The Intelligence Community Is Developing New Uses for AI

U.S. intelligence agencies are applying AI and machine learning (ML) technology to boost efficiency and effectiveness across their mission areas. Geospatial intelligence uses imagery and location data to detect human activity.

NGA makes it safer not only to navigate the oceans but also to fly aircraft, fight in wars, locate targets and respond to natural disasters.

NGA recently set up a digital, data and innovation directorate to take a greater role in managing and sharing developments in AI across the intelligence community and Defense Department to enhance interoperability, Daniell says.

“This new directorate is a reflection of how seriously the agency views AI/ML technology and how valuable we think it will be to meeting our most pressing operational challenges,” she says.

What Does Artificial Intelligence Look Like in Federal Agencies?

Traditionally, NGA has put AI to use in specific niches, such as determining whether Arctic ship lanes are covered in snow or ice. “Now, we’re seeing enormous interest in other areas,” Daniell says.

Agencies are now using AI for natural language processing, as with SMAPS, as well as “digital engineering, acoustics and different varieties of data that are not necessarily image-based,” she says.

At the National Security Agency, human language processing is the “poster child of success” in applying AI to operations, says James Lampton, a leader in the NSA’s Capabilities Directorate, during a panel discussion at the Intelligence and National Security Alliance’s spring symposium.

LEARN MORE: How will the intelligence community invest in artificial intelligence?

AI has enabled speaker identification and speech-to-text processing. “This isn’t a hypothetical; we’re using it now,” Lampton says. “We have machine translation mechanisms that serve thousands of users across the intelligence community, the U.S. military and the government, being able to process over 90 different languages.”

The CIA has contracted with cloud computing and infrastructure platform providers including Amazon, Google, Oracle, Microsoft and IBM to support its AI efforts, says Lakshmi Raman, the CIA’s chief of AI, who also spoke at the INSA symposium.

“They offer that backbone for our officers to develop and deploy their own AI and data analytics tools when necessary,” she says. “I think those managed services available via some of those cloud providers are going to be key to how we apply AI to some of our highest-priority mission problems.”

What Can Artificial Intelligence Do for the Intelligence Community?

The SMAPS program relies on natural language processing, or NLP, which is how a computer can understand text and words.

“It has cut in half the time that is needed to process those incoming incident messages and to write alerts,” Daniell says. “An incident message will come in, and we’re able to generate an alert to the mariner within half the time that we were previously, before SMAPS.”

From an intelligence perspective, NLP is crucial for staying one step ahead of the nation’s adversaries, she says. AI and ML pull information from a vast array of sources, improving the precision and reliability of collected data, Daniell says.

DIVE DEEPER: Artificial intelligence is just one of the technologies changing government.

Daniell’s research department recently studied how NLP can assist intelligence analysts with the reams of text and information that flow across their desks. NLP’s capacity to process billions of words in seconds automates the “sifting, sorting, translating and comprehending of all of this data,” she says.

Like SMAPS, this use of AI helps humans do their work faster.

“We have computer vision models to rapidly geolocate images that have no existing geospatial information,” Daniell says. Think of it as “Where’s Waldo?” technology for intelligence analysts, who might uncover an image from an iPhone and want to find where that building or person is located.

“It’s just a photograph and it could be anywhere in the world,” Daniell says. NGA has used this technology in beta test form and is now transitioning it to operations.

The Future of AI in Government

In 2018, the Office of the Director of National Intelligence launched the AIM Initiative, a strategy for augmenting intelligence using machines, to encourage intelligence agencies to organize and share their AI efforts.

“We can start to build synergy across the community in a way that actually makes sense,” Ryan Carpenter, an AIM program analyst for the ODNI, said during the INSA symposium.

The technology has advanced exponentially across the intelligence field, and agencies rushing to develop applications haven’t had much chance to look outside of their own environments, Carpenter says.

“They’re not necessarily talking to each other, so ODNI takes that role and tries to integrate across those different silos,” he says.

REVIEW: How to prepare a federal network for AI.

The intelligence community sees enormous potential to apply AI and ML technology broadly to the complicated challenges of national security, but it’s still in the early stages of that process, Daniell says. Some programs move from development to operational use faster than others, she added.

“What we’re very focused on in research is accelerating that science-to-mission timeline,” Daniell says. “Sometimes it takes years, but we’re focused on that mission objective.”

Raman says she took on her new role at the CIA to create and update the agency’s AI strategy in a more focused, mission-driven way.

“We’re really working toward a whole-agency approach toward AI,” Raman says. The technology has relevance for data collection, analysis, digital innovation, operations and even legal and finance areas, she says. “We want everybody to feel a part of this strategy.”