A Focus on Diffusion Capacity

In “No, We Don’t Need Another ARPA” (Issues, Fall 2023), John Paschkewitz and Dan Patt argue that the current US innovation ecosystem does not lack for use-inspired research organizations and should instead focus its attention on diffusing innovations with potential to amplify the nation’s competitive advantage. Diffusion encompasses multiple concepts, including broad consumption of innovation; diverse career trajectories for innovators; multidisciplinary collaboration among researchers; improved technology transition; and modular technology “stacks” that enable components to be invented, developed, and used interoperably to diversify supply chains and reduce barriers to entry for new solutions.

Arguably, Advanced Research Project Agencies (ARPAs) are uniquely equipped to enable many aspects of diffusion. They currently play an important role in promoting multidisciplinary collaborations and in creating new paths for technology transition by investing at the seam between curiosity-driven research and product development. They could be the unique organizations that establish the needed strategic frameworks for modular technology stacks, both by helping define the frameworks and investing in building and maintaining them.

Perhaps a gap, however, is that ARPAs were initially confined to investing in technologies that aid the mission of the community they support. The target “use” for the use-inspired research was military or intelligence missions, and any broader dual-use impact was secondary. But today the United States faces unique challenges in techno-economic security and must act quickly to fortify its global leadership in critical emerging technologies (CETs), including semiconductors, quantum, advanced telecommunications, artificial intelligence, and biotechnology. We need ARPA-like entities to advance CETs independent of a particular federal mission.

The CHIPS and Science Act addresses this issue in a fragmented way. The new ARPAs being established in health and transportation have some of these attributes, but lack direct alignment with CETs. In semiconductors, the National Semiconductor Technology Center could tackle this role. In quantum, the National Quantum Initiative has the needed cross-agency infrastructure and during its second five-year authorization seeks to expand to more applied research. The Public Wireless Supply Chain Innovation Fund is advancing 5G communications by investing in Open Radio Access Network technology that allows interoperation between cellular network equipment provided by different vendors. However, both artificial intelligence and biotechnology remain homeless. Much attention is focused on the National Institute of Standards and Technology to lead these areas, but it lacks the essential funding and extramural research infrastructure of an ARPA.

The CHIPS and Science Act also created the Directorate for Technology, Innovation, and Partnerships (TIP) at the National Science Foundation, with the explicit mission of investing in CETs through its Regional Innovation Engines program, among others. Additionally, TIP established the Tech Hubs program within the Economic Development Administration. Both the Engines and Tech Hubs programs lean heavily into the notion of place-based innovation, where regions of the nation will select their best technology area and build the ecosystem of universities, start-ups, incubators, accelerators, venture investors, and state economic development agencies. While this structure may address aspects of diffusion, it lacks the efficiency of a more directed, use-inspired ARPA.

Arguably the missing piece of the puzzle is an ARPA for critical emerging technologies that can undertake the strategic planning necessary to more deliberately advance US techno-economic needs. Other nations have applied research agencies that strategically execute the functions that the United States distributes across the Economic Development Administration, the TIP directorate, various ARPAs, and state-level economic development and technology agencies. This could be a new agency within the Department of Commerce; a new function executed by TIP within its existing mission; or a shift within the existing ARPAs to ensure that their mission includes investing in CETs, not only because they are dual-use technologies that advance their parent department’s mission but also to advance US techno-economic competitiveness.

Charles Clancy

John Paschkewitz and Dan Patt make a strong argument that the biggest bottleneck in the US innovation ecosystem is in technology “diffusion capacity” rather than new ideas out of labs; that there are several promising approaches to solving this problem; and that the nation should implement these solutions. The implicit argument is that another ARPA isn’t needed because the model was created in the world of the 1950s and ’60s where diffusion was all but guaranteed by America’s strong manufacturing ecosystem, and as a result is not well-suited to address modern diffusion bottlenecks.

In my view, however, the need to face problems that the ARPA model wasn’t originally designed for doesn’t necessarily mean that we don’t need another ARPA, for three reasons:

1. While it’s not as common as it could be, there are examples of ARPAs doing great diffusion work. The authors highlight the Semiconductor Manufacturing Technology consortium as a positive example of what we should be doing—but SEMATECH was in fact spearheaded by DARPA, the progenitor of the ARPA model.

