Baseball Shows Naval Aviation a Path to Artificial Intelligence

The Navy knows it needs big data, artificial intelligence, and machine learning, but it still is grappling with what it wants AI to do. This must change if the Navy is going to reap the benefits of these emerging technologies.

— Captain George Galdorisi, U.S. Navy (Retired)

Artificial intelligence (AI), big data, and machine learning are popular buzzwords across the military and the defense industry. But as Captain Galdorisi argues in the epigram, the Navy must cut through trendy phrases and broad strategic guidance to figure out exactly how the force can leverage these technologies.

AI could be quickly employed—within the next year—to enhance the Navy’s effort to maintain maritime superiority: sifting through the virtual haystacks of data built up by the Naval Aviation Maintenance and Safety Programs to make naval aviation safer and more lethal by predicting sortie-specific risks.

Such a system could be customized using historical methods proven by a naval aviation great. The United States' oldest major sport suggests how.

‘Hit ’Em Where They Ain’t’ Gets Harder

To understand how behind the Navy is at AI, it is instructive to consider how advanced Major League Baseball is. Baseball is one of America’s oldest and tradition-bound/resistant-to-change sports—but it has embraced innovation and technology to develop advanced ways of assessing players and choosing strategy.

In one of the most visible changes, teams have begun widely using the “defensive shift” against many hitters. The idea is to adjust where defensive players stand to move them from a conventional, generic alignment to wherever the hitter is most likely to put the ball. Against left-handed hitters (such as Cincinnati Reds all-star Joey Votto, or former Red Sox slugger David Ortiz), the shift usually involves moving most of the infielders to the right side of the field or putting one infielder (often the second baseman) in shallow right field.

Statistics suggest that as hitters faced such shifts initially, they saw decreases in batting average (and other, more sophisticated offensive stats) but more recently have begun to adjust their approaches at the plate, reducing the shift’s effectiveness. Defensive-minded managers used spray charts to measure hitters’ tendencies; hitters respond with new approaches and their own statistical analysis of recent events such as homerun trajectories or player movements, using Statcast by Amazon Web Services.

Because baseball is big business, statistics have gone from scouts’ notebooks to computer databases. Other sophisticated performance measures—such as wins-above-replacement, jump step, and weighted on-base average—all form a more complete picture of a player’s skills than what an old-time scout could deliver while watching a game in the hot sun.

Baseball has gone high tech, setting an example for the Navy to follow. 

If Major League Baseball Can Change, the Navy Can, Too

According to Galdorisi, Navy leaders have requested $60 million dollars in 2019 for AI and machine learning.¹ They also have placed their organizational weight behind efforts to integrate these technologies into the fleet. But for naval aviation to break its own tradition-bound ways to fundamentally change the way it operates, it has deeply entrenched attitudes and bureaucracy to overcome.

In “All This ‘Innovation’ Won’t Save the Pentagon,” Army veteran Zachary Tyson Brown sums it up this way:

The authors of the National Defense Strategy recognized that “success no longer goes to the country that develops a new fighting technology first, but rather to the one that better integrates it and adapts its way of fighting.” . . . Of course, that’s easier said than done. The . . . structural reforms [needed] are a near-insurmountable challenge in a rigidly hierarchical institution with hundreds of empowered stakeholders and two million employees. So the department instead remains focused on developing new tools rather than thinking of ways to integrate them.

To paraphrase Admiral Kyle Cozad, you can have the highest technology in the world, but the most important asset for naval aviation is still the people.² And those people all too often operate with a 20th century mindset, relying on gut feelings and subjective decision-making born out of long-established flight training programs and early 2000s operational risk management methodology. 

AI Leverages Data That Already Exists

The proposal does not require a revolution in military affairs, nor does it demand a self-aware “strong AI” copilot (as in the 2005 film Stealth). Instead, as Galdorsi notes, “The area where AI can help warfighters is best described as artificial narrow intelligence (ANI), that is, AI that helps perform specific, discrete tasks.” Naval aviation’s ANI system would compile existing data from across several different databases to predict risks for upcoming sorties. (See Figure 1.)

Deliberate use of time-critical operational risk management (ORM) methodologies to fuse this data is ingrained in aviators from their first flight-training event. According to the Navy’s ORM instruction, squadron commanders are responsible for ensuring operators use ORM to “accept risk when benefits outweigh the cost, . . .  accept no unnecessary risk, . . . anticipate and manage risk by planning, . . . [and] make risk decisions at the right level.” Aviators deployed around the globe use ORM daily. 

If ORM is so successful, a cynical aviator might ask, why do we need an AI system? Making time-critical risk decisions is what we are paid to be good at—and why commercial airlines seek former Navy pilots for their flight decks. But however good we are, humans are limited by time and brain power. A tailored ANI system can make our force even better. This is of the utmost importance as peer adversaries become more capable.

