Causes of the Trend

When asked by Pines about the causes of declining output productivity, Scannell focused on three. First, the drug industry is far more regulated today than in the 1950s. Until 1962, the FDA did not require companies to show efficacy. In 1948, the first randomized control trial (RCT) enrolled about 100 patients versus the tens of thousands required today. Second, new drugs must compete against an ever-expanding catalogue of cheaper generic medicines. This discourages R&D investment in already successful areas, so R&D is pushed to address “difficult” disease areas where private industry sees new and better opportunities for short-term revenue return. Third, these areas are difficult because pre-clinical models of those diseases (e.g., animal models, in vitro models, computational models) are poor at predicting successful drugs. The “predictive validity” of disease models is a powerful determinant of whether a particular disease area is tractable, he said. Scannell cited oncology R&D, where drug candidates have a very high rate of failure in human trials after success in the pre-clinical models.

The private sector lacks sufficiently strong economic incentives to develop better disease models because the knowledge that the models generate can often benefit competitor firms via “knowledge spillovers”. Rather, the private sector tends to focus on the creation of novel compounds which are protected by the patent system, even without good pre-clinical methods to identify which are likely to work in patients. Short term revenue incentives are misaligned with the optimal outcome of curing diseases, he asserted.

When asked what other fields can learn from this phenomenon in the pharmaceutical industry, he noted that formal modeling of the R&D or discovery process can help identify the rate limiting steps. The quantitative importance of these steps may not be obvious without formal analysis.

Disease Models

A participant noted that developing disease models would be an optimal area for public investment but asked whether that was currently being prioritized. Scannell replied that though most people understand the importance of developing good disease models for public good, many researchers who rely on public funds see model development as too mundane.

Another participant asked what kind of partnerships need to be created to set the standards for an accurate disease model. Scannell suggested that the people who are most heavily invested in using the data that the models create should be involved in creating the “target model profiles” and in model evaluation.

Serendipity in Drug Discoveries

A participant cited the impressive new class of drugs targeting the GLP-1 receptors as an example of how basic research can lead to important discoveries. Scannell agreed, and added that historically, serendipity has played an outsized role. Drugs developed to target one disease have been shown to be effective in unexpected ways, with Viagra (originally developed for heart disease) as a prime example. Scannell continued that the “mechanistic story” of a drug—which seeks to predict the rational biological and chemical mechanisms by which a drug is effective—is not usually a genuine prediction of success but is often used to convince investors that the design of drugs is more targeted toward an outcome than it actually is. He noted that if the true ratio of serendipity in the drug discovery process was acknowledged, certain aspects of the process could be redesigned for better outcomes.

Impact of Artificial Intelligence on Drug Discovery

An audience member asked about the impact of AI on drug discovery. In Scannell’s opinion, it will continue to be important but will be incremental and not revolutionary for two reasons. First, pharmaceutical companies have already done a lot of quantitative work in the areas that are most amenable to AI such as protein folding, which leaves less space for large advances. Second, relevant data on the most common reasons most drugs fail is poor, sparse, biased and badly annotated, and thus not suitable to train ML systems to accurately predict outcomes.

A New Look at “Failing Fast and Early”

A participant noted that most drug discovery efforts look for front-end targets, which is seen as much cheaper than clinical trials, and companies have argued that drug candidates should fail fast and early. Scannell disagreed, asserting that good decisions, more than fast decisions, create value. Better disease models could a create bigger spread between the true positive rate and the false positive rate of the R&D process, improving decision quality. He also noted that clinical trials are seen as expensive, but when including the costs of capital and failure, then a preclinical step known as “lead optimization” is, based on some analyses, the most expensive step in the process.

Drug Investment

Another participant asked about investment trends in drug development. Scannell believes pessimism about the drug development pipeline in the pharmaceutical industry and biotech peaked around 2008–2010. An uptick in optimism and investment followed as the rate of new drug approvals increased and as low interest rates made biotech and pharmaceutical investments more attractive. Economic incentives tend to pull companies into certain R&D trajectories, Scannell commented. He cited as an example the financial and regulatory disincentives of developing diagnostic tests for early cancer detection. Early detection has had a larger role in decreasing cancer mortality than new pharmaceuticals, he said, though private sector capital flows more easily to developing new drugs.

Identifying and Overcoming Disincentives to Partnerships

Anderson and Shenoy both identified the misalignment of timelines as a challenge. Shenoy cited energy development investments that were such long-term endeavors that it was futile to plan for outcomes. While recognizing that the government has the longest time horizon, he urged industry to recognize that infrastructure development cannot be as rapid as software prototyping. There are challenges in scoping that all participants must work on, he said.

