Josephine Teo部长在IBM Think on Tour新加坡站的讲话

Moderator Ana Paula Assis, Senior Vice President and Chair for IBM Europe, Middle East, Africa and Asia Pacific: I often come back to the idea that “winning” the enterprise AI race looks different depending on where you sit. With Singapore’s updated National AI Strategy, which introduced the sectoral missions and the new AI Council, led by the Prime Minister, it is clear that you are stepping up the ambition for AI in Singapore. From your perspective, could you please define the version of success that you're aiming for, and what prompted the shift in approach here?

Minister Josephine Teo: I would characterise what we're doing in the updated National AI Strategy as a “double click” rather than a system reboot. When we put out National AI Strategy 2.0 in December 2023, we kind of already had a sense that the developments were going to happen at such a fast clip, that as soon as we put the document out, parts of it would already be outdated, and it would always need a refresh.

But what was really interesting is that, as we implemented the strategy, we realised something had inverted compared to previous technology cycles. With earlier technologies that were being introduced, our first contact as individuals was usually as members of the workforce. Our organisations adopted the technology, then conducted massive rounds of training. We learned how to use the technology at work before we realised that this technology could have personal applications and started using it at home.

The personal computer was like that. Our first encounter with the computer was in the office. For many people, the first interaction with email was in the office too. But AI is quite different. AI is truly democratising access to something that can have a profound impact on our lives. You and I use AI almost daily. I am pretty sure that on your smartphone and other devices, you are already using AI. Today, the converse is true: we might be using AI to a larger extent than our organisations are.

Many of you have probably tried vibe coding. You are building apps to keep track of your children's schedule. You are using it to scan the internet for interesting articles and to surface them to you. I have an agent that does that for me.

So here's where the gap is: individual productivity is improving is at a faster clip than enterprise productivity. Then you think about the bigger challenge. How does this technology impact industries? How does it impact the entire nation? That was what we were thinking about when we refreshed our National AI Strategy.

We said we have to put more emphasis on helping enterprises adopt this technology and integrate it into their existing processes. As is natural for all businesses to expect, there also has to be an ROI. At some point, this has to translate into commercial value, into dollars and cents producing a real, tangible benefit. So that was where we chose to place the emphasis.

Hence, the push to accelerate and boost efforts in helping businesses, especially at the leadership level, think about how they can make this technology come alive, not just for individual employees, but in the enterprise context.

For example, we have also a collaboration with IBM targeting leaders who hold responsibility for digitalisation in their organisations. We want them to think strategically about how AI can be used in their operations and how they might rethink their business models.

We have a National AI Impact Programme that hopes to do this on a wider scale. This programme complements what we are also hoping to do through the National AI Missions, which are more about the end-to-end transformation of sectors. They are advanced manufacturing, healthcare, finance, as well as connectivity, which together contribute 40% to our GDP. We are hoping to push these, so that the benefits don't stay at the individual level, but expand to enterprise, industry, sectoral, and national levels.

Moderator: On to the next question. One point of yours that really resonates with me is this commitment to ensuring AI drives inclusive growth, not just productivity for the few. And again, we're thinking deeply about how AI can augment people and enhance their capabilities, not simply replacing tasks. How is Singapore approaching this opportunity, and what early progress and insights are you seeing so far?

Minister: When we think of inclusivity, it can also cover several dimensions.

There is inclusivity at the individual level. We're concerned about pockets of our workforce that may not have adequate exposure to how AI can transform their domains. If you work in a technology-infused environment, you are very likely to be exposed to AI tools. But in some professions, in some areas, adoption may lag, and it may take some time.

So, when we were implementing our workforce capability development programmes, one area deserved greater attention. We were thinking of AI practitioners, data scientists, machine learning engineers. Expanding the pool of such practitioners in Singapore was a meaningful exercise, and we set about doing it.

We also noticed there are other professionals, experts in their own domains, who had something the AI practitioners did not. They have intimate knowledge of all the problems that they faced in their respective areas of work, but not quite how AI could help solve them.

So we developed this idea of AI bilinguals – experts in their own field who pick up enough AI knowledge that their interactions with the data scientists and machine learning engineers become far more meaningful, and far more likely to translate into and real benefits in their areas of responsibility.

We are talking about people like process engineers, who know the manufacturing line and the production processes deeply. We are talking about lawyers, accountants and HR professionals, who have something to bring to the table, who can help organisations benefit more from AI adoption.

That was one type of inclusivity: how we include people whose organisations and professions were unlikely, on their own, to become early adopters of AI.

But there is also another very important type of inclusivity. I know you would be very familiar with this: technology diffusion, it is very often the companies at the frontier, like IBM, that use the technology in every way imaginable to improve how they function. But there will be many small and medium enterprises who lack the knowledge and resources, and who don't even know where to begin. If we are truly inclusive, we must also have programmes that reach the small and medium enterprises (SMEs).

