SOF, AI, and Changing Western Conceptions of War

AI is an emerging technology that will impact every sector and field across the globe. Militaries have a particularly hard challenge ahead, as mistakes in AI implementation cost lives. More importantly, there will be dramatic implications for the world order if Western adversaries can outmaneuver the U.S. and its allies. For the U.S military, the key to implementing AI will be Special Operations Forces (SOF). SOF has a unique skill set and ability to implement AI quickly and effectively and provide tangible real-world feedback on its performance. In this paper, I assess how AI will impact SOF operators and change Western conceptions of war. My research is informed by interviews with three retired SOF veterans: David Cook, who served in the 4th Psychological Operations Group of 1st Special Forces Command; Pete Chenko, a retired Marine Raider who remains active on the reserve force; and David Maxwell, a retired U.S Army Special Forces Colonel. Through my interviews, I gained an understanding of how AI will be used and how it will impact SOF operations. I focus on training, intelligence gathering/analysis, the role of SOF in the next generation of conflict, proposed uses for AI in combat, and the impact of AI on the levels of war. Ultimately, AI is the fulcrum of a seesaw; the East on one side, and the West the other. How America implements and operationalizes AI will determine which side of the seesaw goes up, and which side goes down.  

Pre-Mission: Training and Intelligence Gathering 

AI Enabled Training

My discussion with David Maxwell covered several applications of emerging technologies to Special Operations Forces namely quantum encryption, autonomous targeting, and the changing nature of SOF operations. However, what stood out to Maxwell the most was the use of AI for training purposes. Currently, SOF training relies on experience. AI can generate and replicate both new and old training scenarios. Military training has evolved from primarily field exercises to now include AI augmented VR training scenarios. AI creates reactive models that respond to operators’ decision making, creating unique training scenarios every time. AI/ML based training will also develop new training scenarios and perspectives continuously as the algorithms receive new data from both training results and information from updated adversary TTPs. Combined with the potential data sharing from allied nations and their training. SOF training can be continually adapted to the modern battlefield in real time, without requiring human capital to generate new training evolutions. The adoption of this next generation training will also remove constraints such as cost, resources, and risk to operators. 

Much of the discussion surrounding SOF and AI is centered around the human element – the operator themselves. Whenever humans are involved, subjective interpretations abound. Gaps in understanding or biases may result in inadequate training. However, AI training provides commanders with objective metrics, providing tangible insight into operator performance. AI assessments of operator performance reduces everyone to a number, more specifically their results. Taking this relatively ‘cold’ approach will remove subjective biases and preferences. Time does not equal success, just because an operator is a seasoned veteran does not mean they will outperform a new team member in every situation. AI shows us information quickly, but it can also show information that contradicts our underlying understanding, such as an unlikely combination of operators for a specific mission. Consequently, operators will be better prepared for their missions and commanders can plan missions with certain skill sets in mind. Undeniably, some variables such as intuition are hard to quantify. Experience certainly provides utility in the operating environment, but AI enabled training will provide commanders with a new perspective and an enhanced ability to hand pick operators for a specific mission. 

However, there are risks to collecting training data. The most important asset to Special Operations Command (SOCOM) is its operational data, both mission and training related. SOCOM seeks to leverage tools like AI and data analytics to advance its operational efficacy but requires an efficient and secure information landscape to achieve its goals. Instead of a siloed approach to network security, SOCOM can employ a strategy of network wide observability (NWO). In contrast to traditional network security practices, NWO is a holistic practice – enabling SOCOM to conduct security assessments of its entire information space. As the modern battlefield becomes increasingly interconnected, SOCOMs internal security must follow the same trend. NWO will provide SOCOM with a more secure environment, and more awareness of their internal systems.   

While SOCOM should implement a more holistic approach, it should still be siloed from the rest of the DoD. Isolating SOCOMs network from the rest of the DoD will protect it from network breaches via lower-level entities. Information and Communication Technology (ICT) security will determine the success and viability of operations in the information age, emphasizing the need to develop emerging technologies in tandem. If adversary nations can attain operational quantum computing first, traditional encryption will be rendered obsolete. Consequently, adversaries could more easily access all SOCOM’s operational data The fifth domain is playing an increasingly important role in modern conflict. Advances in ICT have elevated military capabilities worldwide. However, the same advances have also increased the vulnerabilities adversaries seek to exploit.  

Intelligence Gathering/Analysis 

Through my interviews I was able to identify two important backend functions of SOF that AI will dramatically improve upon. Maxwell spoke about the importance of AI for an ‘area study’ (AS) and Pete Chenko spoke about building a ‘pattern of life’ (POL) for a target. Both activities currently rely heavily on human capital and take considerable time due to the amount of information that needs to be gathered and assessed. AI will dramatically reduce the resources required and time spent to attain accurate intelligence. 

