capabilities and facilities.?
The variety of these initiatives is evidence of laudable efforts at institutional innovation. However, there is little evidence yet for an impact in the aggregate on overall levels of technology-based innovation. Indeed, there is some evidence that while research activity is rising, research productivity is declining.[7] One possible explanation for the striking contrast between institutional innovation and limited overall success may lie in the fragmented, small-scale nature of these initiatives. The U.S. policy environment and the U.S. R&D ecosystem rests on diffuse and distributed efforts. This is a virtue in that many ideas for success are allowed to bloom, and it is consistent with the attitudes of policymakers who tend not to value singular goals and strategies at scale. However, fragmentation brings with it costs to the R&D ecosystem which may be systematically underestimated, costs that can be understood by looking at the industrial research lab from the point of view of a single production function.?
The industrial research lab was home to several capabilities, as outlined above. The important point is that all were present and integrated into a single institutional setting. In this way the lab was able to have a substantial impact on the market. The institutional context is important, since the failure of one part meant the failure of the overall effort.[8] This is a form of the “O” Ring production function, in which the component parts of an economic activity are highly complementary and in which the failure or inferior performance of one part drags down or even completely stymies the effectiveness of the whole.[9] This may sound familiar to anyone engaged in technology transition at a lab or research university. The fragmentation of the innovation ecosystem means that the quality of the components in the system can’t be known. As a result, there is a large number of possible failure points along the path to successful technology transition, and technology transfer experts often encounter a different kind of failure each time.
The industrial research lab of an earlier age is gone forever (with the possible exception of such giants as Google and Apple entities that focus on only one important area, AI), but it is worth noting its history of scientific excellence:
“In the 1960s, DuPont's central R&D unit published more articles in the Journal of the American Chemical Society than MIT and Caltech combined. However, in the 1990s, DuPont's attitude toward research changed and after a gradual decline in scientific publications, the company's management closed its Central Research and Development Lab in 2016.”[10]
One implication of the “O” Ring production function is that high quality human capital—in other words, talent—has an incentive to cluster together, because working alongside other talents of all kinds makes each more productive. It is commonplace to note, for example, the world-class talent that was drawn to Bell Labs during its heyday. Moreover, a lead scientist in a university (i.e., a faculty member) might be responsible for securing the money, doing accounting, ensuring compliance paperwork, being safety officer, teaching, in some cases collecting the data, analyzing the data, generating reports, etc. Even today, scientists in companies have much more time available to them to be engaged in the science/technology and technical discussions (even though they spend a lot of time doing some of those other things).[11]
Bringing in Business
What would a new institutional setting look like, under present conditions, that embodied the capabilities of an industrial lab? Seen from a certain point of view, Stanford and MIT are latter-day versions, with some of the pieces in place to accomplish technology transition, operating at a high level of quality. Key elements of their ecosystem, for example venture funding, operate with tight interdependence. Yet for the most part, their innovation success is almost entirely dependent on students who leave to form successful companies. Other research universities are all trying to emulate their success, but they don’t control all the key elements of their ecosystem, and their faculty are embedded in academic disciplines. Their all-important role is in producing talent; beyond that, the record is limited. More importantly, all of these institutions operate at arm’s length from the demands of industrial production and from the marketplace and thus struggle with the challenge of carrying discovery all the way through technology development into widespread use.??
The startup is considered part of the solution to the loss of the industrial lab. Large enterprises scout the ecosystem for innovations brought to market by successful entrepreneurs, grabbing likely prospects that have succeeded in crossing the valley of death. This is a reasonable solution from the point of view of established businesses, who can pay for de-risked technologies that match their business needs. For society as a whole, however, there may be valuable technologies—publicly funded technologies—sifted out of the ecosystem that never go to scale. Entrepreneurship at research universities is highly prized and success widely reported. But outside the Silicon Valley ecosystem there have been very few disruptive technologies produced at scale through startups. Science-based innovation has unique challenges in that it requires scientific expertise (and all the support infrastructure of labs, etc.) and must also connect with market needs. Startups fail here because they will almost never have the resources for science infrastructure except in the narrow case of drug startups (which are essentially expensive lottery tickets). Those that we know well are often digital platforms that scale very quickly and appeal to the relatively short time horizons of venture funding.??
