What NGA’s Data Brokerage Means for Industry
The National Geospatial-Intelligence Agency (NGA) has a lot of data. Think of it this way — from just one sensor, roughly the equivalent data of three full NFL seasons recorded in high-definition video — every day. It would require millions of analysts to exploit that amount of imagery; a daunting task when your job is to keep the nation safe from adversaries.
Kristin St. Peter, NGA’s new deputy associate director of capabilities, recently spoke to an audience during the United States Geospatial Intelligence Foundation’s (USGIF) event about the agency’s latest initiative – creating a data brokerage with industry in hopes that the commercial sector can use artificial intelligence (AI) and automation to train computers to search data and help analysts identify anomalies faster.
“It takes 100,000 images to be able to train an algorithm to spot a plane, train, or automobile with a 90 percent level of accuracy,” St. Peter said. “We have those data sets and want to be able to use them to partner with people who normally don’t partner with government.”
The pain point for industry startups is not necessarily getting venture capitalist funding, but rather gaining access to the training data they need to develop these algorithms and tools. St. Peter hopes this brokerage will help NGA drive its mission forward, and allow industry to create something meaningful and valuable to their national security imperative.
Industry is ultimately left to question what a brokerage like this means for them. During the question and answer session, St. Peter was asked for examples in which data brokerage models have been successful and how such a partnership would work in terms of finance and logistics, as NGA has never engaged in a brokerage like this before. She explained that exchanging training data for algorithms and other AI tools would allow industry partners to retain their intellectual property, while NGA would retain intellectual property rights in terms of a partners’ ability to sell the data.
“We would place caveats on who you sold it to. We don’t want to give [our adversaries] the training data needed to outpace us in AI,” she said. “We need to be selective in how we pick our partners and how they use the IP. We would place a few restrictions on it but wouldn’t diminish the IP to the point where it’s not worth your while.”
Companies like OGSystems have been asking for data like this for years now, as another step in the democratization of system and application development for Agencies like NGA. The move to separate infrastructure from application development via the cloud (and ISP/ASP previously) started the movement towards allowing small and mid-sized companies to contribute innovative tools in a field once dominated by a handful of very large integrators. According to , OGSystems’ President, NGA is further cracking the door for companies like OGSystems, who are already working on tools that could further benefit from testing against exotic data sets like those possessed by NGA.
“We applaud NGA for their desire to expand partnerships with industry, as we already have commercial partnerships in place to get access to this type of data,” Pagon said. “For example, when you look at the analytics platform, it is predicated on accessing unclassified data and algorithms that exist in the open, with the ability to use APIs to connect to them quickly. This private-public partnership could allow for more robust development of similar analysis as a service toolkits that are compatible with both commercial and government needs. The challenge for NGA is going to be in providing unique data relative to commercially available analogs, and in figuring out IP rights, as it will be tough to create a business case if NGA can restrict use.”