Not all IARPA programs transition to the field
When word of an Ebola epidemic drifted out of West Africa in 2014, much of the world averted its eyes. But the Intelligence Advanced Research Projects Activity (IARPA) and Dr. Jason Matheny, who came to IARPA as a program manager in 2011 after a long career in epidemiology, looked closer.
At the time, Matheny—now director of IARPA—led a program called Open Source Indicators (OSI). One of OSI’s funded research teams, Harvard University’s HealthMap, sounded an alarm to U.S. health officials, who put a relief effort in motion that included the Intelligence Community (IC) to help quell the epidemic.
“With machine learning and open-source data, social media, newsfeeds, financial market data, and web search queries, we can detect disease outbreaks weeks sooner than with other methods,” Matheny said. “I spent most of my life working in public health in epidemiology. We often had disease outbreaks that we detected too late. Thousands of people would die because we detected diseases too late.”
OSI is just one example of what IARPA has accomplished since it was stood up in 2007. But IARPA success stories aren’t often publicized because most of the activity’s generally unclassified research programs become classified when they transition to the IC, according to a spokesperson from the Office of the Director of National Intelligence.
However, researchers who contract with IARPA to work on unclassified programs are free to publish and develop products of their own. This is how Google obtained facial recognition research conducted for IARPA’s Biometrics Exploitation Science and Technology (BEST) program, previously managed by Michael King, when the company acquired Carnegie Mellon University offshoot Pittsburgh Pattern Recognition.
“The government was able to get usage rights for the technology since we supported its development,” said Dr. Chris Boehnen, an IARPA program manager. “It’s an example of how, even when programs end, we still support them and help our partners continue to use them.”
But not all IARPA programs transition to the field unless research shows they are feasible or valuable enough to continue to completion.
The Synthetic Holographic Observation program was one that fell short of its technical goals. Intended to produce full-color 3D workstation displays that could be viewed simultaneously by multiple people without specialized lenses, the goals proved too ambitious and the program ended early. But as IARPA leadership says, technical failures only prove the activity is carrying out its intended purpose of tackling the Intelligence Community’s most challenging problems.
Return to feature story: Quantum Leaps
Black Cape's Abe Usher and percipient.ai's Balan Ayyar provide an overview on current trends in AI/ML ahead of USGIF's Geospatial Community Forum Nov. 16–18.