Introduction

The origin story of this tutorial is that I wanted my 11 year old son Isaac to understand how modern AI really works.

I looked at existing tutorials, but all the ones I could find were either too hand-wavey, required a level of existing expertise he didn't have, or just weren't sufficiently "fun" for him to want to complete them. Maybe something better existed, but I didn't find it. So I decided to create my own.

My goal is that this tutorial should be accessible to anyone with a middle-school level understanding of math, and yet give the reader a deep intuitive sense of not just how AI works, but why it works the way it does.

The trick I'm using to do this is a series of interactive playgrounds to illustrate each new idea. It's much easier to understand an idea if you don't just read about it, but directly play with it. I've also got a Google Colab notebook for each chapter, for more advanced readers who want to play with the real code behind each chapter.

This tutorial isn't yet finished. It currently builds up to the Transformer, then follows it the rest of the way to a working assistant: how post-training makes it helpful, how it holds a conversation, how it reasons, and how it acts as an agent. After that it steps back to look at how AI is changing society, and at how it could go badly wrong. I want to eventually cover all of the ideas behind the current cutting edge models. Subscribe to my substack or follow me on linkedin if you want to know when new chapters get released.

If you're a teacher thinking of using this with a class, see the guide for educators.

And please do email me at rob.ennals@gmail.com if you found anything confusing, find any errors, or want it to explain something better. I do want to hear from you and I'll usually reply.