AI tools are capable of performing, making it challenging to encapsulate the field within a concise framework.

 

AI definition and demystification

The landscape of AI is characterized by dynamic evolution, resulting in a lack of a singular, universally accepted definition. This is primarily due to the broad range of tasks and outputs that AI tools are capable of performing, making it challenging to encapsulate the field within a concise framework. Despite the absence of a definitive definition, several authoritative sources offer valuable perspectives on the essence of AI. A foundational definition, attributed to John McCarthy, who coined the term "Artificial Intelligence," describes it as the science and engineering of creating intelligent machines. AI is a technology that enables computers or machines to mimic human perception, comprehension, problemsolving, decision-making, creativity, and autonomy. Another relevant definition has been provided by the European Commission: “AI refers to systems that display intelligent behaviour by analysing their environment and taking actions – with some degree of autonomy – to achieve specific goals. AI-based systems can be purely software-based, acting in the virtual world (e.g. voice assistants, image analysis software, search engines, speech and face recognition systems), or AI can be embedded in hardware devices (e.g. advanced robots, autonomous cars, drones or Internet of Things applications).” In accordance with these definitions, at its core, AI endeavors to equip machines with the ability to execute and automate tasks that traditionally necessitate human intelligence. This involves creating systems that can perceive their environment, learn, reason, solve problems, and understand human language. AI is a technology that deals with enabling intelligence in machines or computers. The source of this intelligence is usually data, sometimes combined with physics knowledge (e.g. thermal, structural, or mechanical)—especially when machines try to learn about other systems. Learning from data can be seen as analogous to how humans learn from experience in their lives. Today, data-driven AI technology can perform tasks such as language understanding and generation, object recognition and tracking in images or video sequences, time series analysis and anomaly detection in data recorded from sensors, and pattern recognition in vast volumes of data—with state-of-the-art performance in both speed and accuracy. AI technology can be extremely performant at specific tasks while lacking general intelligence or emotional capability like humans. It is vital to understand that AI is conceptualized, built, and deployed by humans, and is thus prone to biases like humans. In most cases, humans also operate these systems in the operational environment. The metaphor of AI as a "black box" underscores the often-opaque nature of its internal workings, where the complexity of connections within AI systems can be difficult for humans to comprehend fully. The concept of the "human primary thinking loop," while not a formally recognized term in cognitive science or psychology, allows us to understand how humans interact with the world cyclically: (a) sensing and perception; (b) processing and prediction; (c) decision and action; while learning throughout the loop thanks to feedback. The main AI domains (a.k.a. Application Classes) can thus be defined by aligning with the primary thinking loop: a) Sensing and perceiving the environment: Computer Vision, Pattern Recognition, and Natural Language Processing; b) Predicting: Hybrid Modelling; c) Deciding and acting: Decision Making and Generative AI. These Application Classes rely on technical fields like Probability and Statistics, Search and Optimization, or Knowledge Representation, while considering key enablers such as data, dedicated software and hardware, the business environment in general, human interaction, and regulatory compliance (through proper demonstration).