Creating the edge in Military Capability through AI & ML
Before the new millennium rolled in and we saw rapid, accelerated development of technology across various spectrums, the idea of ‘artificial intelligence’ and ‘quantum computing’ were the domain of science fiction literature, even though the theoretical underpinnings in quantum theory and mechanics had been around for several decades. Today, however, the debate has shifted from whether it is a possibility to when it will become a possibility. Artificial intelligence (“AI”) and Machine Learning (“ML”) technologies hold the potential to become critical force multipliers in future armed conflicts – with the objective of identifying the right factors, activities, posture, and capability development to advance AI and ML technologies in order to boost national competitiveness and harness them in order to safeguard national security.
The reality of the times that we live in is that the strategic edge provided by AI, to both state and non-state actors on the global stage, is something that is incomparable, for it far surpasses the technological advantage and the sheer destructive capability that the invention of the atomic bomb provided to the beneficiaries of the Manhattan Project. In fact, it would not be incorrect to state that there is no industrial-age innovation that threatens to change the face of warfare more than the application of AI technologies.
Therefore, it has become imperative, now more than ever perhaps, to drive investment and implementation policies; and ensure that investments and policies contribute to and protect AI/ML growth in our country’s commercial, academic, and government sectors. The challenge of utilizing AI and ML advancements in a manner so as to add an edge to our military capabilities requires us to leverage said enhancements in the commercial and research sectors, transforming them into concepts that can be applied to the Indian military, undertaking an assessment of their efficacy and suitability to our operations, and evolving existing operational, functional and deployment strategies whilst developing new ones that can take advantage of these newly created capabilities.
Keeping in mind this prerogative, it is evident that there are five main parameters that drive the country’s AI/ML development and play a substantial role in creating and retaining a competitive edge in its military capability: first, fundamental research breakthroughs; second, commercial industry advancements; third, undertaking development and engineering required to transition AI & ML technology to the military; fourth, implementing an evolved, comprehensive verification, validation, test, & evaluation metrics; and fifth, the conceptualization and development of an operational stratagem.
Nonetheless, there are five key areas and initiatives that require execution at a national policy level to ensure that the country develops and retains the AI/ML edge in the battlefields of tomorrow, namely, (i) through initiatives that can push fundamental research breakthroughs in the commercial space; (ii) having the requisite AI & ML Developmental Roadmap for a vision that the stakeholders can work towards; (iii) initiatives that can help in creating a military AI/ML pipeline for the defense establishment of India; (iv) a focus on developing and formulating verification, validation, testing and evaluation mechanisms for assessment of AI & ML technologies; and (v) developing operational concepts and analytical wargaming frameworks to enhance capabilities.
- Fundamental Research Breakthroughs: Whilst fundamental research breakthroughs are vital for the research, development and analysis of newer, previously unrecognized, dimensions of the application of AI & ML technologies, it is imperative to recognize that these breakthroughs are rather unlikely to be the differentiating factors between two nation states vying for military supremacy. This inability to provide a differential large enough to attain supremacy is partly down to two key reasons: firstly, the underlying research is driven primarily by academia and commercial industries, therefore the thrust areas are rarely ever in sync with what would be determined as important by the military; second, these industry-driven findings and developments are generally published and therefore form a part of open literature, making them accessibly globally, and therefore reducing / nullifying any advantage that may have been sought from such development. Nonetheless, it can form the foundational pivot for any AI & ML technologies that are created in a nation, which would necessitate certain steps must be taken to create a framework that can drive such research forward. There are two key areas that require action: creating avenues that can foster collaboration between Indian and foreign academics; and the integration of industry and military interests – which seems to be a larger, more pressing problem that needs to be tackled.
- AI & ML Developmental Roadmap: The importance of having a stoic vision, only increases further when a particularly critical technologies, such as AI, are considered, as they are the subject of grossly exaggerated and nonstop coverage in the media. The lack of clarity about the current capabilities and limitations, of the contemporary AI & ML technologies that we envision, may doom any attempt to incorporate them into useful systems. Therefore, as a first step, it is necessary to exercise great care and caution whilst ensuring that the lines between fictional speculation and attainable reality do not blur, because AI & ML systems have immense potential beyond all the hype, even if it may not be past the threshold that science fiction enthusiasts would envisage. To keep the country’s military AI & ML efforts attainable and resolute, it would be recommended that a forward-looking AI roadmap (referred to hereinafter as the “AI Roadmap”) is created. The proposed AI Roadmap must be able to provide a balanced perspective on realistic goals for military AI deployment across different / various time horizons. This AI Roadmap must be, ideally, organized according to three periods: the near term (attainable in the next one to two years), medium term (attainable in the next three to five years), and long term (attainable in the next six to ten years) terms. This Roadmap must undertake the following:
The AI Roadmap must be treated as a dynamic and living document that must be constantly maintained and revised as per the operational requirements of the military and as the AI and ML technologies evolve through their lifetime. If executed adequately, the AI Roadmap, the AI Application Areas, operational planning and R&D requirements will coevolve symbiotically. The intent behind the AI Roadmap should be to ensure that the military’s equipment modernization and recapitalization planning does not get too far ahead of the pace of the application growth that the scientific AI/ML-focused R&D community provides. If deployed correctly, the AI Roadmap would also help the tri-services and the defense industry from squandering precious acquisition funds on concepts that are not achievable with current AI/ML applications and keeping their vision and development grounded and realistic.
