The Unstated Population: When the robot clocks in

On 18 February, Prime Minister Lawrence Wong stood in Parliament and told Singapore not to be afraid.

Artificial Intelligence (AI) is advancing at "remarkable speed," he said.

The potential is "immense." And Singapore's response would be decisive: a National AI Council, chaired by the Prime Minister himself.

Four national AI Missions. A Champions of AI programme. A new AI park at One-North. Six months of free access to premium AI tools for workers who complete the right courses. An entire section of the Budget dedicated to harnessing artificial intelligence as a "strategic advantage."

It was, by any measure, an impressive list.

It was also, by any honest reckoning, a list assembled for a problem that has already outgrown it.

There is a comforting fiction embedded in every government AI briefing, every SkillsFuture pamphlet, every ministerial assurance about the jobs of the future. 

The fiction goes like this: AI lives in the computer. It answers your emails, drafts your reports, summarises your meetings. The human still shows up. The human still moves, lifts, builds, serves, repairs. The human body — warm, dexterous, present — remains in the equation.

That fiction has an expiry date. And it is closer than any Budget speech is prepared to admit.

The Robot in the Room

In early 2026, videos began circulating out of China that should have stopped every labour economist in their tracks. Humanoid robots — not the clunky, scripted automatons of factory floors past — were performing wushu. 

Martial arts. Fluid, coordinated, physically demanding sequences that require balance, spatial awareness, and real-time motor adjustment. They were jumping. Tumbling. Moving through space with a fluency that, even two years ago, would have been considered a decade away.

Now pair that body with the mind.

GPT-4 passes bar exams. Claude reasons through complex multi-step problems. Gemini navigates ambiguous instructions and adapts on the fly. These are not narrow task-specific systems anymore — they are general-purpose reasoning engines that improve with every iteration.

The question that nobody in Singapore's policy apparatus appears to be asking out loud is: what happens when you put one inside the other?

A humanoid robot with frontier AI reasoning, able to see, navigate, communicate, problem-solve, and physically execute — operating 24 hours a day, seven days a week, at a fraction of the cost of a human worker. No CPF contributions. No medical leave. No maternity or paternity benefits. No annual wage supplement. No union to negotiate with. No resignation letter. No bad days.

We are not talking about five decades from now. We are talking about five years. Possibly less.

Run the Numbers

Consider what Singapore's labour market actually looks like at its base. Construction workers, cleaning staff, food and beverage service workers, warehouse operators, delivery riders, security guards, eldercare aides — these roles account for a substantial portion of the resident and foreign workforce.

The government has spent years trying to raise wages in these sectors through the Progressive Wage Model. Noble in intent. Structurally irrelevant if the employer's next option is a robot that costs a one-time capital outlay and a monthly electricity bill.

Lawrence Wong told Parliament that Singapore would help workers "use AI to take over routine tasks, so that they can focus their time and energy on higher-value activities — work that requires judgment, creativity, and human insight."

At the white-collar level, that argument has some runway left. The accountant who automates data consolidation and moves into advisory work — fine. The lawyer who uses AI to handle discovery and focuses on courtroom strategy — plausible, for now.

But what is the equivalent argument for the construction worker? The hospital porter? The elderly care assistant who bathes and feeds residents in a nursing home? What is the "higher-value activity" they pivot to when a robot with AI can perform their physical role more consistently, more cheaply, and without the legal obligations that come with a human employment contract?

The honest answer — the answer no politician will give you in a Budget speech — is that there may not be one. Not for everyone. Not within any retraining timeline that is politically survivable.

The Cost Structure That Cannot Compete

Let us be precise about what employers are actually weighing when they make hiring decisions.

A full-time Singaporean worker at the lower end of the wage spectrum costs somewhere in the range of S$1,800 to S$2,500 a month in base salary.

Add CPF employer contributions at 17 percent. Add the mandatory public holidays, annual leave, sick leave, childcare leave, hospitalisation entitlements. Add the administrative burden of managing a human being — performance reviews, grievances, turnover, retraining. Add the fact that humans work roughly eight hours a day, five days a week, and require rest.

Now quote me the operating cost of a humanoid robot that works around the clock, requires only scheduled maintenance, and improves its own performance through software updates pushed remotely. A robot whose "salary" is a capital depreciation line item and whose "benefits" are a maintenance contract.

The economics are not close. They will not become close. And the moment that robot crosses the threshold of being physically capable enough to do the job reliably — not perfectly, just reliably enough — the calculation for any rational business operator is not complicated.

