Algorithmic Warfare: “Imagine” and “Implement”

| |

Though conceptual Artificial Intelligence (AI) has been around since the 1950’s, its methods were not adopted by the United States Department of Defense (DOD) until recently. Given the current buildup of AI products by Silicon Valley, California, it should come to no surprise that the DOD initiated the development of AI systems as one of the military’s top modernization efforts.

The efforts of China and other adversaries of the United States has created a competitive environment for bigger and better AI technology, comparable to that of the nuclear arms race. Artificial Intelligence/Machine Learning (AI-ML) is being incorporated into Geographic Information Systems (GIS) and their associated databases, ushering in a new future for how we analyze and solve complex geospatial problems. 

Machine Learning and Algorithms

            The current discussion of machine learning largely centers around the use of algorithms. According to Machine Learning for Dummies, IBM Limited Edition, “algorithms are a set of instructions for a computer on how to interact with, manipulate, and transform data.”

A typical algorithm would require a computer programmer to input the algorithm to enable the computer to execute a function. In machine learning algorithms, the algorithm is the product of data inputs. As the programmer inputs more data, the more accurate the algorithm becomes.

This concept is one of the foundational components of Project Maven, which will be examined further below.  

AI/ML Implementation and Project Maven

            Classifying data from remote sensing technologies dispersed across multiple combatant commands typically requires serious manpower to produce products that are relevant and beneficial to military operations. The primary concern is the vast amount of data compared to the number of analysts. The incoming data, received in the form of full motion video (FMV) from drones is also mind-numbing, and requires thousands of analysts.

            Project Maven was an initiative stood up by the Joint Artificial Intelligence Center (JAIC) in April of 2017. Project Maven is dubbed the “Algorithmic Warfare Cross-Functional Team” and is aimed at integrating AI into all branches of the military as well as additional governmental agencies such as the National Geospatial Intelligence Agency (NGA).

Its framework centers around 12 lines of effort, a majority of which directly involve the collection of aerial imagery utilizing remote sensing technology. One of these significant efforts is the utilization of machine learning in data processing systems. For example, a data processing system can be fed thousands of images of a Chengdu J-7 Chinese Fighter Jet, and it can “learn” to identify it from FMV data feeds in a fraction of the time it would take an analyst. 

Classification of these images can also influence decision making by Commanders, provide early detection of threats, detect changes, and possibly predict enemy strategies based off terrain and spatial features. According to Lt. General Michael S. Groen, current Director of the JAIC, AI is limited only by our abilities to “imagine and implement.”

Imagine having the ability to pre-emptively locate a buildup of Russian air defense artillery assets on the border of Poland, or the preparations for the launch of nuclear arms from Iran through an autonomous classification system. Lt. General Groen’s current efforts as Director of the JAIC additionally include incorporating AI/ML into what are called the “warfighting functions”, such as artillery fires and logistical operations.

Machine learning may also influence capabilities in determining line of sight, slope, and even parameters of superficial features on the battlefield. Work is currently being done on utilizing machine learning in conjunction with GIS data sets to render 3-D models of superficial features, such as bridges and stadiums, to provide further analysis for decision-makers on the ground.

Of course all these advancements require substantial amounts of data, but additional efforts are being made in the use of machine learning in serialization of data sets and building these platforms. These are only a few examples of what machine learning platforms can do. 

            The Project Maven vision capabilities have already been utilized to respond to natural disasters and the COVID-19 pandemic. According to an interview with Lt. General Shanahan (Ret.) in July of 2020, first responders could utilize FMV and remote sensing technology to quickly establish fire lines in the West Coast wildfires of 2019.

He additionally articulates that these same remote sensing platforms can be utilized for suicide intervention, prevention, and mitigation. Project Maven is continuously updating “almost on a biweekly method.” The project is additionally expected to expand over time, to further integrate other uses of AI and machine learning into the Armed Forces.

Combatting Stove-piping

The Department of Defense’s Artificial Intelligence Strategy (2018) outlines the use of Machine learning for DOD organizations. Organizations that are largely incorporating the use of Machine learning include the DOD’s Joint Artificial Intelligence Center (JAIC), the Army Future Command’s Artificial Intelligence Task Force (AITF), the National Geospatial Intelligence Agency (NGA), the U.S. Army Corp of Engineers (USACE), and numerous others.

            Lt. General Shanahan’s interview warned of the problem of “stove-piping” of data within the DOD. The issue of DOD subsidiaries keeping data for their own use and not sharing it in a common platform is historical in nature. With AI and machine learning, stove-piping is particularly problematic because it can lead to failures for organizations that cannot access relevant information. To solve this, the JAIC has been actively working on a data repository for military and government agencies. This shifts the JAIC’s priorities from a focus on developing AI to enabling the sharing of AI information across the DOD.

This concept of a centralized “data repository” has flourished into Lt. Gen. Groan’s “Joint Common Foundation” which aims at providing an integrated commercial cloud that enables end users to access algorithms that have been previously developed by the DOD or subcontracts. This allows for easy sharing of algorithms, specifically algorithms that can be applied to different data sets that may commonly occur. Developing this catalogue of algorithms and training data has huge implications for leaders on the ground.

In the past, DOD organizations were disadvantaged because they were unable to access relevant geospatial intelligence products. Now they will be able to utilize these products to make informed decisions on the ground quickly. According to Lt. Gen. Groen, the Joint Common Foundation is expected to be at “initial operating capability” by the end of February. 

References

For more on Machine Learning, see: Hurwitz, J., Kirsch, D., 2018. Machine Learning for Dummies, IBM Limited Edition. 

For the interview of Lt. Gen. Michael S. Groen, see episode of the podcast Eye on AI, link: https://www.eye-on.ai/podcast-archive

For the interview of Lt. Gen. Jack Shanahan (Ret.), see episode of the podcast Eye on AI, link: https://aneyeonai.libsyn.com/episode-45-jack-shanahan

For more on the U.S. Army Futures Command policies on the ethics and use of AI, see this link: https://armyfuturescommand.com/wp-content/uploads/2019/11/AITF-Scope-Areas.pdf

About the Author

Brandon Crossett is an Active Duty Captain in the United States Army, and has 5 years of experience as an Army Engineer Officer. He is currently pursuing a Masters Degree in Geological Engineering from the Missouri University of Science and Technology, and has an undergraduate degree in Criminology (B.S.) from George Mason University in Fairfax, Virginia.

Related


Enter your email to receive the weekly GIS Lounge newsletter: