Minister Josephine Teo's Comments at Financial Times Future of AI Summit

KEYNOTE INTERVIEW BY MINISTER FOR DIGITAL DEVELOPMENT AND INFORMATION JOSEPHINE TEO AT FINANCIAL TIMES’ FUTURE OF AI SUMMIT 6 NOV 2024

Geoff Dyer, Analysis Editor: Minister, good morning. It’s a great pleasure to have you here opening our conference this morning. Thank you very much for being here.

Minister Josephine Teo: Thank you for having me.

Dyer: Let's start with a very big picture question about Al. Just give us a sense of how a country like Singapore thinks about it. Singapore has a fascinating history of the last few decades of constantly reinventing itself to stay competitive, but there are some things about Al that alarm people -- about deep fakes and about privacy. So, when you think about Al, do you think of it as like a new opportunity to retool your economy? Or do you think of AI as a risk or something that needs to be managed or a threat, even?

Minister: I think in this audience, it's very well known that Singapore, at the economic level, is already operating on a very high base. So, for any companies in Singapore, they have to be very competitive. This cuts across financial services, manufacturing and all other sectors that are successful in Singapore. So when we think about AI, we considerits potential to raise productivity, its potential for precision in the way things are done, and greatly, personalisation.

There are many countries that would like to gain leadership in AI, for example, by making sure that they are involved in the development of the most cutting-edge, frontier models. We adopt a different approach. The approach that we want is, to put emphasis on enterprise use. We want to see many use cases being experimented upon, and it is this emphasis on having real activities, with companies and organisations benefiting from the use of AI tools that drives the way we think about AI. So, when we put together the National AI Strategy, it was updated to take into account the prospects of generative AI, but more importantly, it emphasises growing the level of activity in Singapore. So, the AI adoption has to be in sync with the kind of talent that is also being brought into the Singapore ecosystem, as well as the development of the various guardrails that need to be put in place. So, very much an opportunity, provided we also find ways to ensure that there is AI safety, provided we also ensure that the risk that societies are very concerned about are well taken care of.

Dyer: So, you're not thinking purely in terms of your tech sector, but of the economy as a whole.

Minister: Absolutely. In fact, some of the most interesting use cases are found elsewhere (outside the tech sector). We had this observation when we were trying to grow our digital economy, it is not only in the infocomm sector, where you see the most exciting activities. When we try to measure the size of our digital economy, a very significant portion is contributed by digitalised activities happening across financial services, healthcare, and manufacturing. In fact, many banks in Singapore today will think of themselves as being technology companies. If you take away the brick-and-mortar expressions of a bank - and consider even the way we interact with them online - you will see that this is indeed the case. So, it has to percolate to the other aspects of our society.

Dyer: You mentioned the National AI Strategy. You were one of the first countries, I think, to come out with an AI strategy way back in 2019 which seems an awful long time ago. And you’ve updated a new one at the end of last year. What's the one big lesson you think you've drawn that you’d advise other countries? And what's maybe the one thing you might have changed your mind about?

Minister: Well, I wouldn’t call it advice, but maybe some insights on what needs to be put in place. AI is incredibly expensive. From the initial investments to build AI up to the model level, and also, subsequently, when you try to finetune the models and make them useful in your enterprise- or organisational-context, you've got to constantly be thinking about the benefits. You've got to constantly be thinking about in what way is it helping you to grow markets and reduce costs? Unless you are able to make a business case for it, this is not a sustainable effort. No company, no organisation, constantly invests in new technology unless it is able to provide some sort of return. And when you aggregate it across the whole of the economy, I think the same will apply. So, we are practical in the way we pursue these things. On the one hand, we invest in research and development, but equally, we want to emphasise on use cases being experimented upon by all kinds of organisations. And I should say that since we updated the Strategy in December last year, I have been very encouraged by the level of activity that is happening. If you take the example of an airline that many will be familiar with, Singapore Airlines, they've decided that they must also think about how AI can help them to optimise their processes. Airline operations are actually very complex. So, if they are able to use AI for the purposes of analytics, for the purposes of enhancing the customer experience, that could potentially be very powerful for them, but more important than helping them to move to the next level, I thought what was encouraging was them setting themselves an ambition to be a leader in AI adoption for the airline industry.

Now this type of ambition we also find in financial services companies. For example, Prudential in Singapore, even our own homegrown bank DBS, they have identified hundreds of use cases, and are actively developing their own AI models to be applied to those use cases.

