NGA has placed automation and machine learning at the top of its list of strategic priorities
For the commercial sector, analytics-as-a-service (AaaS) is a game changer. By delivering spaceborne intelligence to industries that have never had access to it, AaaS companies can offer customers a competitive edge sharper than a Japanese sushi knife. It’s not just the private sector that stands to benefit from AaaS, however; it’s also the defense and intelligence communities—including the National Geospatial-Intelligence Agency (NGA), which has placed automation and machine learning at the top of its list of strategic priorities, Director Robert Cardillo reported at this year’s GEOINT Symposium in San Antonio, Texas.
“If we attempted to manually exploit all of the imagery we’ll collect over the next 20 years, we’d need 8 million imagery analysts,” Cardillo said during his Symposium keynote, adding that NGA already collects with a single sensor—every day—the data equivalent of three NFL seasons recorded in high-definition video. “Imagine you’re a coach trying to understand the strategy of his opponent by watching a game—every game and every play for three seasons, all in a single day … That’s exactly what we ask our analysts to do when we don’t augment them with automation. All this data, combined with dramatic improvements in computing power, represents a phenomenal opportunity.”
NGA is counting on AaaS upstarts to help it seize that opportunity, according to Scot Currie, director of NGA’s Source Mission Integration Office.
“For the last 40 years, NGA has been applying a rather brute-force approach to dealing with all of our various data sources,” said Currie, who called AaaS “the most exciting part of what’s happening in industry right now.”
“We’re extracting value manually out of pixel streams … That’s not something that scales when you talk about moving to the kinds of rapid-revisit suppliers that are coming forth now across the commercial community. So, we’ve got to look at things like machine learning and algorithm development,” Currie continued.
Unlike commercial AaaS customers, many of which lack human resources to analyze satellite imagery, NGA has a deep bench of human analysts it will continue to leverage going forward. As the volume of commercial imagery swells, the agency sees AaaS as an analytic metal detector that will help analysts sift through sand in search of buried treasure.
“We want machines to do what machines do best,” said Currie, adding AaaS will be ideal for rote tasks like counting the tanks and aircraft in a target location, while human analysts will be retained to determine why the military vehicles are there in the first place. “We’re trying to get analysts freed up so they can do higher-order thinking to answer broader questions.”
NGA analysts will still need pixels. Using their machine learning algorithms for automated change detection, however, AaaS companies will be able to flag for NGA analysts which pixels they should look at and when.
Although he declined to name them, Currie said NGA is already testing some AaaS offerings via its Commercial GEOINT Activity (CGA), a joint program with the National Reconnaissance Office through which the agencies are evaluating new commercial capabilities.
“Between us and CGA, we’re building an assessment process that’s going to eventually tell us who the best-of-breed is among these new analytic services,” reported Currie, who said the most important attribute for AaaS providers to demonstrate is veracity. Until providers can all but eliminate false positives, he said, AaaS will be on NGA’s wish list instead of in its toolbox.
“For us to make the decision to move resources [away from humans and toward machines] we’re going to need a fairly significant confidence level … And quite honestly, what’s good enough for [commercial customers] may not be good enough for us.”
Its high standards mean NGA may lag commercial industry in AaaS adoption. But even if the agency moves slower, it will forge ahead, Currie promised.
“I’ve been a part of [NGA’s] commercial imagery team since 2010, when I was the first program manager for the EnhancedView contract that started this whole transformation,” he said. “We saw [AaaS] happening then, but it was only on the periphery. There was a lot of vision, but that vision is now starting to show up in execution and delivery of real capability … We’re excited to become a stronger part of these commercial offerings and to bring them to our customers.”