Federal Funding Supports the Flow of Innovation

What’s the return on investment of federal research funding in computer science? As politicians weigh proposals that call for steep cuts to the National Science Foundation (NSF) and the National Institutes of Health, we spoke with someone who can answer the question more precisely than most. Elizabeth Mynatt, Dean of the Khoury College of Computer Sciences at Northeastern University, co-chaired a 2020 report that analyzes how IT research impacts economic growth. The latest in a long-running National Academies series, it explores the many ways that research innovations become impactful commercial activities.

The so-called “tire track report series” dates back to 1995, when the National Research Council’s Computer Science and Telecommunications Board produced the report Evolving the High Performance Computing and Communications Initiative to Support the Nation’s Information Infrastructure. Can you describe how that effort got started and how it has continued to evolve?

Current times are fairly unique, but there has always been this myth in Congress: “Look at Silicon Valley, look at all these tech companies that got started in a garage. Why do we need to fund computer science research? This is a self-sufficient thing.” The first version of the report was produced in the dotcom heyday, and a graphic in that report—often called the “tire tracks” diagram because of its appearance—provoked an extraordinary response by clearly linking government investments in academic and industry research to the creation of new IT industries with more than $1 billion in annual revenue.

Both the report and the diagram were updated in subsequent years to depict more clearly the flow of innovation within and across academic research, industry research, and commercial activity.

The original idea was to show some crystal-clear examples of how a new consumer technology, which feels like it’s five minutes old, is actually based on decades of research. The 2002 report broadened these examples, and by 2009, the report detailed the uptake of computing innovation into a growing cadre of IT companies that had an estimated $500 billion in annual revenue.

You chaired the most recent version of the “tire tracks” report, which was released in 2020 and differs from its predecessors in a few important ways.

Our main goal was to move from anecdote and example to showing the whole ecosystem. We had a significant time span, now 60 years, to look at larger patterns. We tracked down over 800 examples depicting the flow of ideas between academic research, industry research, and commercialization. Also, tech companies had become more complex. When you look at research tracks, you no longer see something that starts at one end and a company pops out the other. Instead, companies like Amazon, Google, and Microsoft all have multiple offerings that integrate across multiple tracks.

So you and your colleagues tried to go beyond discrete research contributions and trace the more—and more complex—ways that companies are consuming them…including, for the first time, in industries like healthcare, automotive, and even agriculture.

There’s a lot of talk about how other industries have been “transformed” by computing, but it’s not like we computer scientists walk around with a big wand of transformation. Rather, there are people in the middle who act as innovative translators. For example, modern agriculture is actually quite dependent on computing innovations. Researchers at Cornell, Georgia Tech, and countless other universities have worked through “extension” programs that extend into their respective state industries. The university’s dual role of research and education provides fertile ground for technology transfer to industry.

In the report, you identified two key patterns to describe the interplay between research innovation and commercial activity, which you called resurgence and confluence. Let’s start with resurgence. What is it?

Resurgence describes a pattern where economic return follows a period of diminished interest and investment in an area—followed by an uptick of new ideas and enablers. AI is the classic example, and the report has a specific chapter that details multiple AI summers and winters. But many other research areas follow the same pattern, including virtualization, VR, and formal methods.

And confluence?

Confluence describes a pattern in which IT innovations combine with deep domain expertise, design and production knowledge, and new business models to create transformative results in other major sectors. The report includes examples from healthcare, automotive, agriculture, and even sports—where wearables, data analytics, and mixed reality have had huge impacts.

You visualized these downstream effects in the form of “subway maps” that are connected to the classic research “tire tracks.”

Princeton’s Margaret Martonosi was the one who nicknamed it “subway maps.” We incorporated more than 800 data points into a diagram that depicts the flow of innovation and consumption. The IT companies we illustrated represented just shy of a trillion dollars in annual revenue in 2019. The companies in non-IT sectors represented an additional $1.3 trillion in annual revenue. The “classic” tire tracks now transition into the subway maps of confluence as innovations are integrated into different industry domains.

In other words, a very small investment on one side yields a very large amount of dollars on the other.

Yes, the basic math summarizing this report is that a relatively small amount of funding leads to exponential returns that continue to grow. These developments can take a long time, but they show up in these really important ways in U.S. industries across the economy.

You’ve made the point before that this research ecosystem is both robust and finely tuned.

As industries began to need more information technology capabilities, they evolved their processes to attend relevant conferences and creatively recruit new talent. As one example, in the report, we talk about how automotive executives wanted to hire VR talent, but they couldn’t just go to a top 20 campus and say, “Come work for us,” because everybody wanted to launch their own gaming company. Instead, they would stalk these gaming startups and come back in three years to swoop up talent from the places where things hadn’t worked out.

A lot of other industries have learned how to stalk computing talent, too.

They look at national labs, they look at startups, and they figure out how to get their people. People are a great way to demonstrate how information flows from one part of the ecosystem to another. Intellectual property matters, open-source matters. There are all of these things that matter, but in reality, it is the dual mission of the university that we educate and do research that turns out to be a highly efficient way of funneling innovation.

Let’s talk a little more about why you feel that industry can’t just take it from here in terms of funding.

For me, the headline is that just because folks are spending a lot of money on AI, it doesn’t mean the rest of the ecosystem—healthcare, transportation, agriculture, manufacturing—is going to be taken care of with the same approach. Today’s ecosystem works due to a partnership between academic research and industry that’s mediated by funding from an independent set of agencies. We can’t confuse the amount of money that’s being invested in AI with the value of those long-term investments.

Our research ecosystem has also made many contributions to society that are not quantified in terms of annual revenue. For example, academic research has played a tremendous role in creating assistive technologies that provide access to workplaces, services, and public transportation. There’s no way we can expect for-profit companies to ever do that kind of work on their own.

Around 80% of basic research funding in CS comes from the National Science Foundation. Industry picks up very little in comparison.

Data like that are crucial to have when petitioning members of Congress. Are there other things you feel the computing community should be doing to advocate for the health of the ecosystem?

I don’t think we’ve done a good job in telling the full story. The impact of our research is pervasive across the U.S. economy. The university’s dual mission of basic research and education is foundational to these short and long-term economic benefits, and there is no feasible way to take a shortcut and outsource this work to industry. We have to explain the benefits of our work in terms of lived experiences, namely safe and convenient online commerce, connectivity to work, family and friends, safe and affordable food, groundbreaking and lifesaving health technologies, and safer and more efficient cars. We have over six decades of evidence to show that this powerful ecosystem works.

What else should the computing community be doing in response to challenges to federal funding?

One of the lessons I think we need to take to heart has to do with the relevance of computer science innovation to other industries. Historically, there’s been such massive growth in our field. Universities didn’t really have to do much besides take in smart kids, train them, and hand them off to a growing set of industry leaders, while receiving over 80% of our academic research funding from one place. Whatever the future state will be, it has to include more diversified research and training that matches our field where it really is. That may not meet the challenge of the moment, but it will at least point us in the right direction.