🚀 Alguni Start learning

🤖 AI and machine learning for kids

From a hand-written rule that runs out, to nearest neighbours, decision trees, gradient descent, a neural network you train yourself, backpropagation, k-means, Q-learning and attention.

Start the AI track

How this track is kept honest

Nothing in this track imports a library. No numpy, no scikit-learn, no torch — because the claim it makes is that a neural network is only multiply, add and one curve, and a lesson that called fit() would make that a claim nobody could check. The maths is re-derived twice: analytic gradients are checked against numerical ones to six decimal places.

Best after the Python track. Every sample is Python short enough to read on a phone.

How it starts

You have written rules: if, else, while. The computer never decided anything — you decided, and it obeyed. Here is a rule for spotting a cat.

The full AI syllabus

13 units and 52 lessons, in the order they are played. Each unit has a page of its own with every explanation in it, the code samples and what they print, and a few questions to try.

Unit 1: Rules or Examples

What makes a program learn

  1. When Rules Run Out
  2. Everything Becomes Numbers
  3. Was I Right?
  4. Your First Learner

Unit 2: Ask the Neighbours

Guessing by what it looks like

  1. How Far Apart?
  2. Copy the Closest
  3. The Big Number Wins
  4. Honest Testing

Unit 3: Twenty Questions

A model you can read

  1. One Question Is Not Enough
  2. Which Question Is Best?
  3. Growing the Whole Tree
  4. A Tree That Learned Too Much

Unit 4: Right and Wrong

One number is never enough

  1. Two Ways To Be Wrong
  2. Precision and Recall
  3. Moving the Line
  4. Which Mistake Is Worse?

Unit 5: The Neuron

Three numbers that fix themselves

  1. A Neuron Adds Up
  2. Learning the Weights
  3. It Draws a Line
  4. The XOR Problem

Unit 6: Downhill

Getting less wrong, one step at a time

  1. How Wrong Am I?
  2. Which Way Is Downhill?
  3. Sliding Downhill
  4. How Big a Step?

Unit 7: Layers

Neurons feeding neurons

  1. A Smooth Switch
  2. A Layer in the Middle
  3. The Shape of a Network
  4. Reading a Network

Unit 8: Backpropagation

The network finds its own numbers

  1. Who Is to Blame?
  2. The Slope of the Squash
  3. It Teaches Itself XOR
  4. When Training Goes Wrong

Unit 9: Seeing

Pictures are grids of numbers

  1. A Picture Is a Grid of Numbers
  2. Sliding a Filter
  3. Nine Weights That Work Everywhere
  4. Reading a Number

Unit 10: Groups Nobody Labelled

Learning with no answers at all

  1. No Answers Anywhere
  2. Guess, Then Move
  3. Watching It Settle
  4. Where You Start Matters

Unit 11: Learning by Reward

Nobody says what to do, only what it was worth

  1. No Answers, Only Rewards
  2. What Is This Square Worth?
  3. Try Something You Have Never Tried
  4. You Get What You Pay For

Unit 12: Words

Guessing what comes next

  1. Words Become Numbers
  2. What Comes Next?
  3. Writing a Sentence
  4. Why It Makes Things Up

Unit 13: Attention

How the big models work

  1. Meaning as Numbers
  2. Paying Attention
  3. What a Transformer Is
  4. Using AI Well

Other things to learn

Start learning — units 1–2 are free