Singapore's tripartite model, the NTUC partnership, the Progressive Wage framework — these are all instruments designed to mediate between human workers and human employers. They have no jurisdiction over the decision to automate entirely. And no amount of co-funding support or skills credits changes the underlying cost curve.

We Trained Them for the Wrong Exam

Here is what makes this especially cruel. Singapore did not sleepwalk into this vulnerability. It was engineered — carefully, rationally, over fifty years — by an education system that produced exactly what the economy required.

The ten-year series is the perfect symbol of what was built. The highest form of exam preparation in Singapore is pattern recognition of previously correct answers. Students do not learn to think through genuinely novel problems. They learn to recognise what thinking looked like in past instances and reproduce it under time pressure. Convergence on the model answer is not just rewarded — it is the entire point.

What the system produces: compliance, execution speed, risk aversion, and precision within defined parameters. These are not trivial skills. For five decades, they were precisely what employers — local and multinational — needed. The system delivered them at scale, and Singapore prospered accordingly.

The problem is not that the system failed. The problem is that it succeeded so completely, for so long, that dismantling it now feels like dismantling the country itself.

And yet — apply the robot test. A humanoid robot loaded with frontier AI, trained on every ten-year series paper since 1975, will outscore every Singaporean O-Level candidate.

It will complete the assessment in a fraction of the time, with perfect consistency, zero exam anxiety, and no need for a SkillsFuture top-up. The skills the system spent fifty years perfecting are precisely the skills that are now most efficiently automated.

What cannot be automated — at least not yet, and not in the same way — is something the system never prioritised building: the willingness to be wrong publicly, the comfort in proposing ideas that might be stupid, the capacity to sit with a question that has no model answer, the cognitive disposition to diverge from established frameworks as a starting point rather than a failure mode.

These are not just different skills. They are opposite orientations. And the gap between what Singapore's education produces and what genuine innovation requires is not a minor curriculum tweak. It is structural.

The contrast with Nordic education models is instructive — not in the romanticised way it is usually invoked, but for a specific, concrete reason.

The Finnish system does not produce better students by teaching more content. It builds different cognitive dispositions by deliberately removing the pressure to converge on correct answers. The child who spends an afternoon arguing with classmates about why their bridge design collapsed is not falling behind. They are building the architecture of a mind that can generate questions no algorithm has been given — because the wandering itself is the point.

Singapore's Institutes of Higher Learning (IHLs), to their credit, are attempting to adapt. Budget 2026 gestures at this: students should "learn to use AI wisely, not as a shortcut, and are equipped with rigorous thinking and deep disciplinary skills."

That is a sentence. A worthy sentence. But it is not a strategy, and it does not reckon with the fact that the national examination system — which still determines university admission, career trajectory, and social standing — runs entirely on the logic it is asking universities to undo.

You cannot build innovators at polytechnic while you are still building pattern-matchers at primary school. The system is not confused. It is coherent. That is the problem.

The MNCs will confirm this if you ask them directly. Foreign employers operating in Singapore consistently describe local workers as excellent executors and, in the same breath, weaker innovators.

This is not a cultural critique. It is the rational outcome of a rational system — intelligent people adapting sensibly to the incentive structure they were raised in. Follow the rules. Get the grades. Take the stable job. Upskill when told to.

The Singaporean who did all of this did everything right by the system's own logic. That person is now the most exposed when the system's logic stops working.

The Political Economy of Who Gets Protected

Here is where the analysis takes a turn that should make every Singaporean voter deeply uncomfortable.

Think carefully about which jobs are structurally protected from robotic displacement — not because they are intellectually demanding, but because they require something that physical AI systems cannot yet replicate: political legitimacy, democratic accountability, or the specific human theatre of power.

A robot cannot stand for election. It cannot deliver a National Day Rally speech that moves a population. It cannot sit in Parliament and be held responsible by constituents. It cannot chair a National AI Council and project the authority of a government mandate.

Politicians — and by extension the senior civil servants, statutory board chiefs, and ministerial office holders that orbit them — occupy roles that are protected not by their cognitive complexity, but by the social and institutional architecture that requires a human face on power.

The construction worker who built the MRT tunnel has no such protection. The cleaner who maintains the hawker centre has no such protection. The eldercare aide who wipes down residents at 3am in a nursing home — none.

Let that asymmetry sit for a moment. The people making decisions about how Singapore navigates the age of robotics are among the few whose jobs the robots structurally cannot take. Everyone else is working out their exposure on their own time, with their own anxiety.