Nevertheless, the institutional landscape across the United States remains strong compared to other countries, even though in absolute terms the dollars provided by the federal government have declined over time. The U.S. R&D ecosystem is home to great institutional variety and to technology areas of clear strength. This variety of arrangements across technologies and institutional settings reflects the unique distributed characteristics of a system that offers many opportunities for institutional innovation, as discussed above. The task before policy-makers is to chart a path that knits these assets together into new institutional designs that address the challenge of taking science-based technologies to scale[12]
This kind of challenge has been addressed in the past across many advanced economies through the formation of industrial consortia. In many ways the manufacturing institutes launched over the last decade are a deliberate attempt to improve on past efforts aimed at binding together businesses facing technology-based competitive threats with research partners. Part of the institutional challenge is to replicate the attributes of the industrial research lab in the setting of a consortium: Combine researchers from different backgrounds around a set of problems, do it at scale, and integrate firms that produce at scale for the market. Where these institutes have struggled is in coordinating around a shared vision of the problems to be solved. Furthermore, the funding levels for such initiatives have been small and time limited. We no longer have industry labs connecting interdisciplinary research to market needs and we have not found a new way to do that.
Consortia face another challenge that an industrial lab, as an integrated institution, did not. The principals in any consortium are engaged in a complex set of contracts, designed to embody shared expectations about the purpose, costs, and benefits of the project. These contracts are intended to lower transaction costs among the partners, especially among competing businesses. Variation in the attributes of different technology domains and in industrial structures means that the web of agreements required will vary in every case. The consequence is that every consortium is different and likely to vary in its level of success. The results are complex governance structures that layer slow decision-making onto the organization.
The Microelectronics and Computer Technology Consortium (MCC), launched in 1982, is a well-known example of an industrial consortium in the United States that paved the way for many others. A thorough analysis of its history shows how the problems it was designed to solve echo the discussion above of industrial research labs: stung by the rise of Japan in micro-electronics, key industry players wished to combine their resources to develop a broad base of fundamental pre-competitive technologies for consortium members who would then go on to compete in downstream markets.[13] It would develop key technologies from discovery to advanced development, pooling firm resources, avoiding duplication of efforts, and allowing firms to complete in product development and production.
The challenges it faced are familiar given the discussion above: It did not have significant operating capital at the outset—it was expected to quickly become self-sustaining—and never had permanent staff. Information flows were restricted as a result of anti-trust and competitive concerns. Discount membership for Small and Medium-sized Enterprises (SMEs) was restricted due to fears of unfair access to technologies. More fundamentally, firms were not inclined to work on anything critical to their bottom line with competitors, so the research was inevitably relegated to less important areas, narrow projects not broad ones. In fact, DARPA's Microwave and Millimeter Wave Integrated Circuit (MIMIC) program which it took over in 1988 had greater long-term impact.
In one way or another, these challenges faced by consortia reflect a reasonable fear of what the business literature calls “opportunism.”[14] Opportunism is a transaction cost present where separate parties fear being exploited by other parties to a contract and leads to a break down in contracting. Intriguingly, the literature argues that these transaction costs are a cause of firm integration. Folding disparate and uncoordinated elements with divergent interests into a single institutional entity resolves the problem of opportunism. In the case of the R&D enterprise, it was a way to solve the problem of the “O Ring” production function. All of the necessary elements, at a high level of quality, are inside the same institutional setting. This accounts, in part, for the success of the industrial lab.
Building a New R&D Infrastructure
How would the United States resolve these institutional challenges at scale if it decides to establish a set of new innovation hubs that integrate critical technologies around world-scale problems? There is no silver bullet, but the following principles should inform institutional design:
The industrial labs have not wholly vanished from the scene, for example HRL inherited the legacy of the Hughes Research Laboratories and serves its two major industrial member companies (Boeing and General Motors). SRI International is home to the RCA/Sarnoff laboratories, which now operates alongside a diverse set of other divisions. The benefits of inter-disciplinary, problem-driven research are visible in their work, which straddles the gap between academic research and commercial product development. However, neither entity operates at scale, focused on a singular broad domain of technical challenges. Yet this is the kernel of what a new model should look like.?