Walter noted that policymakers on Capitol Hill get frustrated when they have appropriated funds, but the project hasn’t yet shown impact because realistic timelines are longer than appropriators expect. One obstacle is that alternative methods for contracting and other vehicles that can fill in funding gaps in the short term are little known. He gave a recent example of a commercial technology that won a financing agreement through the SIBR program. Its funding wouldn’t begin until FY 2026, leaving investors with a year with no revenue and thus a pressure to cut the technology, even with promise of a contract in the future. They are searching for a more flexible contract for the year to close the gap.

Anderson said that another challenge is “educational overhead,” i.e., the burden on the private sector to understand how to work with the government. Even finding the right government stakeholders is challenging.

She described how IQT has created a government platform accelerator that collects the foundational information that start-ups need to approach the government market. Their curriculum provides step-by-step processes that have worked well for some of the companies in their portfolio. However, she noted there are tens of thousands of other companies that could benefit from increased awareness about the federal marketplace.

Shenoy observed that agencies find it difficult to stay abreast of the wide range of ongoing R&D investments. He described an exercise last year where different DOE offices asked participants to identify all the pathways they were aware of for the U.S. energy portfolio and solicited feedback on the gaps. Commenters offered approaches that DOE would not have known about otherwise. He urged start-ups and others to respond to these solicitations to become a “known quantity” to DOE and other funding agencies, which will have a positive effect on their future funding applications.

Dealing with Risk

Winslow commented that the National Institutes of Health (NIH) is often criticized for avoiding high-risk projects and asked how that debate affects how partnerships are created. Walter found this to be a difficult question to answer because there are no clear-cut quantitative metrics for success. Rather, he suggested using strategic communications to demonstrate real impact of projects to convince funders to take risk. Anderson noted that funders should accept more risk in some areas vs others. For example, if a non-critical piece of software fails, there are fewer consequences than for a critical system that needs a pristine solution. This is where PPPs can be more forward-thinking.

Effects of the Bayh-Dole Act on PPPs

A participant asked about the impact on PPPs of recent reinterpretation of the Bayh-Dole Act’s “march-in” rights, i.e. the federal government’s right to require a patent holder to license to third parties if the invention was developed with federal funding.⁸ In IQT’s experience, Anderson said, the government exercises its march-in rights much less frequently than in the past and is much more comfortable funding R&D without using those rights. Convincing companies that the risk of this is low is a key part of shepherding these partnerships to fruition. Shenoy said he has not seen this as a friction point with the long-term projects with which he has been involved, although he has seen some points of friction with universities that have shifted toward more commercialization. Walter said he has not seen friction on the research side, although he has on the contracts side when the U.S. government has wanted to avoid being overly beholden to a supplier.

Accelerating Government Timelines

In considering how to accelerate funding from the government, Walter said his program at AFWERX has been able to award in 90 days. Last year, they received 8,000 applications and processed 1,570 contracts, and they must also deal with increasing requirements. Perhaps there should be a look at the commercial market for growth investments, he suggested. Several DOD programs are seeking partnerships with private capital investors for efforts that should be accelerated. Anderson suggested being more open about successful partnerships, including sharing case studies. Shenoy said large corporations also must deal with bureaucracy and multiple stakeholders. If a process is not well designed, there will be bottlenecks. He pointed out that high school students can apply to colleges using the Common App, yet SBIRs in different agencies have slightly different applications, an example of friction in the system.

University Partnerships

Universities may also have misaligned timelines, a participant observed, and asked how to accelerate university involvement in PPPs. Anderson cited NobleReach, a nonprofit, as an example of a way to equip people with technical expertise with the business acumen they need. They “put teams around” programs emerging from of DARPA. Shenoy said major universities in the Houston area have come to him and colleagues to discuss how to help those with technical but not business expertise. Relevant industry experience, whether through government agencies or the private sector, is essential. Walter said the Small Business Administration has made changes in its STTR program to allow more money to be paid out for business expertise.

Shaping a Portfolio

Managing and shaping a R&D portfolio is often discussed but not always implemented, a participant observed. Anderson said IQT looks at historical data and other factors in building their portfolio, including completion rates and stakeholder evaluation. The portfolio has had an 80 percent pilot rate, about half of which go to adoption. Walter added that metrics need to inform decision making. Defining a portfolio and what constitutes success becomes hard with limited information. In terms of success, new entrants may not understand that the real stakeholders are the decision makers or those with the money. Shenoy considered the perspectives of large corporations, start-ups, and venture capitalists in building a diverse portfolio. For a large corporation, maybe 10 percent of its R&D budget is for fundamental blue-sky research. The stakeholders are the board and CEO, and they do not have the patience for analysis. Some blue-sky funds are delegated to the CTO, with an expectation of return in the next 10 to 20 years. Venture capitalists build a portfolio to ensure a high return rate to investors in a shorter timeframe. Since most projects will fail, diversification is crucial. Walter said government agencies also should have diverse and well-funded portfolios.

