If artificial intelligence (AI) and robotic process automation (RPA) were air wings, they’d be flying high. For the Air Force, no less than the other armed forces, the question is not whether to adopt these technologies. It’s where to adopt them first so they yield the greatest return.

“We are in a lot of difference businesses,” said Jay Bonci, the Air Force’s chief technology officer. Those include traditional airpower missions, space operations, weather modeling and prediction, human performance and health planning, and financial management.

“Each one of those areas is a data rich environment,” Bonci said. “We are right now working on making sure that we can get that data into our large data platforms, so that it can be cleaned and indexed and be ready for different communities of analysts.” He added that much of the data is bound for “fusion, for things like AI and machine learning models.”

Bonci spoke on a Federal News Network panel to discuss defense uses of AI and RPA.

When it comes to the information technology domain, National Defense University’s Dr. Joseph Schafer, professor in the college of information and cyberspace, said the staging and provisioning of IT resources is ripe for RPA. In particular, cybersecurity incident detection and response requires AI, Schafer said, because of the volume of attacks and the log data that security operations centers must deal with.

“DISA [Defense Information Systems Agency] gets hit by one and a half billion cyber incidents every single day,” Schafer said. “There’s no way you you could tackle those kinds of incidents without leveraging some artificial intelligence.”

Regardless of the specific AI/RPA application, in many ways these technologies are about people. Melissa Long Dolson, the vice president of worldwide technology sales for AI Ops and Integration for IBM, said AI “is actually infusing human into intelligence into the process of automation. How do we help take something that I learned, and that I’ve been doing all day long in my job for the last 10 years, and build that into a tool that allows me to do it faster, quicker and better?”

Especially when AI applies to workforce management and development, Dolson said, AI systems must at all times ensure protection of personally identifiable information.

Beyond PII, she added, “I think the DoD agencies in particular have a greater challenge… it’s about all of those secrets that we have to protect, to make sure that we are handling our national security interests in the right way.”

Infusing human knowledge into an AI program, panelists emphasized, augments a person’s decision-making ability but doesn’t replace it. For the armed service, Bonci said, ethics centers on the kill chain. An ethical deployment of AI retains human judgement in decisions to fire and what to aim at. This ensures kill chain decisions remain accountable and in compliance with international conventions and U.S. military standards.

“There are many different dimensions around that concept of AI ethics,” Bonci added. “It does apply to weapons, it does apply to the human factors, it does apply to personnel data and how we use and fuse the different sources.”

Another dimension, he said, is varying levels of data classification, with the challenge being how to fuse high and low level data for AI and analytical purposes.

“There are a lot of efforts at the [Office of the Secretary of Defense] level, looking at how we do our data classification and our data protection,” Bonci said.

Ethical considerations and the kill chain even make up part of the NDU curriculum. “At the end of the day,” Schafer said, “nobody is for Terminator,” Schafer said, referring to the Hollywood version of a robotic hitman with autonomous intelligence.

Pick your AI battles

With no shortage of processes that could benefit from AI and RPA, it may be wise for an agency to obtain an objective, third-party view of its workloads to find the best candidates.

In helping clients with where to go first, Dolson said, IBM consultants will survey and assess workloads to determine “where those AI solutions and robotic process automation solutions can benefit them the most quickly, and do so in a way that does treat the data properly, and is ethical by consideration.”

Bonci said that essentially, the Air Force looks at big applications with big returns first.

“The places where we need to think about applying AI and ML, and the places where we are starting down this road, are all the areas that involve the true scale of the military,” Bonci said. That ranges from AI for managing weapons systems to the challenges of cyberspace.

“There’s never going to be enough people who are able to respond to an automated attack,” Bonci said. “To be able to rise to meet that challenge, we’re going to need automated defenses that can make sense of highly noisy, highly, varied traffic with the intent to confuse and deceive.”

Bonci added that the Air Force has established an enterprise service to curate, secure and fuse data at scale for training purposes. He pointed out that in some functions, such as optimizing fuel, packing airplanes efficiently, and scheduling crews, the Air Force resembles commercial carriers. Therefore, outside of the kill chain at least, industry and the military can potentially share training data and AI approaches to common challenges.

Dolson said this type of cross-domain cooperation can help on two fronts.

“We have to share the air across industries, because I think that’s one way to enable AI in your organization.” Dolson said. “But it’s also a way to help manage that cost that comes with AI.”

For the Air Force, AI and RPA are components in the ongoing effort of digital transformation. Therefore, Bonci said, agencies including his own need good strategies for cloud computing, identity management, non-person access, and secure application programming interfaces among systems. And, of course the human skills necessary to make it all work.

Moreover, said Dolson, modernization “starts in one place. And that’s really understanding your processes. Because if you automate a process that’s bad, it’s still a bad process.” Better, she said, to get processes right, automate them, then scale them in a cloud infrastructure.

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