June 21, 2022
Imagining military artificial intelligence (AI) applications can make one dream up scenarios like those in the Terminator films, but in reality, AI solutions for defense are much more mundane and focused on improving decision-making for humans, whether they’re aircraft maintenance personnel ; pilots; or intelligence, surveillance, and reconnaissance (ISR) analysts, says John Canipe, Director of Business Development, Air Force, at SparkCognition Government Systems, during a conversation we had at his company headquarters in Austin, Texas. We also discussed the difference between AI and machine learning (ML), how AI is being applied across multiple military domains, and more. Edited excerpts follow.
MCHALE: Please provide a brief description of your responsibility within SparkCognition Government Systems and your group’s role within the company.
CANIPE: As Director of Business Development, Air Force, my current responsibilities are product development, capture management, price/licensing of products, and generating new and recurring sales.
MCHALE: We often see AI/ML [artificial intelligence/machine learning] in the same sentence, or used to describe the same thing, but what is the actual difference between AI and ML?
CANIPE: Differentiating AI from ML is a struggle everyone is having right now. We see AI as a broad umbrella term, with ML as the heartbeat of AI, enabling actual applications of putting in data and getting an outcome, rather than data and tools. That’s why we refer to SparkCognition and SparkCognition Government Systems (SGS) as machine-learning companies.
When speaking of both, it helps if we remove the fictional AI in films like “Terminator” from the discussion, because that is just fantasy and not reality. Instead, we need to focus the AI on solving the critical, behind-the-scenes problems our military faces, like search features in the depot maintenance world. While that’s not a Hollywood headline, it is still vital for mission readiness. It greatly reduces downtime for aircraft. It enables that maintainer to look back in time and see what happened, how it happened, and pull that data quickly. Without this capability, such problem-solving could take a week or longer, costing time and money.
MCHALE: What are the design and requirement trends driving AI innovation in military applications?
CANIPE: There is a lot of opportunity to improve upon current legacy systems the military is fielding. These platforms and systems are not going away. By adding AI solutions to already existing platforms to drive cognitive capability, data filtering, and the like. it enables the life of the platform to be extended while enhancing capability. An example of a program tackling this concept is Project Kaiju, which is exploring AI solutions to embed cognitive electronic warfare (EW) at the edge.
There is a big push towards improving the readiness of the aircraft. How to solve that readiness challenge at the maintainer level with AI is the million-dollar question. Getting maintenance readiness above 50% to 60% will impact the entire defense community.
For deployed applications, AI requirements will focus on battlespace-management scenarios to speed up decision-making for warfighters at the edge of battle. This is done by AI algorithms that help filter out ISR [intelligence, surveillance, and reconnaissance] data close to the sensor, so the human operator monitoring the feed from an unmanned aerial system (UAS) sensor pod – for example – can more quickly decide what intel is actionable and get that actionable intel to the commanders in the field, speeding up the sensor to shooter process.
It comes down to speeding up human decision-making – whether the human in question is a maintainer at the depot level, a fighter pilot, or an ISR specialist analyzing data from a UAS sensor. There are so many decisions that need to be made that can overwhelm the cognitive load of a human mind.
MCHALE: How does AI enhance autonomy?
CANIPE: From an F-35 pilot perspective, the goal will be to have a wingman that is autonomous, but with instructions on what and how to attack a target. The autonomous wingman will alleviate cognitive weight for the F-35 pilot, improving decision-making speed. For the past few years, we’ve seen research and development around this technology. Funding has focused on prototyping to see if these capabilities are actually feasible. Project Kaiju has done a good deal of this type of R&D, prototyping, and testing.
MCHALE: Predictive maintenance, decision-making, and autonomous navigation are three of the most well-known AI capabilities. Are there others that the US Department of Defense (DoD) is investing in?
CANIPE: The DoD is focused on updating most of its software and evaluating its current processes. This will, in turn, allow the DoD a great opportunity to leverage new technologies. In the spirit of speed, however, we build our solutions to embed into the existing systems, which allows a seamless user experience and helps the DoD keep additional software costs down.
MCHALE: What are the acquisition pain points with AI? Are they technological? Bureaucratic?
CANIPE: AI is a software-as-a-service (SaaS) model, which is a new mindset for the government. The DoD is used to owning the technology they buy, such as a tank or aircraft and the electronics onboard the platform, like computer hardware or software operating systems. That is not a SaaS model. Some in government are still trying to understand the sustainment cost of an AI SaaS product, asking questions like: How do I acquire AI? Where is the IP? Who owns it? Can we create it on our own?
MCHALE: How has military technology changed since you served as an Air Force pilot and how is AI enabling that evolution?
CANIPE: Since 2018, there has been a cultural shift within the Air Force, from a culture of procurement and quick wins to one of innovation, so they don’t get left behind. It’s a more long-term strategy. It’s incredible to see the changes over the last few years. What’s crucial is that the push towards innovation is coming from the top, from Air Force leadership down to individual airmen.
MCHALE: SparkCognition founder, Amir Husain, has said that AI can be applied to every stage of the war and almost every activity. How is the DoD progressing in mastering these domains? Where should the focus be going forward?
CANIPE: AI can and is being applied across all areas of war, and it can be exciting thinking about possibilities and future applications. However, the key focus shouldn’t be the exciting, headline-capturing projects. Instead, we believe the DoD will realize exponential value by focusing on the smaller wins that may be less attractive to a wider audience, but deliver real return on investment to the DoD’s cost savings, speed of decision, and mission readiness. These smaller wins will pave the way to some of those more exciting projects down the line when the DoD has a more intimate understanding of the procurement and deployment cycles of AI solutions.
MCHALE: Looking forward, what disruptive technology or innovation will be a game-changer in the AI/ML? Predict the future.
CANIPE: The founder of my previous company in the oil and gas industry said that the future winners will be those who can take all the data into a single location and make sense of it. That was about 10 years ago. That statement holds true today. A major challenge facing the DoD at the moment is disparate data, spread across many different databases and stakeholders. The goal is to streamline a decision-maker’s access to the right data and ensure the correct protocols are in place to act upon that data. Once this access is unlocked, “unknown unknowns” will be easier to identify and act upon, unlocking innovation across the DoD.