There is one other way which we think inclusivity also matters, and that's at the country level. Singapore is very fortunate in the sense that we host great technology companies like yourselves (IBM), and so through you, we have exposure to some of the most advanced products and services available. This is not necessarily the case in other countries, as can be seen among our neighbours, for example.

So through our participation in, and as convener of the Forum of Small States, which has 108 members, we developed an AI playbook. Within ASEAN, we share our open-source model – the Southeast Asian Languages in One Network (SEA-LION), large language model. It is not very large. I think only 17 billion parameters, which is small compared to the biggest models. But it has real usefulness in our context because it captures the languages and the cultural nuances of Southeast Asia.

These are the different ways in which we hope to be able to promote the idea of inclusivity.

Moderator: I always admire the way that Asia is bringing all the people along in the adoption of AI, and I think that this could really serve as a blueprint for the rest of the world. Minister, you talked about SMEs, right? It is not just a technology that is going to drive benefits to the large corporations, and this is a very tough balance to strike: scale versus growth. How do you make sure adoption at that scale creates real impact, not just ticking the boxes to make sure that you really reached all the segments?

Minister: It is really quite interesting because when I interacted with some small businesses, I was quite struck by how creative they were in trying to use AI.

I've been to a small F&B outlet. When I was talking to the business owners, they told me that they were already using AI, and I was very curious how. Well, there are two people, and the way in which they first picked up AI skills was to go for training.

We have an individual learning account, called SkillsFuture. The Government puts credits in it. You can use the credits to go for training, and these two entrepreneurs did exactly that. They went to a training provider and learnt how to prompt. The platform they were doing business on had also created an AI assistant, which they learnt to tap to analyse their sales data.

Their business on this platform was not all of their business - it made up about 20% of their revenues. But that was enough data for them to make sense of their sales promotions. So, through learning the skill of prompting, they used this AI assistant that was provided by the platform to figure out which of their promotions were generating better returns. They also learned to use AI tools to generate their own marketing materials.

So while there may be a resource or knowledge gap, the adoption can be quick once you plug this gap. The programme that we put in place, the National AI Impact Programme, works with partners who perhaps already offer Software as a Service (SaaS) products and services to many small and medium enterprises, and the idea that you can add an AI tool to the offerings is, I think, a really reasonable way for the enterprises, to get to learn how this technology can benefit them.

We are also very keen to think about companies that may have more bespoke needs. In other words, a standard offering, through SaaS may not satisfy their needs, so we have a parallel programme – we call it the Enterprise Compute Initiative, where by working with partners, we make compute available, to try and moderate the expense requirement for SMEs. This is especially when they need to develop something that is more bespoke, something that they can't quite get through an intermediary.

So these are the different ways in which we are trying to give our SMEs a leg up, and bring to life the idea that AI is a democratising technology.

Moderator: You have this great initiative called the Digital Leaders Acceleration Bootcamp. Can you tell us a little bit more about how that helps in that democratisation strategy?

Minister: Well, it comes from talking to many individuals about their experiences in their organisations. Individuals were telling us that “I'm learning AI skills but my bosses aren't”. There are many others who are, perhaps, still slightly behind the curve. So people were telling us that we need to help business leaders understand the usefulness of this technology, and we cannot just rely on theories. We need to give them hands-on experience, introduce them to prospective partners that they can also work with, so as to implement AI in a sustainable way in their organisations.

After all, AI is not a cheap technology. You can easily use up tokens if you did not architect it right. These things will always feed into a business owner's calculations: How much does it take? How much does it cost me to use this tool?

The Digital Leaders Accelerator Bootcamp was designed with this purpose in mind. It was designed to give business leaders, particularly SME business leaders, a fuller and deeper understanding of all that comes with AI, including the risks, not just the benefits, and how they can put in place the right systems and processes in order to use AI effectively.

We decided that this was not something the Government should do on its own. We should work with partners because the technology moves very fast, and we benefit from their experiences in other domains, perhaps in other countries as well. We already have such a programme because the digitalisation journey serves as a very useful foundation for us to promote wider AI adoption. So expanding this programme, infusing more AI elements in it, is a very natural progression.

Moderator: That's very good, and we are very proud to join the programme. I think we're the third organisation to do that, and we are doing that as one idea – bringing together our consulting expertise in business transformation with our enterprise AI technology, to help organisations identify meaningful use cases, build Minimum Viable Products (MVPs) quickly, and most importantly, develop the internal capability to continue their AI journey long after the programme ends. So, transferring the knowledge, and for many organisations, the challenge isn't a lack of AI ambition. It's knowing where to start and how to scale responsibly, and that's exactly what we hope to contribute to the programme.

