AI & Generative AI: a companion reader

Objectives

• Recall the big-picture map of AI → ML → deep learning → generative AI → LLMs
• Recall the one-line reason prompting, verification and responsible use each matter
• Know where to go next in the full course for depth on any single idea

The big picture: AI, ML, deep learning, generative AI

AI is the umbrella term for any system that performs a task we'd call intelligent if a human did it. Machine learning is the dominant technique inside modern AI — instead of a programmer writing every rule by hand, the system learns patterns from examples. Deep learning is a family of machine-learning models (neural networks with many layers) that turned out to be extremely good at learning from messy, high-volume data like images and text. Generative AI is what you get when a deep-learning model — usually a Transformer — is trained to create new content (text, images, audio, video) rather than only predict a label or number. Each layer in this chain is a specialisation of the one before it, not a competing, unrelated technology — this is the single most useful mental map for the entire course.

Why prompting is a real skill, in one line

An LLM generates its response by predicting the next likely word given everything in its context — and your prompt IS that starting context. That is the entire reason a specific, well-structured prompt (context, role, task, constraints, input, output format) reliably beats a vague one: it narrows down what a 'good next word' looks like at every single step of generation. Few-shot examples, decomposition of big tasks, and iterative refinement are all just practical techniques for controlling that context more precisely.

Why you must always verify, in one line

A generative model is optimised to produce statistically plausible-sounding text, not verified truth — so hallucination (confident, fluent, wrong output) is a structural feature of the technology, not a rare bug. The practical fix is not to distrust AI entirely, but to scale your verification effort to the real-world stakes of the specific claim: light for a brainstorm, rigorous for a number that will go on an invoice, a legal document, or a medical decision.

Why responsible use is not an afterthought, in one line

The people who benefit from an AI system deploying successfully and the people who bear the cost when it fails are often two different groups of people — a business saves time, while a wrongly-screened applicant or a customer whose private data leaked pays the real price. Ethics, privacy, bias, copyright, security and human oversight are the practical checklist for noticing and addressing that imbalance before it causes real harm, not abstract philosophy bolted onto a technical course.

Where to go next

This reader is a map, not a replacement for the full course. Open /courses/ai-generative-ai for all 92 original lessons across 13 modules, with diagrams, worked examples, quizzes, and four hands-on projects (AI Study Assistant, AI Business Content Assistant, AI Customer Support Assistant, and a final AI-Powered Business Automation System design). When you feel ready, attempt the course's final skill exam to check your overall understanding.