2. Machine Learning, Deep Learning and Generative AI

2.1 The relationship

Think of the fields as a hierarchy:

Artificial Intelligence
|
+-- Machine Learning
|   |
|   +-- Deep Learning
|
+-- Other AI approaches
|
+-- Generative AI applications

The exact taxonomy is more nuanced, but this is a useful beginner mental model.

2.2 Supervised learning

In supervised learning, training examples contain known targets.

Input + Label
     |
     v
Model training
     |
     v
Model

Example:

Email -> spam
Email -> not spam

2.3 Unsupervised learning

The system looks for structure without a supplied target label.

Examples:

• Clustering
• Dimensionality reduction
• Pattern discovery

2.4 Reinforcement learning

An agent interacts with an environment and receives feedback.

Agent
  |
  v
Action
  |
  v
Environment
  |
  v
Reward
  |
  +----> Agent

2.5 Neural networks

A neural network contains connected computational units organized into layers.

Input layer
     |
Hidden layers
     |
Output layer

Deep learning uses many such layers.

2.6 Generative models

Generative systems learn patterns that allow them to produce new outputs.

For language:

Instruction
     |
     v
Language model
     |
     v
Generated text/code

For image generation:

Prompt
  |
  v
Image model
  |
  v
Image

2.7 Why this matters for developers

You do not need to train an LLM to use AI effectively as a developer.

Modern AI engineering often focuses on:

• Model selection
• Prompt design
• Context engineering
• Tool integration
• Retrieval
• Agent design
• Evaluation
• Safety
• Observability

Exercise

Describe a software product and identify:

1. Where traditional software is sufficient.
2. Where ML could help.
3. Where Generative AI could help.
4. Where an AI agent could help.