For subscribers only
Translated by ChatGPT
https://www.zaobao.com.sg/forum/views/story20260622-9245589?utm_source=android-share&utm_medium=app
2026-06-22
Lianhe Zaobao
Author: Laurence Liew is Director of AI Innovation at AI Singapore.
Author: Willie Shi Jianzheng is a Visiting Lecturer at the Singapore University of Social Sciences and a Visiting Scholar at the Lee Kuan Yew School of Public Policy, National University of Singapore.
=====
"Singapore’s position may be more advantageous than it appears on the surface. It has never been the world’s largest technology market, nor does it intend to compete head-on with China and the United States in developing foundation models. But that does not mean it lacks advantages in the AI era—on the contrary, it possesses several structural strengths that are often underestimated."
=====
Over the past three years, global competition in artificial intelligence (AI) has largely been reduced to a “large-model race.” The United States has OpenAI, Anthropic, and Google; China has DeepSeek, Alibaba, and ByteDance; while Europe has been busy constructing regulatory frameworks. Whether in the news, forums, or capital markets, people keep asking the same question: who can train the most powerful model? As a result, computing power, chips, parameters, and training costs have become the standard metrics for measuring a nation’s AI strength.
But if we shift our attention from laboratories to the real world, a different picture begins to emerge. Some organizations with access to advanced models have not achieved the productivity gains they expected. Meanwhile, some economies that do not develop foundation models are rapidly converting AI into practical capabilities, driving industrial upgrading and improving public services.
This reminds us that the real contest in the AI era may not take place solely in laboratories.
What determines a nation’s future competitiveness may not be who possesses the strongest model, but who can integrate AI into the actual functioning of society and the economy most quickly and reliably. The next phase of competition may not be about technological capability, but institutional capability.
This conclusion is not without basis. Advanced models are becoming increasingly accessible. Today, both large corporations and small and medium-sized enterprises can access world-leading large language models through the cloud. Differences among models remain, but the barriers to acquiring the technology are rapidly falling.
What truly creates separation is no longer whether one has AI, but whether one can make AI work in practice.
Many deployments fail not because the models are insufficiently intelligent, but because organizations are unprepared. Processes have not been redesigned, responsibilities remain unclear, data cannot flow effectively, and employees do not know how to collaborate with machines. As a result, advanced technologies remain stuck in pilot projects and never enter core business operations.
In other words, the greatest challenge brought by AI is often not technological innovation, but institutional and organizational innovation.
We might call the capability that supports such innovation a nation’s “AI operating system.” It is not a specific software package, but a comprehensive set of institutional infrastructure that enables AI to be applied broadly and reliably: digital infrastructure, talent development systems, regulatory frameworks, industry standards, organizational governance capabilities, and mechanisms of social trust.
In the past, technological strength was measured by papers, patents, and computing power. At the application stage, however, a more important question is: how quickly can a society connect new technologies to its existing operating systems?
Future differences between countries may depend less on who possesses the most graphics processing units (GPUs) and more on who has developed the most mature AI operating system.
Viewed from this perspective, Singapore’s position may be more favorable than it appears.
It has never been the world’s largest technology market, nor does it intend to compete directly with China and the United States in foundation-model development. Yet this does not mean it lacks advantages in the AI era. On the contrary, it possesses several structural strengths that are often underestimated.
First, Singapore is a highly institutionalized society. Whether in finance, healthcare, or public services, most organizations have clearly defined processes, mature compliance cultures, and strong execution capabilities. This institutionalization is sometimes seen as conservative, but for AI deployment it is a valuable advantage. AI does not function well in chaos; it is better suited to environments with clear rules, well-defined responsibilities, and stable processes.
Second, Singapore’s economy is already highly digitalized. Digital identities, electronic payments, and data infrastructure were established years ago. This reduces one of the most common barriers to AI deployment: the inability of systems to connect with underlying business data.
Third, population aging and limited labor-force growth make AI adoption not merely a strategic choice but a practical necessity. Historical experience shows that necessity often drives genuine technological adoption more effectively than enthusiasm.
Fourth, Singapore’s size makes cross-sector coordination possible. In many large economies, government agencies, universities, industry associations, and training institutions often operate independently. In Singapore, it is easier for them to move in the same direction. Once a new national strategy is introduced, alignment from policy to training and from industry to education can often be achieved relatively quickly.
This coordinating capability may not have been particularly visible in the past, but in an AI era where the speed of technology diffusion determines success, it could become a crucial variable.
These conditions are not merely theoretical.
Since 2017, AI Singapore has driven more than 300 applied AI projects locally, achieving an implementation rate of about 60 percent. By comparison, an industry benchmark commonly cited from MIT is roughly 5 percent. The difference does not lie in model quality but in sequencing.
Projects that succeed are those that complete process mapping, organizational preparation, and responsibility definition before writing the first line of code.
For example, in AI Singapore’s collaboration with local power company YTL PowerSeraya, the team first embedded the system into existing workflows before addressing technical implementation. The result was faster decision-making and higher operational accuracy.
The gap between 60 percent and 5 percent points to an often-overlooked fact: what determines AI success is usually not the model itself, but a repeatable deployment methodology.
First define the problem and responsibilities, then embed the solution into workflows, and only afterward address technical implementation. This sequence of “integration first” is a concrete expression of institutional capability.
It cannot be obtained simply by purchasing a more powerful model. It can only emerge from an accumulation of organizational experience over time.
This absorptive capacity is also reflected in talent development.
Within AI Singapore’s apprenticeship program, about 80 percent of participants do not come from computer science backgrounds. Instead, they are professionals from various industries who have AI capabilities grafted onto their existing expertise.
The goal is not merely to cultivate a small number of elite researchers, but to expand society’s overall capacity to absorb AI through nationwide reskilling efforts.
At the same time, Singapore continues to invest in another form of infrastructure that is often overlooked: trust.
From the Model AI Governance Framework for Agentic AI and the AI Verify governance testing framework and toolkit, to alignment with international standards such as ISO/IEC 42001, these initiatives are often misunderstood as additional regulatory burdens.
Their true purpose, however, is to give highly regulated sectors such as finance and healthcare the confidence to use AI in important decision-making processes.
The earlier conditions concern whether a society can absorb AI; governance and trust concern whether it wants to.
Both are indispensable.
Of course, none of this guarantees success.
The same institutionalization that provides structure for AI can also become an obstacle if regulators treat every deployment as a risk rather than a capability.
Whether these advantages can be realized depends on whether Singapore recognizes that its comparative advantage in the AI era lies not in building models, but in using models effectively.
The significance of this logic extends beyond Singapore itself.
Across Southeast Asia, most economies will not become centers for foundation-model development, yet they face the same challenge of transforming AI into productivity gains.
If Singapore can become the first to develop a replicable model of institutional capability for “using AI well,” its regional value will no longer depend on the size of its models, but on its ability to export governance and organizational expertise that enables AI to be deployed reliably.
Many people still measure AI strength using patent counts, R&D spending, and model scale.
These indicators are certainly important, but they primarily measure the supply side of innovation.
The real challenge of the AI era lies in the distance between supply and absorption: after technology emerges, can it be widely adopted by society, effectively utilized by organizations, and converted into productivity gains and public value?
Ultimately, these questions test not laboratories, but institutions.
This is where the real watershed becomes clear.
It is not about which countries possess GPUs and which do not. It is about which organizations and nations have developed the institutional capability to make AI genuinely operational, and which remain trapped in one pilot project after another.
Pilot projects can continue indefinitely, but value may never be realized. The ability to embed AI into core processes is the gateway that transforms technology into productivity.
Laboratories determine the upper limits of a technology. Organizations determine whether it can enter reality. Institutions determine how far a nation can ultimately take it.
In the past, people assumed that AI competition was a contest between models.
In the future, we may discover that what truly determines success or failure is how a nation organizes itself.
=====
Author: Laurence Liew is Director of AI Innovation at AI Singapore.
Author: Willie Shi Jianzheng is a Visiting Lecturer at the Singapore University of Social Sciences and a Visiting Scholar at the Lee Kuan Yew School of Public Policy, National University of Singapore.
=====
Posts
No comments:
Post a Comment