This winter, as I watched my foreign language instructor spend 5-10 hours every week filling out routine administrative paperwork, I realized that there were AI/ML technologies available that could not only have automated most of that work but also made it more accurate. This would have allowed her to spend those hours working on class preparation or working on the textbook she dreams of writing. But, because I didn’t know how AI/ML works, I could not advise her on the basics of how to free her time. Who has not stayed late at work, filling out paperwork that should have been automated? Wouldn’t it be nice to not waste time filling out paperwork, but instead use the insights we are supposed to be getting from that data to build better units? This frustration motivated me to begin looking more closely at how AI/ML works, what it requires to be implemented, and how to prepare myself to lead in an AI/ML environment. As I investigated this field, I realized that my technical skills are lacking. I needed to go back to school to be an effective AI/ML leader so I began exploring where I could learn these skills on my own. And if I, one of the slower people in the room, am looking for ways to learn these skills, then other FGOs must be looking as well.

Military applications of Artificial Intelligence and Machine Learning (AI/ML) have become powerful new tools in modern war and great power competition. AI/ML tools change how organizations use data to make better decisions faster. The ability to ask better questions, find novel patterns in data, and automate tasks mean that AI/ML is a compelling tool for improving military capability. The use of AI/ML in intelligence applications is well established. AI/ML applications also are now available for backend maintenance, facilities management, equipment maintenance, administrative tasks, and more. These AI/ML solutions can have real impacts on reducing administrative burdens and improving the lives of our people while increasing warfighting capabilities. The enormous potential of AI applications is not “around the corner” but already here.  As AI/ML moves into every corner of the DoD, all FGOs will find themselves leading AI/ML exploitation and sustainment teams.

Field grade officers need to learn how Artificial Intelligence and Machine Learning (AI/ML) work in order to use, advise, and lead AI/ML development and procurement efforts. The field grade officer has a unique place in the technology development timeline that makes this education especially important. Junior officers should increasingly be joining the services with at least a rudimentary understanding of these technologies, but they will need FGO mentorship on how and why to apply AI/ML to solve military problems. Technical AI/ML experts are increasingly concerned that military leaders are not able to appropriately champion and lead AI/ML projects because they don’t understand the technology.

The following provides a guide to where you can learn AI/ML, beginning with understanding what AI/ML skills an FGO needs. The Department of Defense Artificial Intelligence Strategy states that training will be developed for senior, mid-level, and technical levels of the DoD. Mid-level leaders require knowledge for “directing AI projects, resource allocation, and progress tracking, and for developing technical backgrounds for successful AI project deliveries”[1] (emphasis added). The National Artificial Intelligence Research and Development Strategic Plan: 2019 Update similarly notes that,

Finally, the AI R&D workforce will consist of multidisciplinary teams comprising not just computer and information scientists and engineers, but also experts from other fields key to AI and ML innovation and its application, including cognitive science and psychology, economics and game theory, engineering and control theory, ethics, linguistics, mathematics, philosophy, and the many domains in which AI may be applied.

Research indicates that the multidisciplinary teams described above require members with a basic understanding and appreciation of each other’s skills. This understanding allows members to communicate with appropriate and precise terms, support each other as the burden of work shifts from one specialty to another, and appreciate individual team members’ contributions and limitations. In short, FGOs need the technical understanding to meaningfully participate in these cross functional teams that will become increasingly common in the DOD.

While the management portions of the NIST and DoD AI strategy’s recommendations are simple to understand, the technical background requires more familiarity with how AI/ML works. The “Drew Conway Data Science Venn Diagram” provides a useful framework for understanding and organizing the separate skills that, when used together, create AI/ML understanding. It also has informed my suggestions for finding courses, all of which are online, self-paced, and free. The goal of this guide is not to make FGOs deep technical experts in AI/ML, but to provide enough technical knowledge that FGOs can ask informed questions and tune their BSometers.

Right off the bat I’m going to recommend that you take Coursera’s AI for Everyone class. This free, four-hour course introduces AI/ML concepts and implementation for a non-technical audience and provides enough information to understand the basic landscape of the field. Maybe the most important section is the discussion on the behaviors of AI organizations versus legacy organizations. The three- minute discussion of AI transformation in legacy organizations alone makes this course worth your time. It explains that AI/ML is about more than buying new software because it requires a new understanding of data’s value and how to use data to achieve mission goals. This course gave me solid insight into what AI/ML does as well as the leadership challenges that its implementation will require.

