The case for military adoption of artificial intelligence

^{Illustration: Getty Images}

 

Rio de Janeiro, situated between the mountains and the sea on Brazil’s Atlantic coastline, is a sprawling metropolis that covers roughly 500 square miles of land. Its surface, depicted in satellite imagery, also represents an enormous parcel of raw data. As part of a 2017 internal research and development (IRAD) program, a team of Leidos data scientists developed computer vision algorithms, powered by machine learning, to identify man-made structures in overhead imagery throughout the city. As a manual process, the team estimated it would take an individual analyst roughly 44 weeks to identify and categorize human structures across all of Rio. The automated solution was so advanced, it learned to identify these types of structures on its own, reducing analysis time to a matter of hours.

The success of this program, more than demonstrating how machine learning holds the potential to transform operational processes in visual analysis, foreshadows how smart machines will disrupt defense and intelligence operations on a global scale. The U.S. military, like any other large organization, is flooded with raw data that needs to be converted into actionable intelligence, a challenge which artificial intelligence (AI) and its subset of computing systems (machine learning, deep learning, neural networks, etc.) are uniquely designed to address. While a major segment of the defense and intelligence community is devoted to the exploitation of visual data captured by satellites and drones, a tradecraft known as geospatial intelligence (GEOINT), this kind of analysis is one of virtually limitless possible military applications of AI.

In futuristic scenarios, this same technology will play critical roles in counterterrorism. In the aftermath of terrorist attacks, huge amounts of surveillance footage and social media data are generally available to help identify suspects and create situational awareness. With automated, real-time video and natural language processing AI, analysts can turn these data into actionable intelligence much faster and potentially save lives. In a battlefield context, these systems can help interpret mission critical data including the detection of armaments and weaponry, and, more at issue, be used to devise and defend against military attacks.

It would be hard to argue with the belief that AI is poised to reconstruct critical operations inside the U.S. Department of Defense (DoD). AI has arguably become the preeminent emerging technology the Defense Department is eager to understand and apply. This month, the Pentagon established the Joint Artificial Intelligence Center (JAIC) to aggressively pursue AI in order to assure military superiority and deter threats to U.S. interests. An April 2017 memo from the U.S. Deputy Secretary of Defense called for the integration of these systems to “maintain advantages over increasingly capable adversaries and competitors,” and stated the department should do “much more, and move much faster…to take advantage of recent and future advances in the areas of AI, big data, and deep learning.” The Pentagon spent roughly $7.4 billion on AI and related fields in its unclassified budget that year alone.

But military applications of AI are a conduit for highly controversial ethical considerations. Some experts within the technology sector are grappling with the ethics of deploying AI as an instrument of warfare and ushering our most sophisticated computing technology into the domain of military conflict. Critics express wariness about wielding tools designed to process the world’s most sensitive data, the consequences of machine intelligence surpassing that of the human brain, and trusting life-or-death decisions to computers.

Ciro Pinto-Coelho, who leads research for Leidos in cyber and data science, concedes AI will fundamentally change the calculus of strategic decision-making, unequivocally the most important factor of winning in war. But he believes the true power of this technology is centered on collaboration between humans and machines. Machine learning and deep learning systems, as well as other subsets of AI, are simply tools within a larger computational toolset, building blocks of systems that can solve problems, aid humans, save lives, and minimize tedious work. U.S. military doctrine assures human accountability behind any decision to pull a trigger, he explained, whether that trigger is attached to an automatic rifle or a nuclear missile. But specific regulatory safeguards should be put in place, he believes, to assure the U.S. military may neither relegate life-or-death decisions to machines, nor abdicate responsibility for such decisions.
 

Keith Johnson, Chief Technology Officer for the Leidos Defense business, agrees the power of AI is in its ability to sift through large amounts of data to enhance decision making, an advantage that is just as important on the battlefield as it is in the marketplace. "This technology can and should be applied," Johnson said, "to enable the U.S. government to make the best decisions in diplomacy, economic policy, public welfare, and on the battlefield." Even as the regulatory environment around AI matures, he believes the U.S. government should ensure AI plays only a supporting role when lives are on the line. Regardless, he said, given the ability of smart machines to help humans make faster and better decisions, and to automate laborious processes and workflows, AI will “continue to be an area of strong interest to the DoD given the constant pressure on the department to do more with less budget.”

As these debates unfold, important advances will continue to be made in the real-world applications of AI. Pinto-Coelho and Johnson are optimistic about its wider applications as well, including disaster relief, urban planning, autonomous vehicles, and self-guided robotics. Leidos plans to continue to develop its expertise through research programs and centers of excellence to accelerate customer adoption. “The Leidos philosophy is to take something that’s on the leading edge of science, and apply it to customer challenges,” Pinto-Coelho said. “We’re gaining a strong understanding of how these systems can help our customers well into the future, and we’re identifying areas of further investigation and research. The AI revolution in national security is coming. Leidos will certainly be an important part of that, but it has to be enabled by a strong foundation of science and technology.”