This is not a conspiracy. It is not malice. It is something more insidious: a structural blind spot baked into who holds decision-making authority. The people designing the safety net are the people who will never need to use it. That does not produce bad intentions. It produces catastrophically inadequate imagination about how bad it can get.

And here the education critique returns with full force. The same system that trained citizens to defer to the model answer, to avoid divergent thinking, to trust that the authoritative framework will provide the correct path forward — that system also trained citizens not to loudly challenge the people holding the framework. The population least equipped to interrogate the adequacy of the government's AI response is the one the government's education system produced.

That is not an accident. It is fifty years of compounding.

You Were Warned. Now Warn Someone Else.

Wong is not wrong that fear cannot be Singapore's response. But false reassurance is not courage — it is just fear wearing a different costume. And there is a particular cruelty in false reassurance aimed at the young, because they have the least time to recover from it.

Here is what is actually happening in Singapore's classrooms right now. The same rote learning. The same convergence on model answers. The same exam-obsessed pressure cooker that shaped their parents, and their parents' parents. Except now there is an AI literacy module bolted onto the side. Here is how to use ChatGPT responsibly. Here is how to prompt effectively. The child is now, officially, AI-ready.

They are not AI-ready. They are bus-ready.

And the bus has already left.

The young Singaporean sitting her O-Levels this year is being prepared — sincerely, with enormous institutional effort and genuine parental sacrifice — for a future that is being dismantled in real time. The tools are newer. The laptops are shinier. The SkillsFuture portal has been redesigned for easier navigation.

But the underlying cognitive logic is identical to what the system ran on in 1985: memorise, reproduce, converge, excel within defined parameters. Get the grade. Board the bus.

What nobody is telling her is that by the time she finishes chasing that bus — completes the degree, earns the certification, accumulates the credentials she was told would secure her future — she will arrive breathless at a stop that no longer exists. Not because she ran slowly. She ran the whole way. She did everything right by every measure the system gave her.

But when she arrives, the infrastructure has changed entirely. The bus bay is gone. In its place: a train on different tracks, to different stations, running on a completely different logic — one that was never part of her preparation, never rewarded in her examinations, never modelled by her teachers, never incentivised by the system her parents trusted with her future.

The skills she built chasing the bus do not transfer. The station she trained to reach does not connect to where the train goes.

That is not a skills gap. That is a civilisational mismatch. And it was entirely preventable.

The honest version of the AI Budget speech — the one that was never delivered — would have said something like this: we are buying you time, not building you a future.

The reskilling programmes are real, the intentions are genuine, but we are managing a transition we do not fully understand, on a timeline we cannot control, and some of you will arrive at the destination to find the destination has moved. We do not yet know what we owe those people. We are working on it.

That would have been political courage. What Singapore got instead was SkillsFuture credits, a Champions of AI programme, and a National AI Council chaired by the one category of worker the robots will never replace — because you cannot automate the human face of political authority.

The people designing the safety net will never need to use it. Let that inform how much faith you place in its design.

This piece is not written for those people. It is written for the 22-year-old fresh graduate who has been told he is entering an AI-enabled workforce and believes it, because why would the system lie to him.

It is written for the parent who is pushing her child through tuition and ten-year series papers and supplementary assessment books, believing with everything she has that she is giving her child an edge — not realising she is sharpening a tool for a job that is already being automated.

It is written for the mid-career professional quietly relieved that his company just adopted an AI platform, because at least that means the jobs are safe for now — not understanding that the platform is the proof of concept, and the robot body is the next procurement cycle.

The train does not wait. It did not wait for your parents. It will not wait for your children. And unlike the bus — which at least ran on a schedule you could plan around — the train departed while everyone was still being assured that the bus was coming.

You are not waiting for the future to arrive. You are standing in it, being handed a bus schedule.

Look up from the timetable. The tracks have already changed.

---

This is the first in a series on AI, automation, and Singapore's economic future. The second piece examines the demographic pipeline and what happens to citizen natural increase when the labour market that justified it no longer exists. The third examines the infrastructure being built for a population Singapore's own government has not publicly described. The fourth asks the question the series has been building toward: what population is Singapore actually planning for, and who bears the cost when the plan runs out of road.

This series is the foundation of a longer work. If you think these arguments deserve a full book — with the data, the charts, the parliamentary record, and the regional comparisons this format couldn't hold — support the project at https://buymeacoffee.com/theonlinecitizen. Every contribution goes toward making that book happen.