This is called binary search, and it is incredibly efficient. It can find any word among a million entries in about 20 steps.

How Algorithms Power AI

AI algorithms are more sophisticated than sorting cards, but the principle is the same: follow structured steps to reach an answer. Here are two foundational AI algorithms.

Decision Trees

A decision tree asks a series of yes-or-no questions to classify something.

Example - Is this email spam?

1. Does the subject line contain "FREE MONEY"? → If yes, likely spam.
2. Is the sender in your contacts? → If yes, probably not spam.
3. Does it contain more than five links? → If yes, lean towards spam.

Each question is a branch, and each final answer is a leaf. Decision trees are easy to understand, which makes them popular when humans need to explain the AI's reasoning.

K-Nearest Neighbours (KNN)

KNN classifies something by looking at the closest examples it has already seen.

Imagine you move to a new neighbourhood and want to know if a house is expensive or affordable. You look at the five nearest houses (your "neighbours") and check their prices. If most are expensive, you predict yours is too.

KNN works exactly the same way with data points - it finds the K closest examples and takes a vote.

Big O Notation: Fast vs Slow Algorithms

Not all algorithms are equally fast. Computer scientists use Big O notation to describe how an algorithm's speed changes as the data grows.

| Notation | Name | Example | Speed | |----------|------|---------|-------| | O(1) | Constant | Looking up a value by index | ⚡ Instant | | O(log n) | Logarithmic | Binary search | 🚀 Very fast | | O(n) | Linear | Reading every item in a list | 🏃 Decent | | O(n²) | Quadratic | Comparing every item to every other | 🐢 Slow |

You do not need to memorise the maths. The key insight is this: as your dataset grows, a poorly chosen algorithm can go from fast to impossibly slow.

Why Choosing the Right Algorithm Matters

There is no single "best" algorithm. The right choice depends on:

• The problem type - Are you classifying images or predicting numbers?
• The data size - Some algorithms struggle with millions of rows.
• Explainability needs - Can you use a black box, or must you explain every decision?
• Speed requirements - Does the answer need to arrive in milliseconds or can it take minutes?

A decision tree might be perfect for a simple loan approval system where transparency is required. But for recognising objects in a photo, you would need a neural network - which we will explore in the next lesson.

Key Takeaways

• An algorithm is a step-by-step set of instructions for solving a problem.
• Decision trees classify by asking sequential yes/no questions.
• K-Nearest Neighbours classifies by looking at the closest known examples.
• Big O notation tells us how an algorithm's speed scales with data size.
• Choosing the right algorithm depends on the problem, data, and constraints.

Next, we will look at the most powerful family of algorithms in modern AI: neural networks.