A few months ago, an AI pilot trounced an elite U.S. Air Force fighter pilot in a simulated dogfight. Sound like the stuff of sci-fi nightmares? If the Air Force continues on its current path, it is.
When the Defense Advanced Research Projects Agency (DARPA) concluded the AlphaDogfight competition in August, the Air Force experienced just how advanced AI systems have become.
AlphaDogfight pitted AI companies against one another’s dogfighting algorithms in a tournament-style competition, with the winner earning the chance to face off against a human fighter pilot.
The final tally: 5-0 in favor of the algorithmic “pilot.”
That result reinforces a trend in AI vs. human contests: Humans are losing their edge over machines in more and more tasks. And the trend is accelerating. It’s a question of when, not if, AI will change everything about the way the Air Force must do business.
To make that transition, the Air Force must accelerate its efforts to ready itself for the AI revolution. Otherwise, as Chief of Staff Charles Q. Brown put it when calling for accelerating change: The service will lose.
Senior leaders such as Brown are saying the right things. The 2018 National Defense Strategypushes for “an unmatched 21st century National Security Innovation Base.” The 2019 USAF Annex to the DoD AI Strategy and DoD Data Strategyrecognize that AI will shape the battlefield of the 21st century, and the Air Force must invest in a data-centric ecosystem to facilitate AI capability. The recent National Defense Authorization Act even includesmeasures to direct focus on AI development in critical areas.
This direction has produced initiatives like the Air Force/MIT AI Accelerator, unit-level Spark Cells, the advanced battle management system (ABMS) initiative, and software development organizations such as Kessel Run and Kobayashi Maru.
These innovation hubs have incrementally proven out a series of best practices for changing the culture and practices of the service to prepare for the AI age.
Unfortunately, these exceptions prove the rule: Despite evidence that AI is going to radically change the way the Air Force must operate, and quickly, the service is still only sprinkling innovation, and AI efforts more specifically, across the service.
Bold initiatives such as ABMS and JADC2proclaim an envisioned service that leverages advanced analytics and decision intelligence algorithms to build situational awareness and the ability to act on it quickly, but how does the Air Force take the next step on its journey toward true AI-enabled capability (and prepare for the myriad business process changes that AI will bring), at a pace that will ensure it isn’t left behind? The answer is multifaceted and complex, but the first step should be to scale validated practices from Air Force innovation units and adopt a few overdue changes. These steps should focus on the areas any organization needs to make change to ready itself for AI: people, data and culture.
1. People: Reform talent management for digital skill sets.
In 2020, Kessel Run demonstrated effective ‘digital’ (referring to digital age skill sets critical to AI development like software development and data engineering) talent management by allowing an almost completely remote workforce, movingto virtual hiring, and finding new ways to foster the skills of its people. These moves allowed the organization to pull from a much larger talent pool.
To scale recruiting success demonstrated in the innovation ecosystem, the Air Force needs to allow remote work wherever feasible and maximize its participation in virtual recruiting events, hackathons, career fairs and tech conferences to build more bridges to these communities.
To scale success in building organic Air Force talent already in the service, the assignment selection process should (finally) improve how it tracks, incentivizes and assigns digital talent to appropriate organizations and positions. Tagging skillsets and providing financial compensation will improve retention. The service must also expand continuous learning opportunities outside the professional military education and technical schools.
2. Data: Set the stage for artificial intelligence by becoming data-centric
The service must accelerate its efforts to become more data-centric. The work and advocacy done by Kessel Run, Kobayashi Maru and the Air Force Chief Data Office’s VAULT program embody data-centric efforts. They emphasize data sharing and focus on enforcing healthy data management standards and practices.
Kessel Run’s ODIN effort, the suite of applications replacing the F-35 ALIS maintenance system, demonstrates the importance of valuing data. The ALIS system was plagued by insecurities, prone to error, and left much to be desired in the “user-friendly” category. Within a year of standing up a data team, which was empowered to make appropriate decisions about how to gather and utilize data, Kessel Run deployed a replacement suite of applications that strengthen the F-35 data ecosystem by enforcing standards, creating data stores for analytics, and optimizing maintenance schedules. The new system brings vast data improvements, leading to less maintenance time and more efficient scheduling.
The service faces big challenges to scaling data-driven systems across the Air Force, such as the availability and sharing of data across silos, and archaic methods used to manage huge data sets. Leaders must require and enforce informed data sharing in every digital system, support good IT infrastructure to enable data movement, hire and support software development capabilities to collect and analyze the data, and demand security to protect it. Integrated and operational AI is a system of systems, requiring much more than just algorithms — the right people, data, infrastructure and tools are needed to establish and maintain it.
3. Culture: Embrace and espouse an agile digital mindset
The Air Force innovation ecosystem has demonstrated the value in challenging the status quo, failing fast and iterating continuously. These components, which collectively comprise an agile mindset, have driven these units’ ability to deliver capability to the war fighter. The Next-Generation Air Defense Program recently reported to the world that 6th-gen fighters have been developed, simulated and flown on an incredibly condensed schedule, thanks to sophisticated digital engineering and modeling like that seen in modern Formula One racing design and engineering. These methods simulate real-world conditions with extraordinary fidelity and give designers and engineers the ability to run a huge number of experiments and simulations at an unprecedented pace. A mindset that combines the iterative power of the agile process and the transformational power of digital-first design and evaluation is foundational to integrating AI algorithms.
To scale this across the Air Force, leaders need to publicly and consistently embrace this agile mindset. This means consistently communicating Brown’s “Accelerate Change or Lose” effort and acquisitions chief Will Roper’s call for disruptive agility and a new digital paradigm for Air Force spending. It means taking a cue from Space Force Chief John Raymond’s public push for guardians to help him stand up a lean and agile service. It also means championing success stories like the innovation efforts by the 99th Reconnaissance Squadron at Beale, which implemented radically new processes and demonstrated an ability to provide solutions at speed in the recent U-2 mid-flight software update and AI co-pilot demonstrations.
To recognize and reward these efforts, commanders must discover, then insist on metrics that demonstrate solving problems iteratively with digital solutions, and track speed and efficacy of capability delivery.
The service cannot lose this competition. In an era of renewed great power competition, when peer competitors China and Russia are clearly prioritizing AI development, and the capabilities that come with it, that could be a fatal mistake.
Capt. Jazmin Furtado is an Air Force acquisitions officer, and data science and artificial intelligence leader. She is the current portfolio lead for Wing C2 data at Kessel Run, and liaison to the Air Force and MIT partnership AI Accelerator.