The course that sits at the heart of Train the Machine did not arrive fully formed. It was shaped by a conviction about how people actually learn and how to make the conviction work at scale.
In an earlier post, we wrote about what drove us to build this in the first place, a twelve-year-old, a snapping laptop, and the discomfort of watching complete trust placed in a tool nobody understood.
In the next post, we talked about the decisions that turned that mission into a platform. Why we built our own, why hands-on learning demanded it, and why the teacher had to stay in the loop.
This post is about the course itself. Specifically, Getting to Know AI — The Building Blocks. An introductory course, built for Grade 7 and above, though we have since heard from educators and practitioners that it resonates just as well with non-technical adult learners and professionals.
The question this post answers is a simple one: why is it built the way it is?
• • •
There is a difference between knowing something and understanding it. Knowing the name of an algorithm is useful. Knowing why it behaves the way it does and where it might go wrong is what stays with you.
What we care about is not just what you remember after the course ends. It is what you are left with after the details fade. A habit. A question you now automatically ask that you did not ask before. Knowing a fact is useful for an exam. Knowing how to question a fact is useful for a lifetime.
This is the story of the choices we made and why we made them.
• • •
We Had a Choice
When we sat down to build this course, the obvious thing was to start with definitions. Here is what AI is. Here is what machine learning means. Here is a diagram. We decided not to do that.
Every chapter begins with something you already have an opinion about, even if you do not know it yet. You are asked to try before you are taught. You write rules. You train models. You watch things go wrong. And then ,only after you have felt the problem, the concept arrives with a name.
That sequence matters. When you have already struggled with something, the explanation means something. It is not abstract. It is the answer to a question you were actually asking.
• • •
You Meet Maya First. There Is a Reason for That
From the very beginning, you meet Maya — a baby learning to make sense of the world. She does not have rules handed to her. She observes, tries, gets feedback, and adjusts. She groups familiar faces without anyone teaching her the word ‘clustering.’ She learns that smiling gets a smile back, without anyone explaining reinforcement.
We introduced her early because abstract concepts do not stick on their own. But something you have already understood does. So when you encounter clustering later in the course, you will not be meeting a new idea. You will be formalising something you already felt.
That moment of connection, ‘Oh, this is just like what Maya did’, is not an accident. Each connection adds a layer, until the idea is fully formed.
• • •
A Different Kind of Progression
Most technical courses on AI and machine learning follow a progression built on mathematics. Start with mean and median, move to regression, then clustering. That progression has its merits.
We chose a different path.
When we say ‘intelligence,’ what do we mean? Recognising - not just seeing. Comprehension - not just reading. Understanding the nuances of language - not just generating words. These are the questions the course is built around. The learning progression follows them. Mathematical ideas are introduced along the way, but as tools that serve the concept, not as the starting point.
Start with something concrete: a model that predicts taxi fares. Give it examples, and it finds the pattern. Then make it smarter, add time of day, weather, vehicle type and watch the accuracy shift. From numbers, move to images, then to sound, then to audio and video and then to all together. Each chapter builds on the one before it.
This is a journey, not a collection of lessons.
• • •
We Built the Failure In. Deliberately.
In practice, this shaped three deliberate choices. Start with experience, not definitions. Let intuition arrive before the vocabulary does. And make sure the concept only gets its name after you have already felt the problem it solves.
In one chapter, you label images. Correctly at first. Then you label them wrong — not because you are careless, but because a genuine mix-up is possible. A palm of a hand. A palm tree. An easy confusion to make. You watch the model train on that data and produce confident, wrong predictions. That is not just a lesson about bias. That is you causing an un-intended bias, and watching what it does.
In a later chapter, an AI agent tries to cross a river. Untrained, it fails. With limited training, it finds a clumsy path. With full training, it solves it cleanly. You are not reading about a learning curve, you are watching one, and you control how far it goes.
The moments of surprise are engineered in. Because surprise is where real questions get asked — and asking is where understanding is built.
• • •
And Then It Works on You
At a certain point in the course, something shifts. You draw your own handwritten digits and the model recognises them. You speak into a microphone and the model identifies your voice.
The AI is not working on someone else’s test data. It is working on you. And getting it right.
That is not a demonstration. That is the concept becoming real.
To get to that moment, three deliberate strategies run through the course. The first is visual, step-by-step learning — you see the model adjusting in real time, not just a before and after, but the process itself. The second is relatable anchors: cricket player categories make clustering intuitive before you have ever heard the word; a find-the-odd-one-out game makes anomaly detection feel familiar before it is formalised. The third is personal, the model responding to you, specifically. That is what makes the idea real.
• • •
The Bias and Fairness Chapter and Why It Is Not a Footnote
The course includes a full chapter on bias and fairness. It sits alongside the technical chapters with the same weight, because we believe it deserves that weight.
In that chapter, you train a model on data where colour incorrectly predicts emotional expression. You watch it make confident, wrong predictions. And then you are asked one question: what is one thing a machine learning model needs a human for?
Evaluating the quality of what a model learned, recognising its blind spots, asking whether the training data was fair or representative requires a human. Someone who understands enough about how AI works to know what questions to bring to it.
That is what we are teaching you to become: the human who asks the right questions.
• • •
What You Leave With
By the end of the course, you will have trained and interacted with more than ten real learning machines — no code, no equations, but the same underlying ideas that power every AI system in use today.
More importantly, you will have developed a way of approaching things. The ability to look at an AI system and ask where it came from, what it was trained on, and where it might be wrong. This is not a technical skill. It is a critical one — and it is the one that will stay with you regardless of how the technology changes.
That is what the course is for. Everything else is practice.
The curiosity, the critical thinking, the capacity to learn independently — every learner carries these. What the course does is give them shape and direction. The nudge is what we want to give through this course. The path will always be yours.
• • •
The journey starts here.
For how to get the most of this course and a lesson-by-lesson walkthrough of how the course is structured — the videos, the activities, the pauses, and the take-home readings — stay tuned for our next blog