Area Study

Maxwell described an area study as attaining working knowledge of a prospective area of operations (AO). Understanding the social fabric of a society, the economic factors at play, culture, and general sentiment felt by a local population, inter alia. Not only will AI be used to gather and synthesize this information, but it will also be used to predict outcomes. AI can simulate complex systems and the multiplicity of factors that SOF evaluates in an area study, including how a local population will react to the operation at hand. AI will leverage an attempt to understand an area of operation, into a roadmap for strategic success.  

Pattern of Life

A pattern of life may refer to close surveillance of a specific target, generic activity within a confined space such as a city square, or the compilation of sensor inputs to evaluate nefarious activities. Compiling enough information to form an accurate POL is measured in the hundreds of hours. This practice is not sustainable as the world moves towards a more complex and interconnected geopolitical dynamic. This is where AI will revolutionize POL assessments. AI is uniquely attuned to performing image and signal processing – tasks that currently require a disproportionate amount of human intelligence. AI can also continuously update and re-evaluate an existing POL – analogous to the reactive training.  

Both the AS and POL are back-end activities that serve vital roles in mission preparation. AI can be used to create precise AS/POL faster than humans, increasing operational efficiency and speed while enabling the expenditure of previously used human intelligence elsewhere, a concept best exemplified in Afghanistan.  

In 2020 U.S. commanders were expending ~60% of their efforts towards figuring out how America’s objectives would impact the thought process of the Taliban and the local population. Ultimately, war is about resources, and time is among – if not the most – important resource. AI-enabled analysis models can free up a significant amount of commander’s time by simply providing the information. SOCOM is using AI visualization software to analyze social trends, population agitation, and opinion analysis, inter alia, to give operators a holistic view of their AO. AI can also provide predictive outcome analysis. AI systems can synthesize complex systems and determine the likely outcomes of interactions between governments, populations, and military objectives. Both uses of AI – short and long term – have an immediate impact on the tactical and strategic levels of war. Commanders have more time to devote elsewhere, operators have a greater understanding of their operating environment, and strategic planners can ‘test’ operations and decisions against AI. Maxwell told me AI can speed up our OODA loop by giving us a greater understanding of the operational environment. In all three examples of the impacts on the levels of war, this is exactly what happens.  

Operational Implementation 

New Era of Conflict

When asked about the next era of combat, David Cook emphasized the growing prevalence of irregular war and how conventional deployments and conflict will decrease in frequency. Since the first Gulf War in 1991 Western adversaries have shied away from conventional kinetic battles against the United States. A point that was stressed across all three interviews was that we are returning to an era of great power competition. As such, the strategic focus will be less on the non-state actors who have occupied the United States and its Western allies’ conventional attention for the better part of the last 20 years.  

Given the backdrop of irregular warfare, Cook emphasizes the non-linear nature of SOF. Cook explained that SOF is the only military capability that can impact the information and humanitarian domains. A point that David Maxwell corroborates, saying that conventional capabilities are an important aspect of SOF, but influence plays a large role as well. Exploiting the influence capabilities of SOF will be paramount to strategic success in the next generation of conflict. Given the rise of China and their ideals, we may enter an era reminiscent of Cold War style propaganda. China is already making long term strategic advances like the Digital Silk Road. Fueled by subsidized technology, China is seeking to gain more influence around the structure and regulation of the internet. Therefore, SOF may focus more on influence campaigns compared to the past. As Maxwell stated, it is about winning minds and will – not hearts and minds. Winning over the will of a local population generates change, a challenge that SOF is uniquely equipped for. Countering adversary narratives and influence will determine strategic success or failure. Additionally, Cook emphasized the importance of publicly available information (PAI), commercially available information (CAI), and open-source intelligence (OSINT). Leveraging AI with PAI, CAI, and OSINT will also enable faster and more accurate area studies, and pattern of life analysis. 

AI provides benefits to SOF in both the tactical and information domains. The discussion surrounding AS/POL demonstrates that AI can provide tailored messages and strategic roadmaps to exercise the most influence in a foreign AO. Using AI Natural Language Processing models for language translation can greatly improve an operator’s utility in a country. However, on the tactical side AI can be used to actively survey and report on a combat zone, providing real time updates on adversary movement. When asked about the use of AI on the battlefield, Pete Chenko said that it would have been useful during his deployments overseas as a Marine Raider for mapping placement of IEDs, monitoring adversary voice chatter, and enhancing situational awareness.  

Implementing AI in Combat

The Pentagon’s AI strategy emphasizes software, a departure from a previous reliance on conventional military capabilities. SOF are also known to test early developments of emerging technologies successfully, a point that is echoed by Chenko, who said that the operators on the ground will be deciding how the technology is used as they cannot afford to wait for policy to dictate how these technologies are used when faced with a life-or-death decision. SOCOM’s successful uses of AI can then be scaled across the entire military. Conversely, if SOCOM determines a new AI tool to be ineffective, this will save the DoD time and money. Modern conflict hinges on innovation – particularly in the formative years of AI warfare. If the DoD rolled out a new AI tool en masse but it proved faulty or ineffective, they would need to fall back on previous TTPs. SOF assessment of emerging technologies will assist in continuous innovation for the DoD as a whole, as they can provide real world assessments of the technology’s utility at a comparatively lower cost. 

