2. Types of AI¶
"AI" covers a wide range of very different techniques, from simple rules to the large models everyone talks about today.
| Type | How it works | Example |
|---|---|---|
| Rule-based systems | Humans write explicit if/then rules | A thermostat deciding when to turn on the heat |
| Machine learning | The system learns patterns from examples, instead of being told explicit rules | Email spam filtering |
| Deep learning | Machine learning using many-layered neural networks (Unit 3 goes inside one of these) | Image recognition, voice assistants |
| Generative AI | Trained to produce new content (text, images, audio) that resembles its training data | Claude, ChatGPT, AI image generators |
Narrow AI vs. general AI¶
Every AI system that exists today, including the most advanced language models, is narrow AI — built and trained for a bounded set of tasks, however broad those tasks might seem. General AI — a system with human-like understanding across any domain — doesn't exist yet, despite how often it comes up in movies and media. Keeping this distinction clear is one of the most useful things you can take from this unit.
Try it
Classify each of these as rule-based, machine learning, or generative AI: a chess program following programmed opening moves, a system that flags unusual credit card purchases based on patterns in past data, a tool that writes a paragraph on request.
Tip
When a news headline says an AI "thinks" or "wants" something, that's almost always a shorthand for "produces outputs that resemble thinking" — worth reading skeptically rather than literally.