Workforce Considerations

A participant noted that workforce to produce future technologies is an essential factor. Shenoy stressed this point and pointed out that in the 1980s and 1990s, the focus of PPPs shifted from industrial policy to science policy. Considerations of how innovation leads to a healthy market were not prioritized. In the pursuit of capital efficiency, companies turned away from internal research labs and domestic manufacturing capability. Deskilling of labor resulted. Walter said engaging with students early will help build the workforce, as will increasing U.S. funding of academic commercialization projects.

Economic security is national security, which leads to policies for workforce development and training, an audience member commented. The CHIPS + Science Act is trying to accomplish this. Drawing on his experience as a program officer at DARPA, he explained how he used SBIR and other funding mechanisms to seed research. “Innovation and success can be accelerated by using funding mechanisms and derisking enabling technologies to get to an end product faster,” he commented. He called for policies tied with models, money, and management.

New Strategic Partnerships

A participant asked the panelists for examples of new promising strategic partnerships. Walter described Project Vanguard,⁹ which directs private investment into companies that AFWERX has funded. He recognizes that his job in the government is not to do due diligence or valuations but rather to reduce obstacles for private investors. He supports aggregating and curating data in ways that make sense to an investor. Anderson commented that the philanthropic angle is important. To build the partnerships, she urged all stakeholders to communicate at the start of any process and continue real-time engagement. Shenoy sees a role for philanthropy where the government does not have resources or markets fail. One example is how philanthropy filled the gap in funding for the development of the low-cost CORBEVAX^TM COVID-19 vaccine at Baylor College of Medicine,¹⁰ which can be made available to populations in developing countries. He is also working with a nonprofit on using technology to support lower-resourced schools to meet the needs of underserved communities in the U.S.

Balancing Priorities

Vergano asked Gruber how to balance the need for open collaboration with concerns about competitiveness and security and whether different technology issues are approached differently. Gruber said policy and science diplomacy are very dynamic and decisions on how to handle different technology issues are made on a case-by-case basis. Every policy goes through review and comment stages to look both at promotion and protection. For example, Secretary Blinken has publicly promoted “digital solidarity,” which recognizes that all who use digital technologies should do so in a rights respecting manner through secure, resilient and self-determined networks and infrastructure. The underlying tenets are to use technology for “good”, shape the rules, and build ecosystems that are relevant to serving society.

Openness must go hand in hand with security, she stressed. She noted the commonly used analogy “small yard, high fence”, in which the high fence protects sensitive technology in collaboration with partners. She prefers the analogy of a “shared garden,” she continued, where an open ecosystem is crucial for collaboration. Drawing on the garden analogy, cultivating relationships with some countries relies on harvesting longstanding ties, and others involve just planting seeds. She noted that scientists are comfortable with the idea that collaboration and competition co-exist, and policymakers need to be comfortable with this idea as well.

Bessen expanded on the trends he discussed by describing how software manages complexity and how recent development has more often been proprietary. This enlarges the gap between “the best and the rest,” rather than diffusing the benefits of innovations to the entire economy. The nature of business has changed, geared toward product differentiation that makes a company less likely to license its technology, he contended.

The impact of a top-heavy innovation ecosystem is that knowledge is not flowing and there is less disruption. A more diffused ecosystem would produce more robust growth and productivity, he said. He called for policies to encourage firms to be more open, with a balance between incentives to invest and allowing knowledge to diffuse so society benefits. Although the pendulum is still pointing toward proprietary rights, he identified encouraging signs such as cloud computing that have allowed small companies to access top-of-the-line technology. Private firms can be encouraged to open up, for example by offering APIs that allow access to key components but not to the entire system.

Talent and Workforce

Gruber said foreign talent is essential for technology competitiveness. She noted Secretary Blinken has a concept of “diplomatic variable geometry,” or building the right team to create fit-for-purpose partnerships. The department has become more active in engaging nongovernmental organizations (NGOs), universities, the private sector, and others. The Quad Investors Network¹² offers a model, she suggested, in which private investors in Australia, Japan, Korea, and the United States co-invest in technologies. Other Department of State examples are the Global Innovation through Science and Technology program,¹³ which cultivates international partnerships through in-country entrepreneurial education, and the Office of Global Partnerships,¹⁴ which brings delegations of U.S. small companies and venture capitalists to other countries.

