The SETI Institute, a Silicon Valley-based nonprofit that seeks to explore and explain the nature and origins of life in the universe, is gearing up to host the fifth iteration of its competitive, NASA-funded summer program, the Frontier Development Lab.
FDL brings together a diverse cadre of researchers each year since its inception to rapidly leverage artificial intelligence, machine learning and advanced computing capabilities—all to ultimately help America’s space agency accelerate its own research and discoveries. While SETI envelops the “search for extraterrestrial intelligence,” its inside efforts touch a range of areas across space, science and beyond.
But the 2020 program might run a little differently than those that came before.
“Now, what’s interesting is, we may—for the first time actually—undertake the program virtually because of the COVID-19 pandemic,” Bill Diamond, president and CEO of the SETI Institute, told Nextgov recently. “All indications are that this is going to be with us through at least the early part of the summer, and it may preclude the in-person working system that normally is characterized by the FDL program.”
Diamond detailed how FDL has evolved over the years and how the public-private partnerships that undermine it have helped power space-driven revelations in rapid timeframes. He also offered a glimpse into how SETI insiders are preparing to potentially transform the program in an effort to adapt to the global health crisis caused by the coronavirus.
Roughly 85% of SETI’s annual operating budget is NASA-funded research, Diamond said, and the institute also receives some financial support from the National Science Foundation and others. There are more than 100 people with PhDs presently working for the institute and they represent about 23 different scientific backgrounds.
“We’re involved in everything from basic cosmology and astrophysics, to planetary science, planetary geology, to exoplanet work,” Diamond explained. “The Kepler mission that NASA launched was primarily all data analytics and the data pipeline for that mission was managed by the SETI Institute.”
The AI-focused FDL program first came together in 2016, when it was initiated by the Chief Technologist Office at NASA headquarters. Diamond said the idea behind it was threefold. First, officials aimed to figure out “what might be the efficacy of” applying machine learning, AI and deep learning tools and techniques to NASA science priorities. Second, they wanted to explore whether teams of young researchers from a range of science backgrounds could be put together in an interdisciplinary team structure and “do something meaningful in a very intense summer work program.” And third, officials sought to determine if private industry could also participate and collaborate with the agency in new ways outside of traditional contracting.
When NASA approached SETI and asked it to host the program in partnership with NASA Ames Research Center, Diamond said he was “excited by the idea and delighted to do it.” Funding for FDL has come from different NASA offices and mechanisms since its inception, and the space agency recently indicated that it aims to support the program for five more years, Diamond said.
“Honestly when the program started, surprisingly perhaps … basic research, pure science, NASA research and others had not—and still have not—fully embraced machine learning and AI and understood what those capabilities can bring to the table for scientific research,” Diamond said. “So the FDL program has been a great way to demonstrate the power of AI and ML for basic research and basic science.”
Applications for this year’s program are open until April 14. The top applicants that are chosen to participate will be broken up into different teams of four researchers that will be assigned a certain set of challenges and ultimately participate in what SETI calls an “intense eight-week concentrated sprint.” The challenges are divided into different domains such as heliophysics, astronaut health, planetary science and more and the teams themselves are generally made up of young, early-career scientists, from different countries and different research backgrounds, Diamond noted.
“One of the things we say about the FDL program is ‘the magic happens at the intersection of disciplines.’ So when we get computer scientists, together with domain scientists, that’s when really exciting things happen,” he said. “They inform each other, and teach each other and learn from each other—and create incredible results as a result of that.”
He added that participants in the past have beamed over the opportunity to expand out of the comfort zones of their normal expertise, to dabble in something new. “I remember how much fun it was talking to this young woman whose background is in machine learning,” and economic modeling,” Diamond explained. “She said ‘now I’m helping do image analysis of the lunar surface that might eventually help the rover self-navigate on the moon—you know, how cool is that?’”
Each of the teams is also supported by middle-career expert computing and science “mentors,” which Diamond explained come from NASA, academia and private industry. FDL is inherently a public-private partnership, as its site indicates, with NASA in the U.S. and European Space Agency, as well as other partners including “NVIDIA, Intel, Google Cloud, Mayo Clinic, IBM, Lockheed Martin, Luxembourg Space Agency, XPrize, Kx and USGS who provide expertise and the computing resources necessary for rapid experimentation and iteration in data-intensive areas.”
Diamond said that Google, in particular, has been working with the program for the last several years, and that “that in itself has been transformative” for FDL. The tech giant’s involvement has cut across many of the past team’s work. For example, in one of the past projects, by “having access to Google Cloud compute capabilities, and advanced data analytics algorithms, machine learning and AI algorithms, we were able to speed up the time it takes to discover planets, to improve the accuracy through which we discover planets in the test data, and just overall improve the efficiency and efficacy of the science,” Diamond said.
Making it into the FDL program is pretty competitive, he added, noting that SETI will likely “have hundreds of applicants for a couple dozen spots.”
“I mean, the program is kind of harder to get into than Harvard or MIT at this point,” he said. But those who make it have the opportunity to connect with major industry and academic players and tap into NASA resources to drive new discoveries for the space agency. Diamond added that many of the teams have gone on to publish peer-reviewed articles for scientific journals detailing their FDL work.
“It’s very rare to see anybody do any work in a period of just eight weeks that ends up being capable of being published in a peer-reviewed science journal,” Diamond said. “It’s very different in that regard, and it demonstrates the power of the project.”
Still, the planning efforts for this year’s iteration of FDL, like everyone else’s plans across the globe, have been disrupted by the introduction of the novel coronavirus. Yet Diamond’s approaching the challenge from a bright side.
“I think we’re all going to learn a lot in this situation about how to maximize our ability to continue to do our jobs,” he said.
The institute on Monday will hold a 2020 Challenge Briefing to provide information about this year’s summer program—and address SETI’s aims to adapt to the new barriers posed by the pandemic.
“We are still in the planning process for what a virtual FDL program might look like, but nothing is yet cast in stone,” Diamond said. “We are considering adding a week to the program schedule to compensate for a virtual collaboration environment for the researchers, and will also have a more structured pre-workshop phase to the program.”
The institute is also considering potentially using tools such as Google Docs and Hangout, Zoom, Slack, DropBox, to make collaboration more fluid, and they might even explore WebEx and Microsoft Teams. Diamond said they also plan to make extensive use of cloud-based assets for both compute and data storage and “anticipate the Google Cloud Compute credits will once again be a vitally-important resource to the research teams.”
“A variety of tools will be used, and some of this will be up to the individual teams based on what they are familiar with and comfortable using,” he said. “We don’t want to constrain teams with tools that may be unfamiliar, thus adding more burden to an already challenging environment.”
Yet, at least as for now, it’s all to be determined.
“This year is going to be really interesting and I think it’s going to test the boundaries and the limits of team-working in remote environments,” Diamond said. “But we plan to go full steam ahead.”