Minister, I want to shift gears slightly because everything we've just discussed in terms of ambition, inclusion, scale only holds up if there's trust underneath it. Trust, governance, and agentic AI. One thing that I emphasize in my talk is that trust really comes first, because without it, AI adoption simply doesn't scale. As you move from experimentation to embedding AI in real operations, governance becomes critical to sustain the trust. You launched the world's first framework for agentic AI at WEF Davos and you have already updated it. What are you learning as you translate these principles of trust and governance from framework to real world practice?

Minister: Well, if we take a step back, even before we put out the first National AI Strategy, we had articulated an AI governance framework. So, for me as a minister, I was also quite struck that my predecessors had already thought about governance even before we put out a set of strategies. We have continued this emphasis on creating a trusted environment for AI implementation.

The truth of it is that AI creates risks that are different than the risks we are used to. Our risk management frameworks, safeguards we put in place to all kinds of systems-building safety, fire safety, aircraft safety – all of these were conceived of and designed to tackle the kind of issues that humans create, when humans are in the loop.

But when you use agents, when agents are designed to make decisions on behalf of humans, the kinds of errors, risks, and safety issues that arise, are honestly not completely well understood. It will be quite a long time before we fully appreciate the range of problems that could be created when agents are unleashed.

We know the kinds of problems that we humans create, and we have designed to prevent those problems from becoming disasters. We don't exactly know whether agent systems will produce the same range of errors. Therefore, we have to take a step back and ask ourselves: What then do we do in order to be able to promote agent adoption?

So, it is very important for us to understand what we are designing the agent workflow for, and we really have to spend time thinking about the harnesses, the principles, and how to mitigate problems as they arise.

So it is still early days. I would say that in this regard, it is our intention to work with partners who are in the business of helping their clients and networks use agents in a more effective way. I think through learning from their actual experience, we have a much better chance of improving our methods and designing tests before these systems are fully put into production.

Moderator: Minister, we are starting to come to a close in our conversation. Let's now take a peek into the future – let's talk about quantum. Singapore has its own national quantum strategy. Where do you see the biggest opportunity for the country in quantum, and how are you thinking about preparing for it today, not just five years down the road?

Minister: As it turns out, from about 20 years ago, we already invested in starting up a centre for quantum technologies, and for the longest time, we had experts who were working to deeply research the topic and bring their capabilities up to world-class levels. I'm really glad that they've been able to do so. If you talk to the international scientific community, they are aware of the capabilities within Singapore. So that's a really good foundation.

When we started thinking about levelling up our quantum capabilities and preparing for the quantum age, there were a few areas that we were concerned about. The first was, do you have the option of quantum safe communications? Because with powerful quantum computers, whatever is decrypted today may no longer be protected. How do we also, apart from building up quantum-safe communications network, help people to migrate to quantum-safe cryptography? So those were the activities that have been engaging us.

Two more recent developments bring a lot of excitement. One is that I think we are at the stage of development in quantum computing that it is reasonable for us to ask how do we build up the skills in order to develop quantum algorithms and to be able to grow, build the applications that will be relevant in our context, and what would be relevant in Singapore's context.

Every year we are dealing with millions of container movements, and if you have been to a container yard, you will know that these containers are stacked by now almost ten levels high. So, it makes a difference whether the box that you are unloading from one vessel is stacked at level two or stacked at level eight, because when you need to reload this container to the next vessel, if you stacked it at the correct place, there's a lot less work.

But these are very complex computations. It would take a classical computer years to make one assessment each time the box could not be loaded or was loaded at a different time. So there is a use case for us. Certainly, our financial sector also has uses for quantum computing. We think that there is already a basis for demand for algorithms and applications. That's one area that we're putting effort into.

The other area is equally exciting. Until recently, we did not know that Singapore could participate meaningfully even in the manufacturing of components for certain types of quantum computers. It turns out the production process has adjacencies with our semiconductor manufacturing capability, particularly in advanced packaging.

Not long ago at Davos, I was sitting next to Nobel Physics Prize winner, John Martinis, explaining Singapore's interest in quantum, when he shot his hand up and said, "I'm going to Singapore!” I was very curious, so I asked him, "What are you going to Singapore for?” He said, "Because you have advanced packaging, semiconductor manufacturing”. His startup, a company by the name of Qolab, is in fact working with our federated foundry to design and produce components for superconducting computers. So there are very exciting developments here.

Moderator: Super exciting. To conclude, Minister, what is the one thing that you hope people in this room take away about Singapore's future direction and what could it mean for all of us?

Minister: Well, in a simple sentence, we would like Singapore to be the home of AI innovation, and we welcome you to be part of our vibrant AI hub.

Moderator: We accept the invitation, Minister. Thank you so much for your time, the clarity of the vision that you brought to this conversation, it's a big privilege to partner with Singapore and with you.