"I know my child should be learning AI. I just don't know what that means, or where to begin."
I hear some version of this from parents almost every week, and it's a completely reasonable thing to feel. Every few days there's a new AI app, a new AI feature in something you already use, or a new course promising to teach kids "AI skills". It's hard to tell what matters and what will be forgotten by next year.
Here's the simplest way I know to cut through it.
Quick Answer: If you want your child to learn AI but don't know where to start, start with how AI thinks, not with the latest AI tools. New AI tools appear constantly, but almost all of them are built around the same core brain: a large language model. The design behind those models has stayed largely the same for years. What keeps changing is the tools wrapped around them. A child who understands how a language model works can pick up any new tool quickly, because they know where the "thinking" is happening. That's why our AI & Python course for ages 13–17 has students build their own mini language model in Python, from scratch. Younger children can start with coding first, through Roblox or game development.
AI Seems to Be Moving Incredibly Fast
It certainly looks that way. A few years ago, most people's experience of AI was typing a question into a chatbot. Today, AI can:
- Use your browser for you. Tell it what you want, and it can click through websites, fill in forms and move around your screen on its own.
- Edit your documents. AI can write and rework a Word document, build formulas and charts in an Excel spreadsheet, or put together a PowerPoint deck, quickly and with almost no effort from you beyond chatting with it.
- Write code, make images, summarise meetings, plan holidays and a long list of other things that each seem to arrive with their own new app.
No wonder parents feel they're falling behind. If your child is meant to "learn AI", which of these are they supposed to learn? And what happens when next month's tools replace this month's?
What's Actually Changing: The Tools, Not the Brain
Here's the part that most of the noise hides. The core technology underneath nearly all of these tools hasn't changed very much.
Today's chatbots and AI assistants run on large language models, or LLMs. Most of them are built on a design called the transformer, which researchers introduced in 2017. The models have become much bigger, they're trained on far more data, and they've been refined in many clever ways, so they really are more capable than they used to be. But the basic architecture, the way the model takes in words and works out what comes next, has stayed mostly the same.
What has exploded is everything built around those models. Companies and developers have realised that one very capable model can be plugged into almost anything: a browser, a spreadsheet, a slide editor, a code editor, a phone app. So they keep building new tools, and then more tools, and then tools that connect the other tools together. That's where the feeling of breakneck speed comes from.
Tools Are Wrappers. The Model Is the Brain.
A helpful way to picture it: the AI model is the brain, and the tools are the hands.
The AI that controls your browser and the AI that edits your spreadsheet look like very different products. Underneath, they're often the same kind of brain, sometimes literally the same model, wrapped in different tools. One wrapper lets it click buttons on a web page. Another lets it type into a spreadsheet cell. The wrapper decides what the AI can do. The model decides what it thinks.
That has a big implication for what kids should learn. If you teach a child a tool, they learn one set of hands, and those hands will be redesigned or replaced soon. If you teach a child how the brain works, they understand every tool built on top of it, including the ones that don't exist yet.
Why the Brain Is What Kids Need to Understand
The model is the crux of it. It's where the "thinking" actually happens, and it's the part that most AI courses for kids skip.
When a child understands what's going on inside a language model, a lot of things that seem mysterious about AI start to make sense:
- How it writes. A language model produces text by predicting what's likely to come next, one small piece at a time, based on patterns it learned from enormous amounts of text. Our post on how ChatGPT actually works explains this in more detail.
- Why it can sound confident and still be wrong. It's choosing likely words, not checking facts. Once a child really gets that, they stop treating AI as an oracle. We wrote about this in learning with AI: powerful, not perfect.
- Why the same question gives different answers. There's chance built into how it picks each word, on purpose.
- What a new tool is really doing. When the next "AI that does X" arrives, they can see it for what it is: the same kind of brain, given a new pair of hands.
A child with that understanding doesn't need a new course every time a new app comes out. They can open it, work out what it's doing, and use it well, often faster than the adults around them.
Why We Built a Course Around a Mini Language Model
This is exactly why we designed our AI & Python course the way we did. Instead of teaching students to use someone else's AI, we guide them to build their own mini language model, line by line, in Python.
Students start with no coding experience at all. They learn Python as they go, because they need it for the project. Then they put it to work on one question: given these words, what's likely to come next? Their program reads real text, a book, song lyrics or a film script they choose, counts the patterns, and starts writing sentences nobody typed.
Over 10 lessons, the engine gets smarter. Students add chance so it stops repeating itself, give it more memory so the writing gets sharper, rebuild it on tokens (the word pieces real models use), and turn words into numbers so the program can tell which meanings are related. Each step mirrors something the big models do, at a size a teenager can hold in their head and see working on their own screen.
Their mini model is far smaller and simpler than the ones behind ChatGPT, of course. But it answers the same question in the same spirit, and that's the point. Building it ingrains how the large language models out in the world work. Students leave knowing where the thinking happens, which means they can pick up and use any new AI tool, because they already understand the technology underneath it. If you'd like a closer look at what happens in class, see what our AI students actually build.
So, Where Should Your Child Start?
Our advice comes down to age:
- Under 13: start with game development. Roblox game development teaches the same thinking that AI is built on (logic, loops, variables, fixing bugs), with instant visual results that keep younger children hooked. Our guide on AI and coding at each age explains why.
- 13 and over: start with AI. AI & Python takes them from their first line of Python to a mini language model of their own. No coding experience is needed.
- Finished AI & Python: move on to AI Image & Video Generation, where students build the idea behind AI image tools.
Don't worry about your child being the youngest or oldest in the room. We group students with others around the same age, so the pace, the examples and the conversation all suit them.
Whatever their age, you can start at home this week. Sit beside your child while they use a chatbot, ask it something you already know the answer to, and when it slips up, ask them, "Why do you think it said that?" It's a small conversation, and it's the very first step towards understanding the brain rather than just using the tool. Our parent's guide to AI for kids has more ideas, and free AI coding for kids lists ways to start without spending anything.
You Don't Need to Understand AI Yourself
One last thing, because many parents worry about it: you don't need to be technical to help your child here. You don't need to know what a transformer is or how to write Python. You only need to steer them away from chasing every new tool and towards understanding what's underneath. The tools will keep changing. The brain inside them is what's worth learning.
Frequently Asked Questions
Where should my child start with learning AI?
Start with how AI works, not with specific AI tools. For teens aged 13–17, that means learning Python and building a simple language model. For younger children, it means learning to code first, through something visual like Roblox, and coming to AI later.
Is AI changing too fast for kids to keep up?
The tools change fast. The core technology changes much more slowly. Most AI tools run on large language models built on the same basic design, so a child who understands how those models work can keep up with new tools easily.
Should my child learn to use AI tools or learn how AI works?
Both are useful, but understanding comes first. Using a tool is easy to learn and goes out of date quickly. Understanding how a language model thinks lasts, and it makes your child better at using every tool.
What age can my child start learning AI?
Kids can learn what AI is and how to use it sensibly from primary school age, with an adult alongside. Building AI in Python usually suits ages 13 and up. Our AI & Python course is for ages 13–17 and needs no coding experience.
Does my child need to be good at maths to learn AI?
No. Our course builds a language model with counting, lists and probability that students learn as they go. Curiosity and patience matter more than advanced maths at this stage.
Start Learning AI in Singapore
Our classes run in small groups of up to 6, in person at our Jurong East studio or live online. Every course starts with a free 45-minute trial lesson, where your child writes their first text generator and you both see whether it's for them. Book a free trial class and help your child understand the brain behind every AI tool they'll ever use.