Entering this new era of conflict, SOF must attain cognitive overmatch against adversaries. This necessitates operating in a networked environment of sensors and ICT. Society has become increasingly reliant on modern ICT, a dependence that adversaries will continue to exploit, particularly at the nation-state level.  

The CIA Triad is at the forefront of cybersecurity when discussing how to protect assets and networks. When asked about the role AI will play in deception operations, Cook emphasized the poisoning of Language Learning Models (LLM), or the integrity of the data the AI models are using to calculate decisions. Securing the AI models must be a top priority. As AI escalates the capability of SOF and the military, it equally increases the negative consequences of a cyber-attack or misuse of technology. This is why the human element is so vital to AI implementation. If we rely solely on AI and it becomes compromised, so too will all decisions commanders make. A loose analogy to Stuxnet demonstrates this. 

Stuxnet, the alleged U.S/Israel cyberattack caused Iranian nuclear centrifuges to effectively destroy themselves. Although not in an alarming way, it was through subtle manipulation such as altering internal pressure or speeding up/slowing down the rotor speed. The result was a failure to produce the desired uranium and additional expenses incurred to replace broken centrifuges. Because of the subtle nature of the manipulation, Iranian scientists could not identify the cause of their equipment failure for over one year. If an AI model is compromised in the same way as the Iranian centrifuges, we may follow its direction to our own peril. 

Changing Western Conceptions of War 

Levels of War

An additional change AI may lead to is a shift in the structure of the levels of war. The levels of war are defined as strategic, operational, and tactical. The strategic level represents the overarching strategic goals of a nation state, the operational level is where specific operations are planned within a particular war, and the tactical level is where individual missions are executed. Currently, the influence is top down. Policymakers provide strategic goals, planners create operations, and warfighters carry out missions to achieve objectives. I argue that AI implementation – particularly in the early stages of adoption – will change this linear structure.  

The advent of AI in the military domain is still new, and its full capabilities are still unknown. Strategic level planners cannot set long term objectives without an understanding of how to achieve said objectives. America’s strategic objectives cannot be purely theoretical. SOF will pioneer the use of AI; only by using AI in the field will operators determine the practical use and limits of AI. Therefore, tactical operations will inform the strategic objectives set for the next generation of warfare, resulting in a ‘bottom-up’ adaptation to the levels of war with the tactical level serving as the main source of influence. Once we plateau from the learning curve and AI capabilities become commonplace this dynamic may revert to the ‘top down’ – strategic, operational, tactical – dynamic we are familiar with.  

As Clausewitz stated, the nature of war is consistent but the character of war changes. Fundamentally, AI is the latest ‘character’ shift in warfare that will have permanent effects on the Western conceptualization of warfare. Pete Chenko told me there are initiatives at SOCOM to decrease the gap between the tactical and strategic levels as the gap between the two levels results in a bureaucratic process that is too slow to adapt to modern technology. As an example of this, Chenko pointed to the ongoing war in Ukraine. Cook states that technology will skip the operational level, and that success – or failure – at the tactical level will become strategic very quickly. Cook cited Ukraine’s FPV drone footage at the tactical level being used on the strategic level as a part of narrative warfare and even raising money. This theme will likely remain consistent throughout AI generation warfare. While the use of on the ground footage to influence hearts and minds across the world is nothing new, as Cook points out, adversaries can now propagate information at a pace and scale which is unprecedented. Combined with deep-fake technology, this dramatically shifts the narrative environment.  

Increased interconnectivity has slowly closed the gap between the tactical and strategic levels as militaries and non-state actors can no longer ‘hide’ their actions. One insurgent can upload a video to the internet, with the potential to reach the entire globe. Strategic level commanders can attain a live and accurate view of the tactical level and communicate directly with those on the ground. Thus, what we see is a symbiotic relationship between the tactical and strategic levels of war. When discussing POL assessments, Chenko stated that it used to be about attaining one phone number or IP address. Now, there are so many more data points to connect and understand. This trend is not unique to POL assessments: the entire conflict ecosystem is more complex than ever before. The levels of war can no longer be understood as a linear function.  

Conclusion 

Each generational shift in technology impacts military operations. Consequently, a shift in military training, command, and promotion structure should follow. Much of the conversation surrounding AI makes it seem like an unprecedented esoteric concept. While this is partly true, the same was said about steam engines during the Industrial Revolution. Simply put, AI is the next technological breakthrough and there will be more after it. As Clausewitz stated, the character of war changes, not the nature of war. A willingness to adapt while following strategic tenets will enable us to weather the storm and thrive in AI generation warfare. Failure to do so will only bring obsolescence while America’s adversaries gain global hegemonic status. Proper implementation of AI will result in faster decision making, more accurate intelligence, improved resource allocation, better spatial awareness, more effective messaging, and more impactful strategies. The key to reaching this level of success is SOF. SOF is uniquely equipped and trained to implement AI quickly and effectively, delivering results that can be scaled to the rest of the military.