AI Basics: Understanding the Technology and Its Military Use
By Kirthi Jayakumar
AI: A Quick Primer
In simple terms, Artificial Intelligence (AI) is the ability of any computer system to perform tasks that typically require human intelligence - including things like problem-solving, decision-making, and understanding natural language. It mimics certain cognitive functions that are typically associated with human intelligence, but cannot think for itself. To understand AI, it is helpful to ground ourselves in a few key terms with one of the most commonly used metaphors: food.
To perform these functions, AI needs certain key components. First is data, which are individual pieces of information that can be processed by computers, including numbers, text, images, audio, video, and sensor readings, among others. Sometimes, data can be structured into a collection of related data items that are organised for a specific purpose – this is called a dataset. Think of these as ingredients to make a dish. Second, is an algorithm, which is a set of instructions or rules that guide the process of solving a problem or completing a task. Algorithms process data to produce results. Think of this as a recipe to guide how the ingredients will be used to make a dish.
A model is the mathematical representation or algorithm that is trained to make specific decisions. It processes the input data and produces the output data. Think of this as an experienced chef that uses the recipe and produces the final dish. Machine learning is a subset of AI, where the system learns patterns from the data instead of being programmed to approach every scenario – so instead of offering recipes for a variety of dishes, you offer up the dishes itself as examples, and the system learns from these examples to figure out what to cook. Think of this as learning to cook by doing, rather than by reading recipes. Natural Language Processing is a branch of AI that focuses on helping a computer understand, interpret, and generate human language – it is what helps AI systems to work with text and speech in meaningful ways. Think of this as communicating about food – where you’re helping the chef understand food preferences and dietary needs, for example. Natural language processing uses machine learning to train models to process language.
Precisely what happens within an AI model, such as how it learns and produces outputs remain unclear. There is some understanding that AI uses computing systems that function a bit like the human brain – where there are interconnected nodes that operate like neurons and pass information to each other. Each connection has a “weight,” which determines precisely how much influence one node has on another. When the neural networks have many layers, AI engages in what is called deep learning, where each layer learns progressively complex patterns.
AI can be classified into three categories from the point of view of capability:
Narrow, which is designed to perform a single, specific task - like a chatbot or a recommendation engine
General, which is a theoretical form of AI with human-level cognitive abilities that is capable of understanding, learning, and applying knowledge across a wide range of tasks
Superintelligence, which is a hypothetical AI that possesses cognitive abilities that far exceed those of the most intelligent humans.
It can also be classified by function into:
Predictive, which analyses existing data to make predictions and decisions on future outcomes and trends
Analytical, which extracts insights and knowledge from historical data to provide information but does not produce new content
Conversational, which helps machines understand and respond to human language through natural language processing
Computer Vision, which allows AI to see and interpret visual information from the real world, such as images and videos, and are used in medical imaging and self-driving cars
Natural Language Processing, which deals with the interactions between computers and human language, enabling AI to classify, understand, and generate text
Rule-Based, which uses pre-defined rules and logic to make decisions, with limited ability to transcend programmed instructions
Agentic, which combines analytical, generative, and automation capabilities to act autonomously, using AI to perform tasks and achieve goals.
The Military Use of AI
AI was originally developed for military use, and continues to be used for military purposes. Several forms of AI that are deployed for civilian uses have emerged from military innovation: such as GPS navigation, the internet itself, speech recognition software.
Currently, military engagements use AI for a variety of purposes. A significant area where AI is involved is intelligence, surveillance, and reconnaissance. As AI systems process vast amounts of data, they’re used to process large volumes of satellite imagery, intercepted communications, and sensor data to support the identification of patterns, threats, and intelligence targets. The combined capacity for precision and analysis of vast amounts of data make AI useful in drone surveillance as well – AI-powered drones can identify, track, and engage targets autonomously, unlike traditional remote-controlled drone systems. Similarly, AI has also been used for facial recognition to identify “persons of interest,” enhance base security, and distinguish civilians and combatants in complex operational environments. No system, however, is free of error and human bias, which means that the use of military AI is not fail-proof.
AI has also been put to use in weapons systems, namely in autonomous and semi-autonomous ones. For example, defensive systems like Israel’s Iron Dome automatically identify, classify, and intercept threats in the form of incoming missiles mere seconds after launch. Existing autonomous weapons such as “hardkill” active protection systems use AI, too, to autonomously identify and attack oncoming missiles, rockets, artillery fire, and even aircrafts.
Moving away from the battlefield, AI is also increasingly used in cybersecurity and electronic warfare. AI can detect and defend against intrusions, while also carry out offensive cyber operations. AI can support with signals intelligence as well – where it can automate planning and strategy, and fuse and interpret signals far more efficiently than can humans.
Ethical and strategic concerns around the use of Military AI
As AI continues to evolve, the need for effective governance mechanisms to manage its use and mitigate potential hazards also grows. The sustained military applications of AI raise significant questions about accountability, escalation risks, and the future nature of warfare. It also throws up many ethical and legal questions that the current legal framework does not adequately answer.
The dual use nature of AI, where it can be used for military and non-military purposes, use evokes many ethical dilemmas. On the one hand, these technologies can enhance national security, but doing so often comes at the cost of human rights. Should decisions in armed conflict contexts must be consigned to a machine at all? Who is responsible for military AI and its impacts: Is it the one who developed it, or the one who deploys it, or both, or the one who gives the command to do any or all of these things? When AI is trained on data that reflects social realities, including biases and prejudices, can we trust a machine to make life-and-death decisions?
Any attempt to engage with and use AI, both for military and non-military processes, would be incomplete without recognizing the role of power, and systemic and structural violence, as well as understanding the differentiated impacts of the same technology when seen through an intersectional lens. Over the course of the next 11 posts, this series will look at military AI, its history, and myriads of implications through a feminist lens.
References
Duchateau, N., Puyol-Antón, E., Ruijsink, B., & King, A. (2023). AI and machine learning: the basics. In AI and Big Data in Cardiology: A Practical Guide (pp. 11-33). Cham: Springer International Publishing.
Moore, J. Beginners Guide to Artificial Intelligence. https://condor.depaul.edu/jmoore/media-files/jmoore/tech-tuesday/Beginners-Guide-to-Artificial-Intelligence.pdf
Murgia, M. (2023). Generative AI exists because of the transformer. https://ig.ft.com/generative-ai/
TechnoSocio Advisory. 2024. GenAI Essentials: A roadmap to understanding inclusive LLMs. The Inclusive AI Lab.
TechnoSocio Advisory. 2024. The Building Blocks of Inclusive Artificial Intelligence (AI). The Inclusive AI Lab.
Yilmaz, H. (2024). What Is Artificial Intelligence?: Analyzing the Drawings of Preschool Children. Online Submission.
