The Brink: What are some of the NSF’s priorities in computing and data sciences—what are the challenges you’re trying to meet?

Martonosi: The field is really navigating a few key inflection points that are shaping how NSF approaches its priorities for investing in computing. The first is navigating the end of Moore’s law, or what we call the seismic shift. The ability to cost-effectively fit more and more transistors on a chip is getting harder and harder to scale. The second key inflection point is about AI for everyone, what is sometimes called the transcendence of AI. AI is almost as old as electronic computing itself—it’s profound that Alan Turing was thinking about whether machines could think when they could barely compute. AI is not this side, subtopic area. And obviously there’s this wide aperture of impacts: it’s affecting science, society, the economy, national security. How do we fund for this much broader view of AI? And how do we think about engaging the nation’s full talents in advancing AI properly?

The third one is about designing beneficial sociotechnical systems. Where we are now as a field, and as a society, is that computing and information science and engineering permeate so many aspects of how we live, how we transact business—and it’s very hard to opt out. And so we need to think carefully about, not just, can I do this with my technology, but should I do this? How can we achieve the benefits from computing that are so important to advancing society, but with guardrails that make sure that the harms are mitigated?

The Brink: BU recently founded a stand-alone, University-wide, degree-granting computing and data sciences unit with researchers coming together across disciplines to solve big societal problems. Why is that sort of approach so important now?

Martonosi: From an impact point of view, the ability to connect from application areas to computing lets us think more richly about creative solutions that cross the typical departmental lines or domain lines to get better ideas that are more suited for real-world use, real-world impact. From the strategic standpoint, I use the Swiss Army knife analogy: Computing is this set of tools and techniques that are plugged into so many aspects of our life. But if we, as a field, aren’t careful, people will see the places we plug in without realizing how much under the covers we are doing to help them. And so, strategically, it’s helpful if we can actually surface the ways in which our Swiss Army knife of techniques is being employed toward sustainability, toward different aspects of societally relevant problems. First, because I think people will develop a greater appreciation for these tools and techniques and, second, because it can help a broader cross section of the population we want to engage. Some people really get excited about sort of nerding out on particular techniques, other people want to see their ideas have impact on a grand challenge problem like climate change. We need to make sure we have ways of working, ways of engaging with the field that reach out across all these different preferences.

The Brink: Is that something that’s being reflected in how the NSF is managing its own organization and does it have implications for what it may fund in future?

Martonosi: About six years ago, NSF launched a set of 10 big ideas. Those big ideas—things like multi-messenger astronomy, harnessing the data revolution, understanding the rules of life—are not tied to a directorate; they’re truly just: here are some grand challenges we need to go after as a scientific nation, as the scientific community. With those big ideas, NSF developed ways of stewarding investments across many directorates, and those ways have actually been adopted and extended as an operational approach that helps us take on broad challenges. A note of history: NSF was formed in the aftermath of World War II to be the nation’s all-of-science, basic science agency. Other agencies like NIH [National Institutes of Health] have a mission that’s much more focused; NSF is literally astronomy to zoology and everything in between. It’s broad and so that’s an incredible opportunity space for us to work across directorates—and we do all the time—so this notion of the big ideas was one mechanism for working across directorates. So many of our funding opportunities, when you read the solicitation, you’ll see that it isn’t just one division, or one directorate, but it’s a whole list of us that are coming together contributing resources to fund them.

The Brink: Is that part of your message when you come to places like BU or the advice you give to researchers on how they’ll need to work with the NSF in future?

Martonosi: To some degree, I hope they don’t need to know—I hope we will put out solicitations that are interestingly broad and they can simply respond to them and, behind the scenes, we figure out which collection of dollars is going to fund the most meritorious of those proposals. I do always encourage, especially young faculty members, to reach out to program officers to have conversations, to get a sense of the best way to send in their proposal, just so that it doesn’t get categorized in a way that won’t optimize for success. The program officers are really good at knowing how to route things accordingly.

The Brink: Before we spoke, I asked Azer Bestavros, our inaugural associate provost for computing and data sciences, if he had any questions for you. He wanted to hear what CISE was doing to help tackle a brain drain of early- or mid-career researchers from academia to industry.

Martonosi: Azer is on the CISE advisory committee and he’s an extraordinary force for the community in that role, and in any other role—so wonderful. When I think about what keeps the community up at night, what keeps people like Azer and me up at night, one of these issues is how do we make academic scholarly research rewarding and impactful for the next generation. One key thing that I’m focused on is infrastructure. I believe passionately that if we can get academics access to the right data and compute resources, that can help level the playing field between different institutions, and then also between academia and industry. And so we invest heavily in cyberinfrastructure for science: resources like specialized computers, like discounted and facilitated access to a commercial cloud. The data side is even more challenging and so we’re looking at a range of ways to facilitate better access to data and to federate data that comes from many different sources. For example, NSF has a set of major facilities—telescopes, environmental sensing sites, research vessels—and our vision is that you should be able to grab data from any of those facilities and understand whether there are scientific opportunities to be had by combining that data.

The Brink: Computing and data sciences is a particularly white and male-dominated field. What steps are you taking to help diversify it?

Martonosi: NSF has long had two primary factors by which its merit-review process runs—one is intellectual merit and the other is broader impacts. One of the potential subtopics of broader impacts is this notion of broadening participation, diversifying. We cannot advance computing as fast or as well as we need to if we’re only using about 30 percent of the US population. We need to change that for the good of the field, and for the good of the results that come out of the field, for the nation and the world. One of the things that CISE has done over the past close to five years now is our broadening participation in computing plan effort; if you write a CISE proposal and you ask for more than $600,000, we ask you to write what’s called a BPC plan, a broadening participation in computing plan.

CISE puts a lot of work into the BPC efforts. We invest in what are called BPC Alliances, which are large-scale sustained efforts on different aspects of broadening computing, some of which are looking at gender or racial or ethnic demographic groups, some of which are looking at things like improving computing as a field that welcomes people with different accessibility and disability issues. We also invest a lot in curricular issues, some of which have strong aspects of diversity and inclusion: how can we structure intro computer science classes that don’t presume you’ve had a computer in your room since age three, how can we broaden who we pull in at the undergrad [level], at the associate’s degree level, and at the grad level throughout career pathways to finally change what has been a very recalcitrant set of data?

The Brink: Lastly, any message to BU community members to encourage them to come along and speak with you when you’re here?

Martonosi: I have done close to 80 virtual campus visits in 48 states—that’s how I went through the pandemic. One of the things that I really appreciated about those virtual visits was a chance to have conversations with a wide range of students and faculty in this broad set of what I would call the CISE-relevant topic areas, which goes way beyond computer science. Although this one is not virtual—this one is for real—I still would love to have that kind of conversational tone. If people want to come, bring their questions, bring their ideas, that would be great.