Integrating AI into military operations discussed at Time Machine 2020
NewsSeptember 04, 2020
TIME MACHINE 2020 AI SESSIONS. Fighter pilot shortages, operation security concerns, and the need for rapid experimentation in the military have acted as driving forces behind the implementation of artificial intelligence (AI) in the Department of Defense (DoD). These themes were discussed by military industry AI leaders and data scientists to highlight how AI can bolster mission readiness at SparkCognition Government Systems’ (SGS) virtual Time Machine 2020 AI Sessions event, held this week.
A panel, hosted by Logan Jones, president and general manager of SGS, highlighted how AI- and machine learning (ML)-powered logistics and maintenance could support the military in achieving the demands presented by critical national security missions. In addition to Jones, the panelists included Secretary Lisa Disbrow, former under secretary of the U.S. Air Force & board member of SGS, Major Alex Goldberg, chief innovation officer of the Texas Military Department (TMD) & DARPA fellow for the Air Force chief of staff, and Chitra Sivanandam, vice president of analytics and simulation at SAIC.
Jones opened with the DoD’s definition of readiness, “The ability of military forces to fight and meet the demands of the assigned mission.” Meeting these “assigned mission” demands was once entirely powered by humans and the legacy systems that the warfighters themselves were required to control and operate. Today, some of the top engineers and data scientists in the defense space are grappling with not only the demand to innovate, but the speed at which the commercial industry is already doing so.
“We don’t have enough fighter pilots,” says Goldberg, who is responsible for accelerating the fielding of commercial technologies for both national defense and state emergency response efforts at TMD . “And the throughput, because we only have so many aircraft and so many pilots, we’re not able to produce as many as we need. So, from looking at AI-powered instruction to where 40% of the budget is for the Air Force, which is really about sustainment of our aircraft where many of the parts are becoming obsolete and breaking in ways that they never have before. So, can we look at predictive troubleshooting? Or can we predict parts that are going to break before they do? Those are among the problem sets we’re trying to solve within our community and help be the Sherpas to navigate through the bureaucracy which is relatively challenging to get through the DoD.”
The bureaucracy behind the DoD acquisition process, namely with AI capabilities, has presented a challenge for technology companies to overcome. Jones went on to note that the DoD allocates over $53 billion a year for operations and maintenance costs alone, totaling more than research & development, procurement, and military construction combined.
As former under secretary of the U.S. Air Force, Disbrow offered a honed perspective of implementing those new predictive technologies while meeting the security and affordability mission behind the defense supply chain.
“We require a faster response, and we require more force ready than we have in the last 15 years,” Disbrow says. “And when you look back at the 90s, our inventory of aircraft today is about 40% of what is was back in the Desert Storm days. So, it really becomes a global force management problem for the Air Force. There are supply and demand challenges. And to meet those force response demands while managing the cost of the equipment, sustainment, and personnel, we still have to maintain what we call operation security.”
While serving during the 2017 timeframe, Disbrow claims that the Air Force was working toward establishing the Office of Commercial & Economic Analysis to analyze the vulnerabilities across the enterprise and determine the best response. While this office is still in operation today, the data it brings in to assess and analyze is often commercial or market related. However, the military is still using such methodologies to secure the supply chain.
“The Air Force, particularly, is partnering with industry to plug some of these vulnerabilities across the whole supply chain. Things like IP and proprietary concerns,” Disbrow says. “There are some exciting things going on. DARPA is working with the logistics management institute to model the joint logistics supply chain enterprise, and that will help the government and the Air Force to better anticipate risk and to better quantify the value of AI in different scenarios.”
Initiatives to use AI-powered assets to offload time-intensive tasks from highly trained operators are in place to legitimize the investment the DoD makes when putting warfighters through costly training programs. Predictive and unscheduled maintenance as well as data analysis can be done more quickly and more cost-effectively through the use of AI, and industry initiatives are aimed at phasing in more AI to lighten warfighter workload and better use that untapped, expensive talent.
“Pilots are leaving in droves,” Goldberg says. “When you create a pilot, it takes 10 years to get them to a four-ship flight lead and the experience of where the person who left was. So, how do we get them trained to where quality is still the constant, but time is the variable? Where we’re using artificially intelligent instructors to close the gap, Pilot Training Next is an example. The program used simulators that were $5,000 a piece versus $5 million.”
At the heart of AI, whether it be simulation and training or predictive maintenance, is the data and how it is implemented. Companies like SAIC are on the forefront of this AI-driven innovation throughout the defense industry, and Sivanandam acknowledges how pivotal accurate implementation of data can be to an AI-powered system.
“Data is king,” Sivanandam says. “And it isn’t always available to solve some of the more complex problems for our customers and partners. There is a paradigm. There are a lot of open data initiatives across our federal agencies and we’re starting to see a lot more value coming out of those open data environments. And there’s a push to get to more data transparency, more accessibility, and more open sourcing because open platforms for data make is tremendously more valuable.”
Both Sivanandam and Disbrow agreed that the more complex the systems get; the more difficult accessing data becomes in terms of the vulnerabilities and cybersecurity concerns that they create. Training operators to understand the nuances of multiple capabilities from sensors to logistics also presents its own set of obstacles.
“It is an optimization problem,” Sivanandam says. “I might be trying to narrow down time or reduce cost and complexity as I move something through a system. But how do I build and prepare the AI and the agent and the capability for the unknown now and the unknown for tomorrow’s problem set? And that’s what we’re trying to figure out. How does the data enhance the thing I build for tomorrow to retain the agility and the nature of rapid experimentation?”
Accessing data in defense environments is often far more difficult than doing so in the commercial industry. Sivanandam goes on to explain that this has hindered a culture of rapid experimentation in the military, further illustrating the limitations that the DoD faces when trying to leverage commercial AI advancements in military technology.
In spite of the inevitable hurdles groundbreaking innovation like AI will bring, the DoD is seeing significant advancements due largely in part to industry leaders like those who spoke at Time Machine 2020 and the companies they represent. The drive to pioneer AI-powered systems for the military remains, and the “higher level of thought” that SGS aimed to portray with this year’s virtual event was conveyed through the ingenuity that the panelists assured would be carried on by generations to come.