She echoed concerns about the workforce and noted an important aspect of the CHIPS + Science Act is to consider future workforce needs. The department organized a semiconductor workforce symposium in Costa Rica, and she envisions similar symposia for other emerging technologies. She also noted that engaging the future workforce, i.e. the younger generation and particularly underrepresented groups, is crucial.

Dealing with Disruption

A participant suggested to Bessen that companies may be “disrupting themselves.” Bessen sees evidence of this in big tech but not in other sectors of the economy. Noting that Amazon decided to release its AWS capabilities to others, a participant asked whether any regulations or processes might help smaller companies. Bessen said some large enterprises are providing access to data for small companies. Industry-specific regulations, such as in banking, and anti-trust regulations also have a role to play. A government nudge may help, he said. One participant differentiated between diffusion in software and hardware development. In hardware, a lot of learning takes place “on the factory floor,” which is harder to share. The barriers to entry for software are lower than for hardware, which needs more infrastructure, she noted.

Foreign Policy Amidst New Technologies

When Vergano asked how U.S. foreign policy needs to change given the powerful role of technology, Gruber pointed to the Executive Order on Artificial Intelligence as an example.¹⁵ Drawing from the voluntary set of principles contained in the order, the U.S. introduced a resolution at the United Nations to put guard rails around responsible use of AI. Active discussions are needed to address the gap identified by Bessen, she added. There is a wide range between not asking companies to give up frontier models and allowing access to other innovators.

Shutting doors through import and export controls, for example with 5G, limits innovation, a participant commented. Gruber said this issue is constantly debated in concert with other agencies to weigh the pluses and minuses. An example is restricting import of Chinese electric vehicles (EVs), which conflicts in some ways toward reaching an EV economy. Bessen acknowledged the ongoing debate about security and open-source solutions. He noted there is some evidence that open source enables bad actors to create vulnerabilities but also can identify and shut them down more quickly.

Role of Academia

Greater engagement of the academic community in global science diplomacy would be a value add to the United States, a participant suggested. Another role for academia and small companies is in the global standards processes, an area where he said those sectors are underrepresented. Gruber responded that one role of academia is to build trust in science and instill in students the responsibility to communicate with the public about their science. She also noted U.S. efforts to increase engagement in Africa and other developing countries. Her office liaises with the domestic academic community to explore other ways to bring academia to the table. A participant also cited the contributions of scientists from Ukraine as an example of the benefits of importing technology and other innovations.

Funding Models and Processes

Córdova asked Sanford his view on important design elements or processes for investing in scientific research. In the case of the Hypothesis Fund, he said, the goal is bold science and new and divergent ideas. This is why they developed the scout model, rather than rely on more traditional approaches.

She asked Gamick how alternative funding models work with policymakers and the government to create markets. Gamick used ocean-based carbon capture as an example. To create a market, policies are needed around pricing and incentives. A hole in this case, however, is that it is not possible to directly measure and verify how much carbon has been removed. Development of a modeling platform is needed but is not incentivized in for-profit settings and is too big for a single lab or agency to undertake. To solve this bottleneck, Gamick helped set up a nonprofit to develop new software for measurement and verification, funded by philanthropic and public funding and that is partnered with for-profits.

Levine suggested an important consideration about funding models is the intended result. The country has a goal to be preeminent in foundational science, but also to “sit in the translational space.” As a framework, she suggested thinking about value creation and value capture. In her view, the U.S. is good at value creation of the S&T enterprise but poor at value capture. In the security space, for example, many innovations developed in the U.S. are now manufactured in other countries. Application should not be decoupled from foundational science, she urged. Access to capital, infrastructure, and talent are all needed. Large industries can create a need for talent, but if the economy cannot absorb STEM students, they will enter other fields. A lot of the dedicated capital to support infrastructure for fundamental research is now in China, not the U.S., she added.

Time Frames and Risks

When asked about long-term and short-term investing, Holloway called for ways to incentivize long-term thinking and behavior, which he posited requires Congressional action. Decisions flow from policies and funding, such as when universities pick faculty who will evolve research labs and develop graduate students. If the government cannot maintain a balance, it is difficult for the rest to flow. Sanford said the country needs to both use existing knowledge and invest upstream in fundamental science that is the lifeblood of translational technologies. A portfolio strategy should spread risk and invest in different risk profiles, he said. Different funders have different tolerance for risk. Holloway suggested GUIRR as a place where different groups can coordinate, even on an informal or intellectual level, to create a national portfolio of research.

