AI, Commercial Firms Key to Future of GEOINT

iStock photo

The United States relies on its constellations of spy satellites to provide decisionmakers with critical information and data. But why should taxpayers fund advances in collection capabilities if the United States cannot process, exploit and disseminate all of the raw data it already has?

Most of the raw data collected by the intelligence community is not processed, which arguably defeats the entire purpose of having extensive technical collection systems. Geospatial intelligence processing involves the conversion of “data into a useable form or formats suitable for analysis, production and application by end users,” according to the Joint Chiefs of Staff’s Joint Publication 2-03, “Geospatial Intelligence in Joint Operations.”

The value of time sensitive data collection is limited if the material is not processed quickly. However, providing more efficient GEOINT processing can be achieved through commercial partnerships and the use of artificial intelligence.

In an interview, Mark Lowenthal, former assistant director of central intelligence for analysis and production from 2002 to 2005 and author of Intelligence: From Secrets to Policy, noted that the National Geospatial-Intelligence Agency has been using a variety of techniques to help process increasing amounts of imagery.

Bolstering GEOINT commercial partnerships can help to process most — or perhaps all — of the raw geospatial data collected by the intelligence community. NGA and the National Reconnaissance Office have already partnered with some commercial organizations, and these efforts are likely to expand in the future. While there are no plans by the NGA or NRO to list all their potential partners, larger contracts are anticipated to be negotiated within the next few years.

The NRO is sponsoring studies to determine the extent of companies’ respective capabilities and determine which firms can provide reliable support to the agency’s mission. The next step will be establishing procurement contracts that provide a clearer picture of the future partners for the spy organization.

Lowenthal noted that the NGA created a project called GEOINT Pathfinder, which sought to answer key intelligence questions from unclassified sources only. One result has been NGA Tearline, which involves NGA partnering “with expert private groups to create public-facing and authoritative open-source intelligence on various strategic, economic and humanitarian intelligence topics that tend to be underreported within in-depth or long-form formats.’”

While commercial imagery may have lower capabilities compared to the satellites operated and maintained by the NRO, commercial geospatial organizations can provide imagery on the same location more often. The improved consistency of imagery can provide information on dynamic changes in areas of interest.

“The intelligence community has been committed to using commercial imagery whenever possible for many years,” Lowenthal said. “Not every issue requires exquisite imagery — although commercial imagery today is very good. By using commercial imagery where possible, we reserve the classified systems for the issues where they bring the needed intelligence.”

Some intelligence-gathering methods require more data or processing power than others. For example, activity-based intelligence, also known as ABI and sometimes referred to as the “pattern of life,” requires an extensive amount of data, Lowenthal said. ABI intelligence collection is based on observed behaviors that are more likely to indicate that an activity of interest is taking place in that location. It can look either for activities that seem to differ from the norm in a given location or for patterns that indicate an activity — such as teams planting improvised explosive devices, he said.

Besides utilizing commercial industry, artificial intelligence will also provide ample opportunity for growth in the efficiency of GEOINT processing. Analysts need to be able to integrate increasing levels of data from multiple technical sources but are impeded by modern inefficiencies such as time limitations that hinder effectiveness. AI has exhibited the potential to be more efficient at processing big data by recognizing patterns and relaying that information with similar attributes to a human source for analysis.

However, “there are now concerns about images being manipulated by AI to falsify the information in the image,” Lowenthal said. “The United States has placed export restrictions on certain AI programs that search for ‘points of interest’ in geospatial imagery. NGA is also looking at AI to detect false imagery. But remember, all of this AI has to be programmed by humans.”

By allowing analysts access to narrowly defined data sets determined by AI algorithms, there are greater possibilities for breakthroughs that will provide a strategic advantage to war­fighters and policymakers. Although artificial intelligence has tremendous potential to provide innovation to modern practices, it can never completely replace the role of a human in every aspect of processing. AI does not — and may never — have a firm grasp of situational understanding that provides insight and intuition in complex scenarios, which may prevent prolonged development of this potentially innovative solution.

As collection capabilities continue to increase, partnerships between the intelligence community and private industry could provide much needed improvement to GEOINT processing capabilities. In the future, the leveraging of commercial systems with AI and big data analytics could facilitate improvements in the efficiency of such processing.

Ultimately, this capability will be key to developing timely, relevant and actionable intelligence at scale to take advantage of the operational concepts envisioned for the future such as joint all-domain command and control — capabilities that may keep the American warfighter advantaged for another generation.

Nicholas Lapinski is an NDIA junior policy fellow.

Topics: Robotics, Robotics and Autonomous Systems