What AI actually is

Artificial intelligence is software that performs tasks we normally associate with human judgment: recognizing a face, translating a sentence, ranking job applications, drafting an email. The key difference from ordinary software is how it usually gets there. Classic software follows rules a programmer wrote down. Most modern AI, including every generative AI tool in this course, is instead given examples, huge numbers of them, and works out statistical patterns from those examples.

This is why AI behaves unlike most software you have used before. It can handle situations nobody explicitly programmed, which is its power. And it can fail in ways nobody programmed either, which is its danger. A rules-based system tends to fail loudly and predictably. A pattern-based system more often fails quietly and plausibly: it produces something that looks right.

The vocabulary you actually need

• Machine learning: the family of techniques where software learns patterns from data instead of following hand-written rules.
• Model: the trained result. When people say "the AI", they usually mean a model.
• Training data: the examples the model learned from. Everything a model knows, it learned from this data, including the data's gaps and prejudices.
• Generative AI: models that produce new content (text, images, audio, code) rather than just classifying or ranking existing content. ChatGPT, Claude, Gemini, and Copilot are generative AI.
• Prompt: whatever you type or upload to a generative AI tool. Prompts are the main thing YOU control, and the main channel through which data leaks happen.

That is genuinely all the theory this course needs. Every risk and every habit in the coming chapters follows from one sentence: AI is pattern-matching on past data, so it is confident, fast, plausible, and sometimes wrong.