Leveraging the Compute Artificial Intelligence Needs

The confluence of computer technology, artificial intelligence (AI), data management and concerns of growing risks from advisories has U.S. Air Force officials examining more ways to support airmen with the capabilities they need.

A new administration is putting in its priorities, but in the realm of information technology (IT) and AI, some of those goals are not different from what Alexis Bonnell, the chief information officer (CIO) and director of the Digital Capabilities Directorate of the Air Force Research Laboratory (AFRL), and Venice Goodwine, Senior Executive Service member and CIO for the Department of the Air Force (DAF), were working toward previously.

What is different are the efforts of the Department of Government Efficiency, or DOGE. For that, the DAF CIO is responding to the efficiency and effectiveness measures, strengthening portfolio management and performing an investment review, among other steps.

“[With the new administration], a couple of things we know are efficiency and effectiveness are the things,” Goodwine noted. “So, I'm tracking there. We have already started doing investment reviews for our enterprise IT portfolio. But as your CIO, I am not just enterprise IT. I'm actually IT for the enterprise. And so they're challenging me to understand all the dollars spent, not just within my five portfolios, but what the department is spending on IT and how is it moving us ahead to be prepared for national security.”

In addition, the administration is signaling an urgency—not unlike the message from previous Air Force leadership—given the threat from near-peer adversaries.

“The first thing I did was to align with what is the administration thinking about doing,” the DAF CIO said. “The other thing is, they want to make sure that we foster this environment of being agile, having a sense of urgency. You've heard that term before because the previous administration talked about a sense of urgency.”

In addition, the DAF and AFRL will continue to pursue AI at scale for the department.

“The foundations of what the administration wants of AI is still the same,” Goodwine explained. “How do we treat AI like you do electricity, that you don't think twice about it, like the air you breathe, you just have it? For AI, that is really embedded.”

Last year, Bonnell, Goodwine and other DAF and AFRL officials ushered in an era of AI experimentation into the Air Force, with tools and practices spreading quickly to the U.S. Combatant Commands, such as U.S. Central Command.

To build on that momentum, the DAF CIO is setting up an AI Center of Excellence. “I'm doing that in partnership with industry and in partnership with academia,” she said.

For example, the DAF CIO recently visited Purdue University to speak with university officials about what the school was doing with AI and the possibility of military and civilian AI engineering and training.

“It's not just about Gen AI [generative AI],” Goodwine continued. “It's not just about the back office, but it's really focused on, ‘How can I use AI to make sure that decision-makers have data readily available, that they can understand, and more importantly, that they trust?'”

In addition, the officials said they wish to leverage private sector investments in AI, especially given the president’s $500 billion AI initiative.

“I was really excited about three weeks ago, President Trump announced a private sector investment of up to $500 billion fund infrastructure for AI,” Bonnell said. “That is ‘billion.’ And it obviously is signaling a continued emphasis on AI and really doubling down on AI innovation.”

“We are hoping that we can capitalize on some of that $500 billion to help us within the Department of the Air Force. But I'm also not doing this in a silo. I work with my other CIO partners as well, with Jane [Rathbun] over at the Navy and Leo [Garciga] at the Army. We conspire and collude together about the things that we want to do with AI,” Goodwine laughed.

To go forward with all of that AI innovation, though, the DAF needs powerful computers, and Bonnell, Goodwine and others are looking at solutions.

“What I think we learned together was that making progress on digital transformation really meant that we had to have the jet fuel that some of these emerging technologies require,” Bonnell noted. “And we have a lot of lessons learned on compute being so critical. We were able to collectively pioneer hybrid compute options for some of the first large language model tools we did.”

In fact, Bonnell added, the high-performance compute that the AFRL had in mind for the early AI tools was not the exact fit. “I've made a lot of really bad assumptions and then gotten to learn. As an example, we're lucky enough to have a high-performance compute center in our team. But what we found is that many of these machines have been factored for batch loads or processing. They actually were not factored or containerized for the kind of the inference query of AI. I joked a little bit that it was a little bit like passing M&M candies back and forth and back and forth."

“I think it was also eye-opening to recognize from a compute lens that we actually have to be really intentional about moving GPUs [graphic processing units] into cloud,” Bonnell stated. “I'm excited about how this might scale, but I want to own the fact that we are learning as we go.”

Lastly, the leaders promised as they go forward to always keep the importance of their mission in mind.

“It is very important to me that are we going to produce [solutions] for the airmen and the guardians at the edge and even as they prepare to go to the edge,” Goodwine shared. “And so, I'm in places and in tune with what the administration is saying and doing so that I can make sure that I'm advocating for each of you.”

The Rocky Mountain Cyber Symposium is organized by AFCEA International's Rocky Mountain Chapter. SIGNAL Media is the official media of AFCEA International.