I work with AI everyday. Build systems, brainstorm ideas, refine frameworks.
But if today a parent were to come up to me ask if their child should be using AI for learning maths - my default answer is no.
I'm not gatekeeping here, nor I'm rejecting AI. I share from a position where I watch what happens when thinking gets outsourced at the wrong stage of development. I'm a tutor and mentor to others, but I'm also a student when it comes to learning and using AI.
Let me bring you through this process.
What's happening out there
AI didn't arrive like a product launch, or like how Steve Jobs revealed the first iPhone or Macbook Air. It arrived like an epidemic, think Covid and SARS. Fast, everywhere, and impossible to contain once it started moving.
From the first ChatGPT to agentic AI, the evolution took less than 2 years. Before anyone could ask "should we use this?", the world had already moved to "how can we use this?" The deliberating got skipped. Everyone was already on the ground- testing, experimenting, exploring.
This made me think about JARVIS analogy from Marvel. Tony Stark built JARVIS as a tool - brilliant , capable and operating in service of human intention. But the same architecture also produced Vision and also Ultron. Not because the technology was wrong, but because it depended entire on whole held it, and why. The same goes for AI. The tool is the same. What it produces depends on who is using it, and for what purpose.
Here's what I know for certain about the global picture: I don't have the authority to change how the world handles this. Neither do you. The trajectory is set by forces we can't control - policy, capital, geopolitics, infrastructure. My students also raised environmental concerns with me - the water consumption, the energy load, etc. I feel the same helplessness they do. But one's conviction, no matter how strong, cannot redirect a worldwide current at this speed.
The only thing I can do is zoom in. To Singapore. To education. To Maths. To your child.
Singapore's position
Singapore is doing what Singapore does. Watchful. Cautious. Moving deliberately — which in this case is not a weakness.
Under the National AI Research and Development Plan, the Singapore government has committed over S$1 billion over five years from 2025 to 2030, as part of its National AI Strategy 2.0, with the goal of harnessing AI for the public good. On the workforce side, the government has set a target to triple Singapore's AI practitioner pool to 15,000 over five years, with over $20 million invested to enhance AI practitioner training for students. SkillsFuture has become the delivery mechanism for this ambition — AI courses are now heavily subsidised and accessible, with many Singaporeans able to complete programmes with little to no out-of-pocket cost through SkillsFuture Credits and SSG subsidies. These are signs of a government that does not want to be reckless with its most important asset: human capital.
"AI literacy" has become a campaign punchline. But what it actually means in practice — what it looks like in a classroom, what it requires of a student — remains genuinely undefined. The private sector is moving faster. Geniebook, Koobits, and others have integrated AI into their platforms. But these are enterprise decisions, not national education policy. They are not the same thing.
The education problem
MOE's approach to AI in education emphasises responsible and age-appropriate use, framing it as students learning about AI, learning to use AI, learning with AI, and learning beyond AI. The infrastructure is being built carefully. Tools delivered through the Singapore Student Learning Space (SLS) include an Adaptive Learning System (ALS), which creates personalised learning paths for students, currently supporting Primary 5 to Secondary 2 Mathematics, with more subjects and levels to be added over time. MOE began its AI initiative with the Pri 5 Maths Adaptive Learning System, piloted in 33 schools, personalising mathematics content based on each student's performance. AI Expert Network MOE's own lead specialist has stated that the design intent is for technology to complement teacher professionalism — "the teacher continues to be the main designer of learning." That framing matters.
Even within MOE's own tools, the human remains the primary agent. The AI supports. It does not replace. On the broader exposure front, from 2025, AI for Fun modules are available to all students in primary and secondary government and government-aided schools, covering generative AI, smart robots, and understanding how to use AI safely and responsibly.
Here is the specific tension every Singapore parent is navigating. AI tools, used well, can accelerate learning. Used poorly, they accelerate the appearance of learning while quietly removing the actual process. For most subjects, the line between these two outcomes is visible enough to manage.
For maths, the line is almost invisible — and the cost of crossing it is high. In Maths, there are definite answers, to show correctness. It's either correct or it isn't. It's easy for AI to appear right, getting correct numerical answers, but with the wrong methods. And because Maths is taught universally, there's no lack of training data to draw from. The abundance of resources is way more than other learning subjects. The confidence is high. But it's not the same as correctness in approach, and in Singapore Maths, the approach is the whole point.
My take on Singapore Maths
Singapore Maths is not universal. It's a specific pedagogy built over the years, but it is not well-represented in AI training data. It cannot be reliably reproduced no matter how clever we prompt it. More importantly: the cognitive load that we want our students to train on and build is exactly the highlight of the pedagogy itself - to frame a question, draw models, etc. that is where the real learning happens. That is when we train the systems thinking in our students.When they open up ChatGPT and offload the question to it, they are not just getting the solution - the whole training ground is removed. Solutions and numerical answers can be read off the screen. Students just cling themselves to it, or worst, take the final answers and leave. That's going backwards. My default position, keep the usage of AI as low as possible for Singapore Maths. It induces cognitive offloading at the stage where cognitive load is the entire mechanism for their growth at this stage.
With that said - there are exceptions. But it depends on where your child currently stands, and how involved you are.
How to use AI - my 4D Framework
Most conversations about AI in education jump straight into "which tools" and "how much to use?" That's the wrong starting point. Before any of that, one must always start with self evaluation. Here I call this the 4D framework.
1. Determine - assess the readiness
Before any form of AI can come in, ask yourself, "What's my child current foundation? Where are the gaps? How involved am I in their learning?"
This should be the proper entry point. Not "what AI tool to use". A student with weak foundations and to use AI to consolidate knowledge is merely escapism from the true work that needs to be put in - the learning process itself. The gap still stays.
Determine readiness before anything else.
I've put together a practical guide that helps you work through this exact step, broken down by your child's current performance level, the intentions, and your involvement as a parent. It gives you a concrete starting point before you make other decisions.
It will be sent via mail to the Learning Resources Mailing List. Subscribe here and select Learning Resources, before 21st May 2026.
2. Discern - observe what is actually happening
Once the AI is introduced, observe the process. Is your child now thinking harder because AI is present - being prompted to reason, explain or to justify? Or otherwise - receiving answers, reading explanations, and skipping the cognitive work? What's dangerous is that, they look exactly the same from the outside. The real difference is whether the student is truly learning or hiding under the shadows. Discernment requires honest observation. Not what you hope is happening. What is actually happening.
3. Distinguish - interrogate the output
AI always shows confidence, but it's not always correct.Even the most advanced models are not able to do maths correctly across the world. Not to even mention Singapore Maths. Generating the correct answers is one thing - but what's more important beneath is the alignment of the presentation of solution ought to match frameworks and approaches taught in school, which AI cannot reproduce even up to this stage of development.
4. Decide - make the call
It's time to make the call. Should AI be used at all, and if so, how much.What matter is that the decision is made with clear intent, not just because everyone else is using it.If the intention is to genuinely build thinking, then proceed. Otherwise, no percentage in the guide will protect against it.
These questions don't guarantee the right outcome. But it does set the starting point right.
It's no longer the game - it's now on the player.
AI is here to stay, whether we like it or not. What's more important moving forward, is for us to decide how we use these tools to our best advantage instead being consumed and overwhelmed by it.
MathSifu Moses