Córdova commented that people usually overestimate how risk-tolerant they are. Gamick said that risk shows up in different ways. When start-ups in Silicon Valley go under, this is considered okay, and the founders go on to start another company. In academia, risk surrounds publication and tenure, and options seem limited, so people are more risk averse. Convergent Research currently operates six projects and is working on several others. She does not expect them all to succeed and, if they do, that means “we have not picked hard enough problems.” While they try to control for operational risks, sometimes the science will not pan out, which funders accept. Sanford agreed with encouraging more risk, but noted how important it is to clearly define desired outcomes and build incentives that lead to those outcomes. It requires culture change to reward those who go after big ideas that might fail.

Levine noted that venture capitalists are good at evaluating business and market risk, but not scientific and technical risk. She urged scientists and technologists to join the venture capital community to provide this expertise. Although there are things to criticize in the U.S. S&T ecosystem, “do not underestimate its dynamism,” she continued. That said, it is important to create incentives for government program managers to take the risk to fund large capital projects for new scientific infrastructure. In another example, Gamick observed that space transportation and imaging technology has made advances that the last Decadal Survey in Astrophysics did not envision. This calls for flexibility when technology is advancing quickly.

Roles for Different Players

In discussing a role for larger, more established companies in advancing innovation to markets, including contributing to the public good and developing talent, Gamick commented that her organization developed some IP based on a larger company’s IP product. The ability to be a small nonprofit and work with that company through a meaningful partnership was important. Holloway identified three areas for larger companies: to employ students to sustain their interest and early engagement in STEM; to acquire startups; and to ensure that their internal R&D does not become moribund.

Sanford explained that the name “Hypothesis Fund” was chosen for two reasons—because they are looking to both support the next great scientific hypotheses, but also test their new hypothesis about how to fund bold science. Their funding is multidisciplinary but with an initial focus in the life sciences to ensure they are not exceeding their capabilities as they test their new process. They are looking for projects where there is a lot to learn and that are potentially transformative.

Defining Innovation

A participant asked panelists for their definition of innovation and their investment strategies to advance innovation to markets. Levine referred to the book The Idea Factory,¹⁶ which defines innovation as value capture, when a technology makes it to market to impact people’s lives. She contrasted this with invention, which she described as the inception of the technology. Presentations of innovation to investors depend on the audience. Philanthropists may be more interested in the impact and mission. More traditional investors are focused on capital.

Gamick quoted scientist Sidney Brenner, who said “progress in science depends on new techniques, new discoveries, and new ideas, probably in that order.” Convergent Research is focused on new techniques, such as tools and datasets that allow discoveries and ideas. Sanford defined innovation as needing both newness and transformativeness. Hypothesis Fund’s view is that there has been insufficient risk capital in fundamental research. They concentrate upstream of the translational moment and think of speed in terms of process to funding.

A participant offered a “contrarian view” that there are also relatively low-risk opportunities that need investments and strategic partnerships. Levine agreed. She gave the example of the global laser market, which is worth $16 billion globally—too small to attract interest from investors. However, the technology is foundational to trillions of dollars of downstream industry. She noted innovation ecosystems in East Asia are more focused on grinding out iterative innovations that solve problems with large downstream applications and agreed there is a need to ask what kinds of incentives are needed to do this in the U.S.

Funding Decisions

Levine said there are “hype cycles” for technologies. The investment community needs to have better rigor to filter what is technologically possible and identify and avoid those “predatory” companies without substance in trendy areas. To remedy this, she underscored the need for better infusion of scientific talent in the investment world. Holloway agreed that investors should be cognizant of predatory behavior and the hype cycle but pushed back against the idea that this was frequent. He noted that every portfolio will have failures and urged the community to not conflate failure through risk-taking with predatory behavior. Sanford suggested reframing the concept from “high risk” to “high uncertainty” to differentiate research that is worth doing from an idea with no substance.

An audience member asked, with a stabilization in government funding and increase in philanthropic funding, whether there should be a move away from grants to program-related investments (PRIs) and mission-related investments (MRIs). Gamick said funding mechanisms should be flexible and dependent on the application of the technology. Venture capital expects a larger return on investment, but PRIs and MRIs can be used for investments that offer a good, if not as large, return on many important technologies. The audience member observed that many donor-advised funds are hard for the scientific community to reach. Holloway said communicating tailored, well-thought-out proposals to investors is critical to getting their attention. Sanford said even sophisticated donors with private foundations do not know all the mission-related or program-related funding model options, and educating them would be worthwhile.