[Perspective] Where should basic mathematics education go in the AI age?

Following the solving of a Millennium Prize Problem, OpenAI released a batch of mathematical research generated by its internal artificial intelligence (AI) models on 6 October this year. Initially, it published 722 mathematical manuscripts on GitHub, grouped into 372 sets of research findings. But it withdrew three manuscripts the very next day, on 7 October, leaving 719 manuscripts in 372 sets. These findings still require further scrutiny. Yet many cannot help asking: what is happening in mathematics? When answers become cheap, which human abilities are truly valuable? Can AI be trusted completely, or must we use it with caution?

Google DeepMind co-founder Demis Hassabis believes artificial general intelligence (AGI) could arrive around 2030, and that AI will become “the ultimate tool for science”, accelerating breakthroughs in fundamental science. OpenAI chief executive Sam Altman also believes AI is no longer merely a prospect for the future, but is entering a phase of accelerating development. Many media outlets have already begun describing this era as an “AI tsunami”.

For ordinary people, the biggest impact is not that AI will suddenly replace everyone, but that their knowledge, skills and work experience could rapidly lose value. Should we first teach children to calculate by hand or mentally before giving them calculators? How should basic mathematics education prepare for this change? Teaching tools have evolved from chalkboards, whiteboards, overhead projectors and transparencies to computer software such as PowerPoint, digital visualisers, interactive whiteboards and now tablets. Should AI now become a tool for everyday teaching and work?

In 2025, the Organisation for Economic Co-operation and Development (OECD) held its triennial Programme for International Student Assessment (PISA), testing the reading, science and mathematics abilities of more than 760,000 15-year-olds from 91 countries and economies. Singapore randomly selected 6,600 students to take part. PISA results released in September this year showed that Singapore students ranked second in mathematics. This was a hard-earned achievement, but there is still room for improvement. We should neither become complacent nor lose our bearings in basic mathematics education amid the AI wave.

Let me use three exercises to illustrate how mathematics can be taught.

Example one: “If five T-shirts take five minutes to dry in the sun, how long will 30 T-shirts take?”

Assume there is enough space to hang all the clothes out at once, under the same sunlight and weather conditions. If AI approaches the problem using only primary-school arithmetic, it might calculate:

30 ÷ 5 = 6

6 × 5 = 30 minutes.

What is wrong with this answer? Teachers could ask students to consider whether increasing the number of clothes necessarily increases the drying time.

Example two: A goes to a restaurant for a meal. When he asks for the bill, the waiter tells him that the restaurant is offering a 20% discount, then presents the bill. Suppose A has spent $100. He quickly works it out in his head: subtracting 20%, then adding 9% GST, means an 11% reduction. So, $100 × 89% = $89.

But this is wrong. The amount on the bill is $87.20.

Teachers could divide students into discussion groups and ask them to:

a. Write out different calculations to find the correct answer.

b. Assume the restaurant has made no mistake. Where did A go wrong?

c. Identify the underlying issue.

d. If a 10% service charge is also added, what results would A's method and the restaurant's method produce? What is the difference between them?

e. Give another example from everyday life and use critical thinking to test their judgement.

Now let us look at the final example.

A robot starts at position 0 on a number line. Each second, it moves forward by a number of steps equal to the number of that second: one step in the first second, two steps in the second, three steps in the third, and so on.

a. Where is the robot after six seconds?

b. Find a rule for the robot's position after n seconds. Describe it in words or express it as a mathematical formula.

c. Use your rule to predict the robot's position after 20 seconds, without adding up its movements second by second.

d. Give another similar example to test whether your rule holds.

Problems of this kind, involving patterns, sequences and early algebraic thinking, do not often appear in standard textbook exercises. To solve them, students need to draw on different mathematical concepts, logical reasoning, critical thinking, questioning and verification, and the ability to express themselves in mathematical language. Children may not immediately spot where an answer has gone wrong. How, then, should teachers design questions so that children know to ask “why”, even when an answer appears correct? This is an ability we urgently need to cultivate in an age of rapid AI development.

Could we also think more boldly, encouraging teachers to devise mathematical problems, share them with colleagues and discuss solutions and answers together, building a system of basic mathematics education that prepares the next generation for the AI age?

As AI becomes increasingly adept at providing answers, how should we rethink the value of learning? In the past, we measured students' mathematical ability by the speed and accuracy with which they solved problems. But in the AI age, these abilities are no longer uniquely human strengths. The ability to identify problems, ask worthwhile questions and explore the unknown may become more important.

This does not mean, however, that we can neglect foundational knowledge and practice in problem-solving. Asking good questions often requires a sufficiently deep understanding of a field. Judging whether AI's answers are reliable likewise requires a solid grounding in knowledge and the ability to think independently. Without going through the process of learning, working things out and making mistakes — building the confidence to cope with failure — how can we develop genuine judgement and creativity?

Education needs to rethink not only what we teach, but how we teach, and whether we should preserve those seemingly “useless” processes of exploration that exercise the mind. AI can help humanity reach known answers faster, but what is education's more important mission? Should it not be to nurture children who can think independently, who are sincere, kind and curious, and who dare to dream and venture into the unknown?

The writer is a retired secondary school teacher and an author of secondary school mathematics textbooks and primary school Mathematical Olympiad books.