Returning to the Venn Diagram, the first circle of “Substantive Expertise” should already be an area that we are actively engaged in. The only suggestion that I will make is to understand and use modern management methodologies like AGILE, SCRUM, Mission Model Canvas, and Interagency teaming. Larger than AI/ML, these skills apply to any software, hardware, and tactical/operational execution and management work today. These are methods for asking better questions about what we are trying to accomplish, prioritizing effort, and communicating processes. They are also the management methods that you will likely encounter if you work with outside AI/ML development teams. They are the “secret sauce” to leading modern teams at the speed and effectiveness that modern combat requires. If you learn nothing else from this article, learn those skills.

Next, the “Hacking” circle on the diagram is filled by learning Python. This is the coding language used by most AI/ML applications. Unfortunately, AI/ML introductory courses require basic knowledge of Python, and this step cannot be skipped. Luckily, knowing some basic Python pays off in non AI/ML applications that are generally useful for FGOs. Kaggle presents an integrated AI/ML syllabus, but the coding was too advanced for me (I’m pretty sure I passed CompSci 101 by camping out in the professor’s office and he gave me a C just to keep me out of his hair for a 2^ndsemester.) Luckily their intro to Python has a link to Python’s own Python for non-programmers page. This page has many resources for various types and levels of learners. I’m personally working through Automate the Boring Stuff with Python, which offers more to the FGO than just Python skills. With 13 chapters that each take about an hour to complete, this is hard in the short term, but easy in the long term.

While computer code is the language that we use to talk to AI/ML machines, math contains the frameworks that this language is trying to communicate. The “Math & Statistics Knowledge” circle is filled by Linear Algebra and Statistics, the foundational math of this field. Having a basic understanding of AI/ML math allows us to ask if a problem is appropriate for an AI/ML solution and which type of AI/ML math we should use to solve that problem. Understanding how the math works, and where its biases and limitations are, also helps avoid the trap of using coding and subject matter expertise to build biased solutions into our AI/ML stack. MIT’s Open Courseware Linear Algebra was the first course recommended to me for understanding AI/ML. An understanding of statistics is also vital, and MIT offers Introduction to Probability & Statistics. The mathematics and computer science faculties of MIT’s Open Courseware also offer advanced computer science and AI/ML focused courses for anyone who wants to dig deeper.

The next step takes you to the Venn Diagram’s center. The two best sources that I’ve found for learning how to build AI/ML are the Kaggle and Google Education lessons. I tried to begin with these courses, hoping that they offered a shortcut to general knowledge, but I found that without Python and the math, I couldn’t go very far into these.  Google’s Machine Learning Crash Course, for example, requires Python and linear algebra knowledge. Even the Finnish government’s recently-released Elements of AI course, that it developed to educate the Finish people on AI/ML, requires some knowledge of Python. There are no shortcuts to the center of the Venn Diagram, but these courses are open enough that you can go as deep into the topic as you would like. I recommend experimenting with the intro lessons of each to determine which suits your learning style and knowledge-level best and then completing that course.

Local universities and schools may also offer similar courses in each of these areas, which will provide built-in networking opportunities. Unstated in the DoD strategy or Dr. Conway’s work is the need to build a network of actual experts in this field. I could not have written this article without the advice of actual data and AI/ML scientists. This field will also evolve and change quickly, and we will all need friends who can keep us in the loop; as in real life, our chances to “phone a friend” are only limited by the quantity and quality of friends that we have.

This list is certainly not exhaustive, but it is a starting place. Remember that this technology is here, now. We must be experimenting with and adopting AI/ML today if we want to build and lead a military that can fight in the AI/ML environment.  There are many resources available, but it is up to us to take advantage of them. And, if you do not end up using any of these skills for AI/ML implementation, there is still value to be had in each of the individual skills. We all need to understand statistics better, even the “fuzzy” social sciences people among us. Python in particular opens doors for automating and simplifying our electronic lives that would make any FGO happier. And, there are valuable new management and leadership tools and paradigms available to officers. Separately, these skills will make you a better FGO. Together, they will make you one of the rare FGOs that also understands Artificial Intelligence and Machine Learning. Please add to the conversation by commenting below if you have a good resource that you have found useful or implementation experiences to share!

Subscribe to The Field Grade Leader!

Maj Kurt Degerlund is a USAF officer. He is a member of the Military Writers Guild. The views expressed in this article are those of the author and do not reflect the official policy or position of the U.S. Air Force, the Department of Defense, or the U.S. Government.

[1] Department of Defense “Summary of the 2018 Department of Defense Artificial Intelligence Strategy: Harnesssing AI to Advance our Security and Prosperity” 2018, p. 14

The post Leader, Upskill Thyself: Learning Artificial Intelligence (AI) and Machine Learning (ML) as an FGO appeared first on The Field Grade Leader.