The AI Pieces of the IMD Puzzle
By Leigh Harris and Ven Adamov
The age of artificial intelligence (AI) in warfare is here. There is no turning back.
The military applications of AI have become an essential part of military operations and will become increasingly important. But realizing the benefits of AI – and minimizing its potential risks – depends entirely on how well the technology is integrated into military operations.
With guest columnist SRC Can Ltd.’s Alex Clarke, we recently focused on the importance of high-fidelity Intelligence Mission Data (IMD) capabilities to maintain both tactical advantage and mission success in joint and coalition military operations. Here, we outline what’s needed to successfully implement IMD in the Canadian Armed Forces (CAF).
READ: Canada Needs IMD Now – DEFENCE BUSINESS
The primary components for a successful AI implementation in the context of IMD are technological interoperability, purposeful selection of suitable AI models, and robust data and AI governance frameworks.
Enabling a Scalable Cloud
The establishment of a scalable and interoperable cloud architecture is a critical component to implementing IMD successfully.
Such platforms must be capable of connecting to and processing data from multiple sources and formats, collected from Internet of Things (IoT) devices and sensors, military databases, geospatial intelligence (GEOINT), and even Open-Source Intelligence (OSINT). This combination enriches the intelligence picture, providing a comprehensive real-time view of the operational environment for modern military equipment. Recent examples of the successful incorporation of OSINT on the battlefield is how Ukrainian and Russian forces use photos and videos posted on the Telegram app for geolocation and drone targeting in near real-time.
The ability to seamlessly integrate multiple data streams is essential for creating a comprehensive and accurate operational picture and decision-making in complex and dynamic environments. Technologies like application programming interface (API) connectivity facilitate seamless data exchange between different systems, ensuring that intelligence can be shared and utilized effectively across platforms. For instance, RESTful APIs and GraphQL are commonly used to enable such connectivity.
A robust cloud infrastructure also ensures that data is accessible in real-time, providing actionable insights to military commanders. Cloud technology improves operational efficiency, leading to more informed and timely decisions on and off the battlefield.
Building AI Controls
Given the sensitivity, security, and importance of IMD, it is imperative to establish solid governance processes and controls for both data and AI models throughout the entire data and AI lifecycles:
- For data, this includes data collection, storage, processing, analysis and dissemination.
- For AI, it includes model intake, prioritization, development, deployment, monitoring and decommissioning.
Effective governance ensures that IMD is managed in a way that maintains its integrity, confidentiality, and availability, while also complying with relevant regulations and standards. In this way, AI applications can be used ethically and responsibly.
To operationalize data governance, utilizing various data management tools is essential. Cloud solutions, such as Informatica, Collibra, Microsoft Purview, or Snowflake, enable comprehensive data governance, including metadata management, data cataloging, integration, lineage, quality, security and accessibility. In addition, Responsible AI monitoring software solutions minimize risks such as hallucinations, inaccuracies and bias, increasing the reliability of AI model outputs.
Scalable cloud solutions that feature robust data and AI governance frameworks and tools also help mitigate data breaches, unauthorized access, and other security threats while promoting transparency and accountability in the use of AI, which is hugely important to foster trust among stakeholders and ‘Five Eye’ (FVEY) partners.
What’s Best for Real-Time IMD?
Another crucial factor is the selection of AI models that are best suited for processing IMD in real-time.
Each AI model has strengths and weaknesses, and it is essential to choose the ones that align with the specific requirements of military operations. Factors to consider include the model’s accuracy, speed, scalability, and ability to handle diverse data types. They must be capable of operating in dynamic and high-pressure environments, where timely and accurate information is critical for mission success.
Canada’s FVEY partners utilize a range of AI models and technologies for processing IMD. These include machine learning algorithms for pattern recognition, language models for analyzing communication intercepts, and computer vision for image and video analysis. These technologies enable the extraction of actionable intelligence from vast amounts of data, enhancing situational awareness and decision-making.
By learning from our partners and selecting the right AI models, the DND/CAF can enhance their situational awareness and decision-making capabilities, ultimately improving operational effectiveness.
The integration of AI into DND/CAF operations is a complex and multifaceted endeavor that requires careful consideration of various factors, including processes and technologies. By enabling scalable and interoperable cloud architecture, building solid data and AI governance processes, and selecting the best-fit AI models for real-time IMD processing, the DND/CAF can effectively leverage AI to enhance their operational capabilities.
As modern warfare evolves, these elements will be critical in ensuring the successful realization of DND/CAF AI strategy’s full potential.
Leigh Harris is a Management Consulting Partner and Lead Partner, Federal Government, at KPMG in Canada. Ven Adamov is a Partner and National Leader of Trusted Data & AI Services at KPMG in Canada. For more information, visit, www.kpmg.ca. The views expressed here are their own and do not necessarily reflect a CDR editorial position.
