The Three Flavours: ML, Deep Learning and Generative AI

Untangling the buzzwords

When people say "AI", they usually mean one of three different (but related) things. These terms get thrown around interchangeably, which causes endless confusion. Let's sort them out clearly, because understanding how they nest together is one of the most clarifying things in all of AI.

The simplest way to picture it: they're nested circles. AI is the biggest circle (any machine doing "smart" things). Inside it sits Machine Learning. Inside that sits Deep Learning. And Generative AI is a powerful application that mostly uses deep learning. Each is a more specific subset of the one before.

┌─────────────────────────────────────────────┐
│  Artificial Intelligence (the broadest idea) │
│   ┌─────────────────────────────────────┐    │
│   │  Machine Learning (learns from data)│    │
│   │    ┌───────────────────────────┐    │    │
│   │    │  Deep Learning            │    │    │
│   │    │  (neural networks)        │    │    │
│   │    │   ↳ powers most           │    │    │
│   │    │     Generative AI         │    │    │
│   │    └───────────────────────────┘    │    │
│   └─────────────────────────────────────┘    │
└─────────────────────────────────────────────┘

Machine Learning: the foundation

Machine Learning (ML) is the broadest and most important term. It means any system that learns from data instead of following hand-written rules. You give it examples, and it figures out the rule itself. This is the core idea that made modern AI possible.

The classic example is a spam filter. You don't tell it "block any email containing 'win a million dollars'" — you'd never catch every spam trick that way. Instead, you feed it thousands of emails labelled "spam" or "not spam", and it learns the patterns of spam on its own, including subtle ones you'd never think to code. That's machine learning: learning the rule from examples. Most practical AI you encounter — recommendations, predictions, fraud detection — is machine learning.

Deep Learning: ML inspired by the brain

Deep Learning is a powerful kind of machine learning that uses something called neural networks — systems loosely inspired by how neurons connect in the human brain. They're built from layers of simple artificial "neurons" that pass signals to each other, with each layer learning to recognise increasingly complex patterns.

Deep learning shines when the data is messy and complex — images, sound, and language, where writing explicit rules is hopeless. The "deep" refers to having many layers. If you've ever used Google Translate, unlocked your phone with your face, or talked to a voice assistant, you've used deep learning. It's the breakthrough that, around the 2010s, took AI from "interesting" to "astonishing", because neural networks turned out to be remarkably good at learning from raw, complicated data.

Generative AI: the newest, noisiest flavour

Generative AI is the flavour everyone's been talking about since ChatGPT arrived. The key word is generative: instead of just classifying or predicting, it creates new things — original text, images, code, music, video. ChatGPT and Claude write essays and answer questions; tools like DALL-E and Midjourney paint pictures from a description; AI coding assistants write software.

These models are enormous — sometimes hundreds of billions of internal values — and they're mostly built using deep learning. The core trick behind generative text AI is surprisingly simple, as we'll explore: they learned from massive amounts of text to predict what comes next, and by doing that incredibly well, they can produce fluent, useful, original writing. Generative AI is genuinely new and powerful, which is why it's reshaping how people work and create.

How to tell which is which

Now you have a practical superpower: when someone says they're "using AI", you can ask which flavour and understand the answer:

• Predicting a number or category from data (will this customer churn? is this spam?) → Machine Learning.
• Recognising images, speech, or understanding language deeply → usually Deep Learning.
• Creating new text, images, or code from a prompt → Generative AI.

They're not competing — they're layers of the same story. Machine learning is the foundation, deep learning is its most powerful technique, and generative AI is its most exciting recent application. Now that the vocabulary is clear, let's look under the hood at how machines actually learn from data — the mechanism behind all three.