Intel agencies advance artificial intelligence in patches, struggle with big breakthroughs

Officials say a more “integrated” approach is needed to truly transform tradecraft using artificial intelligence technologies.

Intelligence agencies are seeing some success in using automation and machine learning for narrow applications, but officials say a more “integrated” approach is needed to truly transform tradecraft using artificial intelligence technologies.

The National Geospatial-Intelligence Agency has been leveraging tools like natural language processing and automation for years to help analyze and share satellite imagery and other intelligence, according to Jim McCool, director of NGA’s Data and Digital Innovation Directorate.

McCool said his eight-month-old directorate is trying to advance NGA efforts in AI, machine learning and computer vision by focusing on customer outcomes.

“What we’re working on today is automating the existing work or the existing workflow or process,” McCool said April 11 at the Intelligence and National Security Alliance’s spring symposium in Arlington, VA.  “And right around the corner is an enormous increase in GEOINT imagery that’s coming.”

“We only have a few areas in which the teams have envisioned moving away from the transactional use of the source, to the streaming, like on Wall Street where they don’t look at the flow of data constantly,” he continued. “When some activity moves outside a norm, or below a threshold or just some measure is met, or a circumstance is met, then that gets the attention of someone responsible.”

Some officials say grassroots AI projects will lead to an eventual, larger scale evolution in how the intelligence community carries out its work.

The National Counterterrorism Center’s Futures Group recently completed a proof of concept on a tool that will allow the center to “try some new stuff” on legacy systems, according to Sarah Hengemuhle, chief of the Futures Group.

“Those small steps are what’s most critical to help that mission customer with their pain point,” Hengemuhle said. “And then we can use those lessons that we learned from that engagement and feed it into development of larger systems. So we’re taking the small steps that we can and trying to amplify those outcomes to the enterprise.”

The National Security Agency is using human language technology in multiple ways, including speaker identification, speech-to-text processing, and machine translation with the ability to process over 90 types of languages, according to James Lampton, who works in the NSA’s Capabilities Directorate.

Lampton said his goal is to “normalize AI” across the NSA.

“Everybody’s familiar with the heroic moonshots and everything like that, but any practitioner will tell you that you don’t really know if something’s going to work until you try it,” Lampton said. “So how do I lower the cost of that experimentation, do it in a manner, in an environment where I transition from pilot phase to experimentation to small operational use to at-scale.”

The Central Intelligence Agency is also updating its AI strategy, according to Lakshmi Raman, chief of AI at the CIA.

“We’re really working toward the whole-of-agency approach towards AI,” Raman said. “We want to go across collection, analysis, operations, digital innovation, S&T, acquisition, legal, finance. We want everybody to feel a part of this strategy.”

But officials conceded they need a better way to communicate their own successes and challenges across agencies.

“How do we in the community that are doing AI quickly come together to learn what each other’s working on?” said Heather Martin, NGA’s deputy director for plans, programs and strategy, data and digital innovation.

“We have all these forums across the IC where we are sharing information on this and we’re working together,” she continued. “But it’s almost not enough. And people are so busy just developing these things that it’s really hard to kind of take a pause and come a level up and say, ‘Okay, let’s get in a room, who’s got what and how can we host it in one place, so that the customer and everyone else in the community can get to it really quickly?’”

AIM initiative attempts to tend a ‘garden’ of AI

The Office of the Director of National Intelligence is organizing the IC’s disparate artificial intelligence and machine learning efforts under a common framework through the “Augmenting Intelligence Using Machines,” or “AIM,” Initiative.

Ryan Carpenter, a program analyst for the AIM Initiative, compared the various AI and ML initiatives across intelligence agencies to a field of flowers.

“The way that the IC has grown, the way that capabilities have grown and moved along, there needs to be a more integrated approach,” Carpenter said. “The field has been planted, but it’s now a disorganized field and now we are trying to make a garden from a field.”

ODNI is helping to develop common standards and API’s to, for example, speed up the notorious “authority-to-operate” security verification process, according to John Beieler, director of science and technology at ODNI.

The aim is “standardizing these things, and then publishing them out and then enforcing those standards across the IC, to say, ‘Hey, here’s how we build things cross agency,’” Beieler said. “So that there’s a model at NGA, it can transfer to CIA and vice versa. So we can build once and apply widely.”

Congress is also asking questions about the intelligence community’s enterprise strategy. The Fiscal Year 2022 Intelligence Authorization Act requires the director of national intelligence to coordinate with agencies and send Congress a plan for an “artificial intelligence digital ecosystem.” The plan is due next March.

The plan should detail the “development and resourcing of a modern digital ecosystem that embraces state-of-the-art tools and modern processes to enable development, testing, fielding and continuous updating of artificial intelligence-powered applications at speed and scale from headquarters to the tactical edge.”

ODNI prepares to roll out talent exchange

Multiple officials said one of their greatest challenges is recruiting and retaining data scientists, a problem that’s further exacerbated by lengthy wait times to get new employees a security clearance.

“It’s really hard to find data scientists, recruiting and getting them in the door,” Martin said. “We certainly spend a lot of time on what is our recruiting strategy going to look like in the next year or so.”

Nancy Morgan, the chief data officer at ODNI, said her team is also looking at ways agencies can retrain their existing workforce.

“I can’t go get enough people fast enough, so what can I do about changing the skill set of the population I have?” Morgan said.

ODNI is also preparing to release details on a Public-Private Talent Exchange that will allow intelligence officers to work in the private sector, and vice versa. Morgan said the exchange will include distinct categories, including AI and data. It will start off with a series of pilots, allowing for exchanges of up to six months.

Morgan said ODNI plans to release a Broad Agency Announcement with further details on the exchange.

“Now we’re going to try it in a pilot phase at a smaller scale, because we’ve got to iron out some process things on both sides,” she said. “We’ve got some lawyers to work through on both sides. But there’s really a lot of energy and excitement about the possibility.”

Copyright © 2024 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.

Related Stories