The Problem with the HazRep Clipboard

An example of where ANI could create significant improvement can be found in the Naval Aviation Safety Hazard Report System, commonly called HazReps. These ubiquitous text-only files comprise a consequence-free reporting system from which aviators can take “lessons observed” and turn them into “lessons learned,” preventing future mishaps, saving lives and aircraft, and making squadrons more mission effective.

The current process is generally effectual if slow, but the amount of HazReps to read always seem to exceed the time available to read them. Many squadrons use “electronic read boards,” but others still print paper copies of the HazReps and post them in the safety department or the wardroom, where each person is required to hand initial that they have read the latest reports for their aircraft type. This system is equivalent to using a fax machine or a dial-up internet modem in a 5G world. It is simply not a path to maintaining maritime superiority.

Programmers could use the searchable text to train an ANI system to supplement human brains—“that is, AI that helps perform specific, discrete tasks.”³

The HazRep database is only one of at least 16 different information sources that an ANI system could continually scour to develop tailored planning and decision-making information for fleet operators. And just like the HazRep system, the other databases have their own idiosyncrasies that should relegate them to 1995, not 2019. With the right requirements, contract, developer, and funding, a prototype pre-flight ANI ORM planning assistant could be live within the next 12 months. How could it work?

Jay Beasley, Honorary Naval Aviator #11

Before delving into a high-tech AI system for tomorrow, a look to naval aviation’s storied and colorful past is warranted. How have Naval Aviators been successfully trained by a supplement to the normal pipeline? History gives us the answer: Mr. Jay R. Beasley. Born in 1914, Beasley was a civilian Lockheed test pilot who flew 31,497 landings in the P-3 Orion, the Navy’s maritime patrol workhorse for almost six decades. According to a 1981 issue of Naval Aviation News, Beasley traveled the world to train Navy pilots in the finer points of flying the four-engine turboprop. Most pilots he flew with were already winged, but Beasley helped as a “fine-tuner” to perfect technique and impart the most important whys of flying “beyond just what the books said.” 

He trained so many Navy pilots that he was named “Honorary Naval Aviator #11,” and his callsign was “Mr. P-3.”  Mr. P-3 was also famous for his gusto at the bar after a long day in the landing pattern; legend has it that he died of a heart attack at the NAS Jacksonville Officer’s Club in 1998 while in town to attend an annual conference about updating the Orion’s flight manual.

Beasley was a jolly, portly, silver-haired test pilot who flew Navy planes well into his 60s, fine-tuning crews and making sure hard lessons from the fleet were learned to prevent future mishaps. It would be fitting if this huge 20th century personality should live on in the 21st as an AI tool by lending his name to the proposed Beasley Artificial Narrow Intelligence Fleet Fusion Aviation Briefing System—or just “The Beasley.”

The 21st Century Beasley

The Beasley ANI should sift, sort, and fuse the overwhelming volume of data that exists across multiple systems to “fine tune” aircrew decision-making ahead of flights. The system would help eliminate subjective ORM risk decisions or inadvertent fixation on lower risk hazards at the expense of awareness of higher risk ones. 

Even with the Beasley, naval aviators would do most everything they do today, except they would receive more timely and relevant information than they currently get. For example:

Maintenance. The worldwide response time for failure rates and maintenance trends goes at the speed of message traffic and .txt files—slow. With the Beasely, a helicopter commander in the Mediterranean could know in real time about recurring generator failure trends identified by the ANI from the fleet replacement squadron in San Diego and forward-deployed forces in Asia.

Safety Reports. The Navy uses the Aviation Safety Awareness Program (ASAP) to compile anonymous post-flight data from aircrew, but the reports come out monthly. Beasley ANI could quickly recognize a problem with air traffic controllers at a certain airfield based on isolated reports from a helicopter sortie, a maritime patrol sortie, and a tactical jet sortie from different squadrons. The ANI could connect the dots sooner to make stakeholders aware of an issue before it becomes a Class-A problem.

Aircrew. The Beasley could access aircrew qualification and experience data to analyze current trends against historical ones. During a high-operational-tempo period for a maritime patrol squadron, a relatively inexperienced aircrew flying in particular conditions could see what their statistical likelihood is for involvement in certain kinds of mishaps, leading them toward better decision-making during the mission. 

The potential of such technology is high—and, doubtless, even better ideas could be built into Beasley 2.0. It could become the foundation for a modern fleet that does not require the grab bag of disparate systems we fight with today.

Adapting and Changing is Imperative

Despite its shiny new F-35s, P-8s, and MH-60s, many within naval aviation operate like it is 1995 and Windows 95 is king. Baseball, big business, and peer competitors are investing heavily in developing and acquiring AI technology. Unfortunately, U.S. adversaries are not just looking to induce Joey Votto to hit the ball at a glove—they want to replace western democracy and the liberal world order. 

The nature of making pilots better has not changed over the years, but the methods should. The Beasley ANI could be implemented in short order with the right support from Navy leaders to keep us ahead of our adversaries. Jay Beasley himself would likely agree—maybe after a beer or two at the O-Club, of course!