The task before the United States is to design and promote a 21st century version of the industrial lab. It will be an order of magnitude bigger than existing initiatives, located outside the major coastal conglomerations, focused on a distinctive problem set, serving the needs of business and federal agencies. In particular, researchers at such entities will be much closer to the demands of the marketplace than is possible at an academic setting. The manufacturing institutes had, at their heart, delicately negotiated governance arrangements carefully coordinating the actions of partners without permanent scientific teams. What is needed is an institutional arrangement with an independent corporate life, a substantial, well-financed, fully staffed and equipped institution at the heart of a business network, so situated as to have some discretion in problems addressed, staff hired, and resources allocated.
There has been a variety of other suggestions for a dramatic increase in technology investments along these lines, in addition to the innovation hubs proposed by the Council.[15] What will be crucial for success is the attention paid to the relevant institutional technology. Investments on such a scale will not all succeed. A 50% success rate would be ambitious. But successful investments on this scale that ensured U.S. leadership in four or five critical technologies, and dominance over the associated markets, would make a substantial contribution to technological leadership far into the 21st century.
When seeking an appropriate design for pushing back against the headwinds that research, development, and innovation face in the United States the role played by Bell Labs in the past casts a long shadow.[16] Bell Labs is gone forever because the kind of national integrated business that sustained the lab has also gone and is not likely to be repeated. Yet the heart of the challenge to be addressed persists. The work of the lab always had application in mind. Any search for a new institutional framework needs to keep use inspired research as the goal, informed by the demands of the marketplace or by the demands of an end-user.
[1] For an overview featuring economists Robert Gordon and Joel Mokyr who take opposing views see: https://www.goldmansachs.com/our-thinking/pages/macroeconomic-insights-folder/the-productivity-paradox/report.pdf
[2] Joel Mokyr. “The Gifts of Athena: Historical Origins of the Knowledge Economy.” Princeton: Princeton University Press. 2002. p. 18.
[3] National Academy of Sciences. “Convergence: facilitating transdisciplinary integration of life sciences, physical sciences, engineering and beyond.” 2014.??
[4] A good example of a problem driven approach is the National Academy of Engineering’s “Grand Challenges”
[5] Ashish Arora, Sharon Belenzon, Andrea Patacconi, and Jungkyu Suh. "The changing structure of American innovation: Some cautionary remarks for economic growt.h” NBER Working Paper #25893. June 2019. p. 4.
[6] Suzanne Berger. “Making in America: From Innovation to Market.” MIT Press. 2013. pp. 18-23.
[7] Nicholas Bloom, Charles I. Jones, John Van Reenen, Michael Webb. “Are Ideas Getting Harder to Find?” NBER Working Paper #23782. September 2017.
[8] Xerox PARC is a well-known example of an industrial lab that achieved brilliant results, but that fell at the final fence. Xerox management never fully understood its results or took them to market.
[9] Michael Kremer. “The O-Ring Theory of Economic Development.” The Quarterly Journal of Economics. Vol. 108, No. 3. August 1993. pp. 551-575.
[10] Ashish Arora, et al. p. 3.
[11] Kelvin Lee contributed this important point based on his experience with manufacturers.
[12]“Innovation Leaders Propose Nine Pillars for Competing in the Next Economy.” Council on Competitiveness. June 18, 2020. https://www.compete.org/news/12-general-news/3408-press-release-nine-pillars-for-competing-in-the-next-economy
[13] David V. Gibson and Everett M. Rogers. “R&D Collaboration On Trial: The Microelectronics and Computer Technology Corporation” Harvard University Press: Boston (1994) pp. 65-69.
[14] Oliver E. Williamson. “Markets and Hierarchies: Some Elementary Considerations.” The American Economic Review. Vol. 63, No. 2. May 1973.
[15]Noah Smith. “The U.S. Gets Serious About Catching Up to China in R&D.” Bloomberg. June 1, 2020. ?https://www.bloomberg.com/opinion/articles/2020-06-01/the-u-s-gets-serious-about-catching-up-to-china-in-r-d
[16] See Gertner, John “The Idea Factory: Bell Labs and the Great Age of American Innovation” The Penguin Press, 2012
Outsource CMO I Transformational Marketing Thought Leader I Trusted Advisor to CPG, B2B, and startups I Brand Builder I Product Innovator I Leader of High-Performance Teams and Expert Project Manager I Profit Driver
2 年Who funds?