- Identify a list of application areas for the military deployment / utilization of AI & ML technologies (referred to hereinafter as the “AI Application Areas”) that are to be achieved for the near term, medium term and long term time-periods.
- Enumerating the details and prioritization basis for the AI Application Areas for each relevant period as outlined above.
- Laying down, for each AI Application Area, a plan which outlines the transition cycle to operations, inclusive of test and evaluation undertaken for the AI/ML technology being sought to be developed.
- Listing down shortfalls/deficiencies in the current Indian military arsenal of AI & ML technologies for each of the AI Application Areas that need to be developed to maturity as planned.
- For each shortfall/deficiency identified in pursuance with the above-mentioned point, identify and suggest possible approaches that could succeed given the nature of the deficiencies of the extant system in the said AI Application Areas.
- Identifying the requisite R&D required for each of the AI Application Areas, along with the manner in which such R&D can adequately deal with the extant deficiencies.
- For each of the AI Application Areas, offer a rough-order-of-magnitude estimate of the total cost to bring the AI Application Area to maturity and successfully develop the same, through the entire developmental cycle process.
- Creating a Military AI Pipeline: To establish and maintain a competitive edge in AI & ML development and application, the vast technical expertise available to the Indian research fraternity must be leveraged, tapping into our extensive base of engineers, scientists and visionaries possessing adequate skillsets (collectively referred to hereinafter as the “Technical Professional Base”) pertinent to AI & ML Application Areas. The intent behind driving this particular initiative should be to equip our Technical Professional Base with adequate know-how, and the understanding (from a military application POV) so as to enable them to undertake four key activities and use their technical expertise adequately: analyze requirements; ideate and develop design, test & evaluation techniques for AI-warfare based operational concepts; create a structural process for engineering AI systems specific to the Ministry of Defence; and conduct the entire cycle of VVT&E of the AI/ML systems from the laboratory to field application. It must be mentioned however, that, creating and maintaining such a professional workforce may prove to be difficult, particularly given the high salaries and superior equipment offered in the private sector. Further, it must be outlined that the Technical Professional Base needs to be grown in a manner so as to ensure that it contains a substantial number of personnel that are skilled in the information technology-related disciplines that are crucial for the development of AI systems and technologies.
While the retention problem with respect to the Technical Professional Base is significant, there are certain measures that must be taken by the Defense Ministry, aimed at the structural lacunae persisting in our current system, that will help ensure that the military has the minds it needs to stay at the forefront of AI/ML research, development and application. Firstly, it is recommended that the current workforce of technical experts be revitalized through the assimilation of newer members. Furthermore, and this may be the biggest challenge for the Indian defense ecosystem as it hits right at the heart of the paralysis caused by the current structure that we have in place, the military-backed R&D programs must offer professionals more-flexible career paths, along with salaries based on performance, not just seniority; and the opportunity to continue conducting research, instead of being forced into management at a certain level of seniority. It would also be a step forward if members of the Technical Professional Base could receive more professional development opportunities, with opportunities to attend conferences, publish papers, and perhaps work on internally funded personal research. Finally, greater funds should be earmarked and allocated for a military construction and infrastructure budget in order to update the old and aging laboratory equipment in many of the country’s defence labs to ensure that lack of infrastructure is not what slows our pace down as we look towards the future. Further, the armed forces can, through their professional military education institutions, ensure that new recruits study AI & ML applications so that, when they are promoted to leadership positions, they will be knowledgeable about AI/ML techniques and capabilities and able to use these capabilities effectively in their future commands. Across a similar tangent, it is advisable that the defense establishment of the country incorporates and kickstarts AI & ML-focused programs at major universities across the country where AI & ML are key subjects to attract newer talent into military research.
