Grow Faster Than Change: Seven Lessons from Professor Leong

When I first arrived at NUS, my supervisor, Professor Leong, had a conversation with us that lasted nearly fifty minutes. Most of it was advice—his views, cautions, and observations. He had already spent thirty years working in AI, from the days when university administrators would ask neural-network researchers, “What exactly are you doing to the computers?” all the way to the present.

I had only just arrived, and I could understand only fragments of Singlish. Fortunately, I had a habit of recording classes. While organizing my files today, I listened to the recording again. With two more years of experience and a different understanding of my own situation, I finally grasped many things I had missed at the time.

Below are the seven points I extracted in my own words. I am writing them down as a record. To me, the seven things he discussed were really seven sides of the same idea.

1. Do not look only at functions; look at non-functional requirements

Anyone can write a program or add a feature. The real work is making something capable of continuing to operate in the real world.

He used Singapore’s MRT as an example. The early system was designed around a projected population of four million in 2020, but the actual population exceeded five million. By then, the station platforms had already been poured into underground and elevated structures. They could not simply be lengthened, which also meant the trains could not become longer. The original function had been delivered—the trains ran and passengers could board—but the system had been built to be just sufficient, with no room left for the future.

The true cost of a system often lies not in what it can do, but in what can no longer be changed later. What is just enough today often becomes tomorrow’s trap.

2. Think about four layers: technology → user → community → society

Electric vehicles were booming in China at the time, so he used cars as an example.

At the technology layer: can the car drive? At the user layer: is it easy for the driver to use? Some electric cars look beautiful but require passengers to twist something like a lock to open the door from inside; an older person without enough strength may not be able to turn it. At the community layer: there are other cars, pedestrians, and cyclists on the road; your vehicle never exists in isolation. At the societal layer: if everyone has an autonomous car, daily travel may double, making carbon emissions and congestion worse rather than better.

A product can function well, feel convenient to its user, and work acceptably in its community, yet still produce a bad outcome at the societal level. The more mature a person becomes, the wider the radius of the questions he can consider—from “Can I build this?” to “If I build it, will the world become better or worse?” Technical people are most likely to stop at the first layer, while real judgment emerges at the higher ones.

3. Practice the fundamentals until they become instinct, so your mind is free for what matters

He talked about playing football as a student. A friend made him dribble the ball all the way to the pitch and then play even when he was exhausted. During a match, your mind cannot focus only on how to kick the ball. You also need to know where your teammates, opponents, and the goal are. Athletes have no time to calculate angles and spin when the ball arrives; they rely on instinct, recognition, and memory.

Programming is the same. When do you become a good programmer? When you no longer have to struggle with the syntax.

He offered another view: a large language model is less a processing model than a recognition model, something like the rear part of the brain associated with memory and recognition rather than the front involved in thinking. Looking at AI now, the more it resembles that recognizing “back brain,” the more scarce the “front brain” becomes: strategy, judgment, design, and deciding where to go next.

Turning fundamentals into instinct and preserving your mental capacity for judgment has not become outdated in the AI era. It has become more valuable. From today’s vantage point, Professor Leong’s understanding of AI two years ago was both deep and far-sighted.

I feel this strongly myself. The clearer you are, the better you collaborate with AI, and the more it amplifies your strengths. But the more confused you are, the more it amplifies your confusion. Our relationship with AI is multiplicative. When the direction is right, AI can magnify an effective action a hundredfold. When the direction is wrong, AI will execute the mistake a hundred times with extraordinary diligence. What mass production truly amplifies is not AI’s ability, but the human judgment made before production begins.

4. AI will not take your job, but people who use AI will replace people who do not

This applies to every industry, without exception. It is not reassurance; it is an ultimatum. It moves the anxiety precisely from “Will I be replaced?” to “Am I learning?” The former is beyond your control. The latter is entirely within it.

In short, I believe that being unable to use AI in the future will be like being unable to use a smartphone today. Why did humans become human rather than remain wild animals? Because humans used tools. AI and smartphones are still tools—more advanced tools, but no different in essence from a hoe or a plow. If you know how to use the tool, you become a more capable human being. We are animals becoming more advanced; this is evolution.

We may think that because we cannot speak another language, compose music, or write well, we are illiterate in those fields. But AI gives us new possibilities. We can become musicians or poets, as long as we genuinely think and work with care. AI can help us do many things. This is a very good moment to begin.

5. University is a safe place to make mistakes, and people learn by making them

“If you make a mistake here, nobody will die because of it.”

If you are afraid of falling and only walk down the middle of the path, you will learn nothing. He said we should not watch the Olympics only for the top three. Someone may finish last, far behind everyone else, and we still respect that person because they actually stepped onto the field and tried.

Learning is the same. You came here to learn, not merely to score. If you use AI to skip the learning process and let it submit your assignments, your parents may have spent a great deal of money while you learned nothing.

When AI can produce the correct answer effortlessly, the right to make mistakes becomes a luxury. The cheaper standard answers become, the more valuable the experience of trying and falling for yourself becomes. AI will not go to prison for you.

6. Do not plan life so completely that you forget to enjoy it

He had met students who mapped out the next ten years immediately after graduation: begin in sales and marketing to travel while young, return to government to gain experience, and finally enter banking because its income would be suitable for starting a family. Vision matters, but excessive planning welds shut the space for exploration before life has begun.

His generation was not as anxious. Nobody started asking in primary school what you planned to become. People today are pushed by a world changing too quickly, and that pressure makes them more afraid.

Yet we often forget: the average person buys a home at 36.9, not 24. Building an ideal body takes three years, not three days of sudden effort. The average millionaire is 55, not 24. Most people change careers at least three times rather than stay in one job for life. The average age of reaching financial freedom is 51, not 24. Most people discover the field they truly enjoy after 30, not at university. The average successful entrepreneur is 42, not in their early twenties. Meaningful achievement in a career requires at least five to seven years of focused work, not six months.

We are not behind. We have simply been swept up by unrealistic timelines. We should not let the anxiety of instant success control us. What we need is self-improvement and a gentler reconciliation with ourselves. Slow is fast, and our lives deserve the kindness of time.

7. Embrace change and enjoy it

Someone asked whether he was afraid. He said no. He welcomed these changes because every change was an opportunity to learn.

Many of his birthday cards and Lunar New Year images were made with AI. For example, when he asked a Western AI to draw a Chinese dragon, it initially kept producing a Western one. He had to give repeated instructions about how a Chinese dragon should move and which features it should not have. He said it was fun—something worth playing with, learning, and trying.

To me, this is the master switch for the previous six points. Faced with the same uncontrollable world, some people treat change as a threat, so their plans grow tighter and their defenses narrower. Others treat change as a game, so every attempt opens more possibilities and every mistake helps them learn faster. The difference is not ability but mindset.

Put the seven ideas together, and he was really saying one thing from beginning to end: do not let tools alienate you, and do not let anxiety tame you. Keep the judgment beyond functionality for yourself. Practice the fundamentals until they become instinct. Treat mistakes as a right rather than a risk. Then, however the world changes, you can grow faster than change.

“Don't try to be as good as Esther. You want to be better than Esther.”

Esther was the lead for our project. Do not aim merely to catch up with the standard beside you. Aim to surpass it.