Chapter 1 Foundations of Applied AI Engineering

Understanding AI, Machine Learning, and Deep Learning

Artificial Intelligence (AI) is a broad domain focused on building systems capable of performing tasks that typically require human intelligence, such as reasoning, learning, and decision-making. Machine Learning (ML) is a subset of AI that enables systems to learn patterns from data rather than relying on explicitly programmed rules.

Artificial Intelligence (AI) is an umbrella field focused on building computer systems that can perform tasks typically associated with human intelligence. These tasks include reasoning, learning, perception, language understanding, and decision-making. Rather than being a single method or technology, AI encompasses a wide range of techniques, from rule-based systems and expert systems to modern data-driven models. The primary objective of AI is to enable machines to act intelligently within specific environments, adapting their behaviour based on inputs and goals. Applications of AI appear across domains such as healthcare diagnostics, autonomous vehicles, financial forecasting, virtual assistants, and robotics, demonstrating its versatility in solving real-world problems.

A central component of AI is Machine Learning (ML), which emphasizes learning from data instead of relying solely on explicitly programmed instructions. In traditional programming, developers define rules step by step; in contrast, ML systems infer patterns automatically by analysing examples. During training, algorithms examine historical data to discover relationships between inputs and outputs, allowing them to make predictions on new, unseen data. This approach is particularly valuable in dynamic or complex environments where manually defining rules would be impractical. Common applications include fraud detection, recommendation systems, demand forecasting, and medical risk assessment. For instance, an email filtering system can learn to distinguish spam from legitimate messages by studying previously labelled emails, continuously improving as more data becomes available.

Deep Learning represents a specialized and highly powerful subset of Machine Learning that relies on artificial neural networks with many processing layers. These networks are inspired by the structure of the human brain, where interconnected neurons transmit signals to process information. In deep neural networks, each layer extracts increasingly abstract features from the data: early layers might detect simple patterns, while later layers capture complex structures. This hierarchical learning capability enables Deep Learning models to excel at handling large amounts of unstructured data such as images, speech, and natural language. Breakthroughs in computer vision, speech recognition, machine translation, and generative AI have largely been driven by deep learning techniques, supported by advances in computational power and large datasets.

Recognizing how AI, Machine Learning, and Deep Learning relate to one another is essential for selecting appropriate solutions. AI defines the overall ambition of creating intelligent machines, ML provides practical algorithms that allow systems to learn from experience, and Deep Learning offers advanced tools for tackling highly complex tasks that require extracting subtle patterns from massive data. Each level has its own strengths and trade-offs: simpler ML models may be more interpretable and require less data, while deep learning models often achieve superior accuracy but demand significant computational resources. Developers and organizations must therefore consider factors such as problem complexity, data availability, performance requirements, interpretability needs, and deployment constraints when choosing among these approaches. Together, these technologies form a layered ecosystem that continues to drive innovation across science, industry, and everyday life.

1. Artificial Intelligence (AI) – Rule-Based System

Example: Chess Game Program

A traditional chess engine follows predefined rules and logic to decide moves.

• It evaluates possible moves using programmed strategies.
• It does not learn from experience unless explicitly updated.

Key idea: Intelligence is created through rules written by developers.

2. Machine Learning (ML) – Learning from Data

Example: Email Spam Filter

An email system learns to detect spam by analysing thousands of emails.

• It identifies patterns like suspicious keywords, sender behaviour, etc.
• Over time, it improves as more data is provided.

 Key idea: The system learns patterns from data, not fixed rules.

3. Deep Learning – Complex Pattern Recognition

Example: Face Recognition System

A face recognition system (like phone unlocking) uses deep neural networks.

• It automatically learns features such as eyes, nose, and facial structure.
• Works well with images, videos, and large datasets.

 Key idea: Uses multi-layered neural networks to handle complex data.

Quick Comparison Table

Technology

Example

How it Works

AI

Chess engine

Predefined rules and logic

Machine Learning

Spam email filter

Learns patterns from data

Deep Learning

Face recognition

Uses layered neural networks

Real-World Combined Example

Self-Driving Cars

• AI: Overall system decision-making (drive, stop, turn)
• ML: Learns driving patterns from data
• Deep Learning: Detects objects like pedestrians, cars, signals