- Formulating ‘Verification, Validation, Test & Evaluation’ (“VVT&E”) Techniques suitable for assessment of AI/ML technologies: The formulation of VVT&E techniques for assessing AI / ML technologies can borrow from extant practices and procedures currently in place in order to ensure a strong framework, for instance concepts from the paradigm of model checking and robust control . The assimilation of such practices and creation of this VVT&E technique would require significant investment in research in order to bring the scientific community together and effectively deal with problems and roadblocks that traditional / extant testing methods may pose. Indeed, efforts must be made to view these four processes as being one entity, wherein validation, verification and evaluation processes can be incorporated into the design process itself, thereby cutting the developmental cycle timeline short and providing for more flexible research, development and deployment framework. A resolute VVT&E system must ensure that there are processes in place that avoid side effects (minimal or nullified effects on areas not directly relevant to the AI Application Area), avoids reward hacking (ensuring that the AI / ML systems do not resort to any shortcuts that may produce less than desirable results), scalability (ensuring that complex processes and objectives do not render the AI system ineffectual by taking cumbersome decision making algorithms and paths), ensuring exploration safety (ensuring that the AI does not attempt to undertake activities that are exceptionally dangerous or dangerous beyond a certain threshold as it builds up its decision-making system through trial and error processes), and reducing operational shifts (ensuring that the training/simulation data and environment is not remarkedly different from the real-world environment or information that an AI/ML system is supposed to encounter).
Keeping in mind the essentiality of effective VVT&E systems for providing military personnel with the confidence and comfort to be able to harness the full potential of AI & ML systems at their disposal, it is necessary to ensure that the VVT&E processes in place are equipped to assess the reliability of autonomous and dynamic systems. At the moment, military VVT&E systems have not been designed with the self-learning behavioral characteristics of AI & ML technologies and will not be able to keep up with the emergent behavior that AI & ML systems exhibit with time. Therefore, testing and evaluation parameters must be relooked at to ensure that they are sufficiently equipped to represent the unpredictable, engrossingly complex, open-ended and adversarial environments that autonomous technologies operate in. This may be undertaken in two distinct parts: first, an emphasis on evaluating both the performance of the operator and the deployed systems as one singular entity to ensure that there is an optimization of human-machine interaction in a practical, efficacious sense; and second, by focusing efforts on evaluating each component of a technology / system as and when they are developed rather than reviewing such technologies once the entire system has been developed and assimilated. Additionally, an enhanced VVT&E is also required as there are no previous platforms or predecessors against which the performance of such AI / ML systems may be assessed against, which means that this may prove to be an exercise of putting blocks in place one by one. In terms of cutting developmental cycles short and providing flexibility, it is imperative to ensure that verification and validation of AI & ML systems plays a larger, more substantial role in the creation of early system requirements as assessed by the defense establishments. Since AI & ML technologies are built on several layers of code and algorithms, verifying and validating its more basic components early on in the development cycle can add to its reliability and prevent the requirement for any costly, time-consuming and growth-hampering fixed at the later stages of development by nipping these problems in the bud. This measure can also help accelerate the production process and ensure that prototypes are rolled out and deployed in a rapid manner.
- Developing Operational Concepts and Wargaming Capabilities: It need not be outlined that decisive military innovations have often separated themselves from their less-strategic & less-desirable counterparts through a display of practical effectiveness on the battlefield. This necessitates that AI & ML technologies are implemented properly through the execution of superior operational concepts that can leverage their potential to create & sustain decisive advantages in rapidly evolving battlefields. One such approach to developing these concepts can be through the use of analytical wargaming. Such games can serve as the cornerstone upon which new operational concepts and doctrines may be developed in order to leverage newer (in this case previously unforeseen) technologies for military superiority. It is advised that these wargaming efforts assess two situations – the current, extant capabilities and one wherein a multitude of AI & ML capabilities (identified and clubbed together based on the identification of the AI Application Areas as outlined in the first point hereinabove) are available – and subsequently analyze the results of the game to build a set of operational concepts for AI-supported military operations (“AI Operational Doctrine”). This set of AI Operational Doctrines can then be evaluated across several rounds of war games which can cover a broad range of conflicts that the military envisages may be possible and evaluate which of the AI Operational Doctrines is most effective in each scenario. For best results, it would be advisable that such doctrines are created keeping in mind the three scenarios envisioned whilst drafting the AI Application Areas, i.e. separate operational doctrines for near-term, medium-term, and long-term applications in order to ensure that the operational doctrine is cognizant of the dynamism presented by AI & ML technologies. Select Operational Doctrines can then be refined into white papers for circulation with the armed forces for their perusal, review and approval.
AI & ML technologies hold the potential to become critical force multipliers in future armed conflicts and the identification of AI will prove to be key to our goal of enhancing our national competitiveness and protecting our national security. If a well thought out, methodical AI plan is successfully executed, India can achieve a substantial military advantage over our adversaries and competitors. Creating this AI Roadmap and concentrating our allied efforts towards AI & ML can result in strong positive strategic implications for the country.