Guide 2 - Extended Edition

Guide 2 - Extended Edition

Artificial Intelligence (AI) refers to systems that perform tasks typically requiring human intelligence—such as learning, reasoning, pattern recognition, and decision-making. It is not one technology, but a broad field encompassing machine learning, natural language processing, computer vision, robotics, and more.

1. Core Concepts

Machine Learning (ML)

ML is the most practical and widely used subset of AI. It allows systems to learn from data and improve over time without being explicitly programmed. ML algorithms fall into three categories:

• Supervised learning: Trained on labelled data (e.g. spam detection).
• Unsupervised learning: Finds patterns in unlabelled data (e.g. customer segmentation).
• Reinforcement learning: Learns through trial and error (e.g. game-playing agents).

Neural Networks and Deep Learning

Neural networks are inspired by the human brain. Deep learning uses large neural networks to identify complex patterns. These models power image recognition, speech processing, and large language models like ChatGPT.

2. How AI Is Used

Business Automation

AI is commonly used to automate repetitive or rule-based tasks—customer service chatbots, invoice processing, document classification, etc.

Decision Support

Predictive analytics powered by AI can help with demand forecasting, risk analysis, and personalised recommendations (e.g. Netflix, Amazon).

Language and Vision

AI powers voice assistants, translation tools, facial recognition, object detection, and handwriting recognition.

3. Building an AI System

To build an AI solution:

1. Define the problem – Be specific. “Classify support tickets” is better than “automate support.”
2. Collect data – The quality, quantity, and relevance of your data directly impact performance.
3. Train a model – Use libraries like Scikit-learn, TensorFlow, or PyTorch. Most models require iteration and tuning.
4. Evaluate – Use metrics like accuracy, precision, recall, or F1-score depending on the task.
5. Deploy – Package the model into an API, integrate into your workflow, and monitor it in production.

4. Risks and Considerations

Bias and Fairness

AI learns from historical data, which may contain human biases. Without safeguards, models can reinforce or even amplify discrimination.

Transparency

Many AI models are black boxes. Stakeholders must understand how decisions are made—especially in sensitive areas like hiring or lending.

Security and Misuse

AI systems are vulnerable to adversarial attacks and misuse. For example, deepfakes and automated phishing are real-world risks.

5. Responsible Use

Responsible AI involves:

• Human oversight: Keep a human in the loop, especially for critical decisions.
• Auditability: Maintain logs and version control of models and data.
• Data governance: Ensure data privacy, consent, and security.
• Monitoring: Continuously evaluate the model’s behaviour over time.

Final Thoughts

AI is not magic, but it is a powerful tool. It’s best used to augment—not replace—human decision-making. Success with AI comes from a solid understanding of your problem, good data practices, and responsible deployment. Start small, iterate often, and always measure impact.