2. New ARPAs can modify the model to help diffusion in their specific domains. ARPA-E in the energy sector has “tech to market advisors” who work alongside program directors to strategize how technology will get out into the world. DAPRA has created a transition team.

3. At the core, the powerful thing about ARPAs is that they give program managers the freedom and power to take whatever actions they need to accomplish the mission of creating radically new technologies and getting them out into the world. There is no inherent reason that program managers can’t focus more on manufacturability, partnerships with large organizations, tight coordination to build systems, and other actions that can enable diffusion in today’s evolving world.

Still, it may be true that we don’t need another government ARPA. Over time, the way that DARPA and its cousins do things has been increasingly codified: they are under constant scrutiny from legislators, they can write only specific kinds of contracts, they must follow set procedures regarding solicitations and applications, and they may show a bias toward established organizations such as universities or prime contractors as performers. These bureaucratic restrictions will make it hard for government ARPAs to make the creative “institutional moves” necessary to address current and future ecosystem problems.

Government ARPAs run into a fundamental tension: taxpayers in a democracy want the government to spend money responsibly. However, creating new technology and getting it out into the world often requires acting in ways that, at the time, seem a bit irrational. There is no reason an ARPA necessarily needs to be run by the government. Private ARPAs such as Actuate and Speculative Technologies may offer a way for the ARPA model to address technology diffusion problems of the twenty-first century.

Ben Reinhardt

John Paschkewitz and Dan Patt make some fantastic points about America’s innovation ecosystem. I might suggest, however, a different framing for the article. It could instead have been called “Tough Tech is… Tough; Let’s Make it Easier.” As the authors note, America’s lab-to-market continuum in fields such as biotech, medical devices, and software isn’t perfect. But it is far from broken. In fact, it is the envy of the rest of the world.

Still, it is undeniably true that bringing innovations in materials science, climate, and information technology hardware from the lab to the market is deeply challenging. These innovations are often extremely capital intensive; they take many years to bring to market; venture-backable entrepreneurs with relevant experience are scarce; many innovations are components of a larger system, not stand-alone products; massive investments are required for manufacturing and scale-up; and margins are often thin for commercialized products. For these and various other reasons, many great innovations fail to reach the market.

The solutions that Paschkewitz and Patt suggest are excellent—in particular, ensuring that fundamental research is happening in modular components and developing alternative financing arrangements such as “capital stacks” for late-stage development. However, I don’t believe they are the only options, nor are they sufficient on their own to close the gap.

More support and reengineered processes are needed across the entire technology commercialization continuum: from funding for research labs, to support for tech transfer, to securing intellectual property rights, to accessing industry data sets and prototyping equipment for validating the commercial viability of products, to entrepreneurial education and incentives for scientists, to streamlined start-up deal term negotiations, to expanding market-pull mechanisms, and more. This will require concerted efforts across federal agencies focused on commercializing the nation’s amazing scientific innovations. Modularity and capital are part of the solution, but not all of it.

The good news is that we are at the start of a breathtaking experiment centered on investing beyond (but in lieu of) the curiosity-driven research that has been the country’s mainstay for more than 70 years. The federal government has launched a variety of bold efforts to re-envision how its agencies promote innovation and commercialization that will generate good jobs, tax revenues, and exports across the country (not just in the existing start-up hubs). Notable efforts include the National Science Foundation’s new Directorate for Technology, Innovation and Partnerships and its Regional Innovation Engines program, the National Institutes of Health’s ARPA-H, the Commerce Department’s National Semiconductor Technology Center and its Tech Hubs program, and the Department of Treasury’s State Small Business Credit Initiative. Foundations are doing their part as well, including Schmidt Futures (where I am an Innovation Fellow working on some of these topics), the Lemelson Foundation, the Walmart Family Foundation, and many more.

As a final note, let me propose that the authors may have an outdated view of the role that US research universities play in this puzzle. Over the past decade, there has been a near-total reinvention of support for innovation and entrepreneurship. At Columbia alone, we offer proof-of-concept funding for promising projects; dozens of entrepreneurship classes; coaching and mentorship from serial entrepreneurs, industry executives, and venture capitalists; matching programs for venture-backable entrepreneurs; support for entrepreneurs wanting to apply to federal assistance programs; connections to venture capitalists for emerging start-ups; access for start-ups to core facilities; and so much more. Such efforts here and elsewhere hopefully will lead to even more successes in years to come.

Orin Herskowitz