DOD Tech Leaders Highlight What’s Next in the AI Wave

Federal and industry leaders from the Defense Department presented their insights and predictions on how artificial intelligence will transform the future of national security.

These AI defense leaders spoke during a lightning round session at Friday’s Defense IT Summit in Reston, Virginia.

• Amanda Bullock, AI Lead, Air Force Research Laboratory
• William Streilein, CTO, Chief Digital and Artificial Intelligence Office, DOD
• Angel Phaneuf, CISO, Army Software Factory
• Jaime Fitzgibbon, AI/ML Program Manager, Defense Innovation Unit
• Bharath Nagaraj, Field Technical Director of AI, Cohesity

Each panelist provided insights on AI and how it’s currently being used in their departments.

‘Cheats Time’

“We’re looking to insert AI into everything we do, from combat operations, to reducing the toil and administrative tasks. We define toil as anything that keeps our people from performing the mission,” Bullock said in her remarks.

She acknowledged that AFRL is navigating its relationship with AI and machine learning, and said it has been using AI to “cheat time.”

She told the story of a dashboard that needed to be cleaned up and combined manually. Even after some macros were created by the inspector general, the process still took eight hours a month for the team to complete.

“How could they cheat time? By creating an automated data workflow that required only a single upload of data. No more merging spreadsheets running Excel macros than uploading the resulting spreadsheet to Tableau,” Bullock said. “They now have a visualization in seconds. It was so well received the other [inspectors general] across the Air Force are looking to adopt it.”

Bullock said that during last summer, AFRL has successfully demonstrated the first ever flight of a machine learning-trained AI algorithm. The algorithms were matured for millions of hours of simulation events, hardware in the loop and ground test operations.

AFRL leaders demonstrated that AI machine learning can perform aerial operations autonomously, resulting in much safer and efficient operations in the future.

“The future looks bright, but to stay ahead of our adversaries and keep the fight unfair, the Department of the Air Force will need to accelerate its capabilities in artificial intelligence and machine learning,” Bullock said. “Leading that charge will be AFRL. … What’s next in AI and machine learning is an environment where we save more lives, save more money and cheat time.”

‘Responsible Agility’

DOD needs to practice responsible agility while it accelerates the adoption of data analytics and AI, Streilein said in his opening remarks. Experimentation and capability at scale are key.

“First off we’re driven by experimentation,” Streilein said. While it is imperative that experiments iterate quickly and need to “fail fast,” the terminology has a negative connotation. Streilein prefers the phrase, “iterate and learn.”

“It means you realize you’re going to try something out and take away from that learning but also leave behind capability there,” he said.

The department will scale capabilities by using scaffolding for data labeling and the replicator initiative. The department takes the approach of the hierarchy of needs in terms of leveraging AI. This means taking care of data first. Data is critical for algorithm and AI development.

DOD is prioritizing responsible AI development through problem-driven data management, talent management and product management, Streilein said.

Future Force

The Army Software Factory protypes the Future Force design, according to CISO Angel Phaneuf.

“[AI and machine learning] are tools that empower our soldiers to make data-informed decisions, revolutionizing the way we approach and execute the mission on the battlefield,” she said. The Army Software Factory selects 25 military and civilian candidates biannually to learn product management, design software development and platform engineering.

Data is a strategic asset. AI and machine learning technologies can give the ability to forecast the enemy’s movements and potential threats in the DevSecOps framework. These technologies can identify patterns, threats, and enable commanders to anticipate and prepare for various scenarios.

On the warfighter level, these technologies empower soldiers to make tactical decisions aligned with overall strategic objectives. AI alleviates that burden of cognitive load on the warfighter. By automating routine tasks and data analysis, soldiers can focus on higher level decision making and strategy on the battlefield based on generated insight.

‘New Version of Oppenheimer’

It is the nations that adopt AI that are going to consume and move forward and faster than those that do not, said Jaime Fitzgibbon, AI/ML program manager for the Defense Innovation Unit.

“We’re in a new version of ‘Oppenheimer,’ we are at a race to get to AI faster, but not just get to it, get to it where we can actually ingest it,” she said.

While AI is not new, the attacks and implications are becoming more tangible and real. Data and processing power are the fuel that powers the machine.

Defense is moving from 2D visualization molecules to 3D. This change brings in bio-mapping capabilities that enable personalized health and medicine, affecting veterans as well as warfighter readiness.

With a new way of doing additive manufacturing, more things can be brought to the U.S. on demand to be operationally ready. AI provides new sensors and geo-intelligence, and new ways of staffing skills lead to more niche opportunities.

“The superheroes of the future are actually going to be the people in the CIO shop that rethink architecture,” Fitzgibbon said. “Anyone in the IT department … and the people in acquisition, that rethink how we can acquire and move faster, and getting these things on contract into our government and getting them tested, and not failing, but failing forward.”

Cyber Resilience

Nagaraj closed the panel from an industry leader perspective highlighting how industry partners are working to adopt AI to help with data management, cyber resilience and cyber recovery.

“[Between 2000 and 2020] the amount of technologies that produced data was great, the amount of data increase was insane, but they all became silos of data that just kept getting exponentially proliferated,” he said. “What we’re trying to do is unify that data, we’re trying to consolidate it, we’re trying to make it addressable.”

He emphasized the need for common language and data management spaces to allow quick and accurate responses to prevent cyberattacks. He also said this type of system would allow for understanding of what potential future risks could look like.

“Imagine bringing security and IT into the same space. Imagine security is one part of the brain and IT is the other part of the brain. And inside of a cyber resilience conversation, we need that brain to work together,” he added.

Nagaraj said he believes cyberattacks are going to increase in the years ahead, and DOD needs to prepare now.

“We have to find ways to unify, we have to find ways to consolidate, we need to bring our organizations closer together,” Nagaraj added in agreement with the previous speakers. “We then need to unleash the power of generative AI, but in a responsible and governed manner, so that we can make decisions faster.”