And then I would also say that within the government, there are exciting developments. We decided at the very early stage that for us, as policymakers, to have a good grasp about the kinds of guardrails that really will be useful, we ourselves have to be users of AI in a very comprehensive and systematic way. We made it clear to government officials that they are welcome to experiment, provided they take ownership and accountability of the results. They cannot just push the problem to the AI and say that if anything happens that is not good, it's the AI's problem. They have to be able to account for this. And it's actually gone quite far. For example, across the whole of the public service, more than 65,000 of our own officers are using a secure version of generative AI, in order to do things that all governments have to do, for example, transcribe meetings. From the productivity angle, this is happening at a fast clip. We also have our own platform that enables government officials to design and build our own AI bots, and I'm glad to say that this experimentation has seen several thousand bots built. This kind of energy, I think, is very much to be encouraged.

Dyer: Is there any popular backlash in Singapore against any of this, other than security concerns?

Minister: I think there will always be a concern. It's not just in Singapore. I was in China recently, and they were telling me how the introduction of robotaxis in the city of Wuhan met with great pushback, because the drivers who were operating private hire vehicles suddenly felt that their livelihoods may be at stake. So, there is a great concern to people, particularly knowledge workers, white collared workers, who feel that in the past, when we talk about technology, when we talk about automation, it's likely to be blue-collared workers, or the rank-and-file workers, who are going to be affected. But in today's context with AI, with generative AI, in particular, it could be their own jobs at risk, and there have been studies that would say that because of the high percentage of professionals, managers, executives and technicians in our workforce, then we are much more exposed. But we kind of think of it differently, and we want to enable our people to think of it differently. I think the competition is not necessarily between humans and AIs. It's going to be much more a human that is AI enabled versus a human that does not have the ability to work with AI. So, the way in which we would want to go about it is to try and strengthen the skills development process to ensure that not only are we growing a pool of AI practitioners - and by that we mean data scientists, machine learning engineers - but also for the entire workforce to be able to make use of AI tools and feel confident in using these tools. That is the effort that we have started to invest in.

Dyer: Do you need migration to do that? Do you need to bring in skilled people from outside, or can you do that with training of people within the country?

Minister: You will always need a combination of both. In fact, I should say that many of the leading tech companies in Singapore are very much a part of this effort. Take Microsoft. Clearly, they have an interest to ensure that there are enough people in the workforce that are able to use Copilot. So, they have committed, for example, to train, I believe, it's more than 10,000 people in Singapore. In the Singapore context, with a workforce of 3.5 million, 10,000 is not small. Take a company like Google, they're all very much invested in these kinds of efforts. So, we see that in a matter of three to five years, it's entirely possible that many more members of our workforce will gain confidence in the use of AI tools. There is no other way to assure people that their livelihoods need not be at stake. In fact, it will give them the potential to advance further in their careers. I think all people, whether they are operating at a junior level or at a senior level, want to see some form of career advancement and growth. And we'd like to find a way to let the mastery of AI skills be a way forward for them.

Dyer: You mentioned guardrails a couple of times. Do you need new laws or new regulations in Singapore to be able to weigh the way you might regulate the potential for deep fakes or fake news?

Minister: The short answer is yes and no. If you're concerned about the potential risks of discrimination, for example, if you have workplace legislation in place against discrimination for fairness, those kinds of laws very often can be applied no matter how the discrimination comes about. So even if it is an algorithm that caused discrimination in the recruitment process, the law should be able to accommodate it. It should be technology agnostic in that sense. Singapore is also one of the few countries in the world that has a law against fake news. We do not remove the content that that is being challenged, but we require a clarification to be put alongside that offending content, and then people can judge for themselves. Again, it doesn't matter how that misinformation is generated. Even if it was generated with the help of artificial intelligence, the law can still apply. In that sense, quite a number of the laws that already exist may be able to deal with the risks that are being brought about by the use of AI, but there will be certain instances where additional guardrails will be needed. So for example, for elections.

Dyer: Just yesterday.

Minister: Yes, yesterday with results that everyone is concerned about. I was just watching television, and it was reporting about the results in North Carolina and how misinformation may have played a part. So, misinformation in the context of elections is particularly tricky and also of great concern. Our Parliament passed a new law just last month that would make it a criminal offense for anyone to depict a candidate doing something that he did not do or saying something that he did not say during an election period. So those are the kind of additional guardrails that may be needed, because we see that AI can be used to generate these kinds of content, and then, when put online, these can go viral very easily. Now we will continue to look at instances where the risks of AI are not adequately dealt with by existing rules, but I wouldn't say that at the outset, we start off being able to know exactly where the AI is being applied that will have wide impact, that will require new guardrails. We adopt the approach of being very watchful and being able to respond to it in an agile and quick fashion.

Dyer: Doesn't there need to be a global discussion about this as well?

Minister: Absolutely, we can't do this ourselves. But sometimes, you know, global discussions take a lot longer, so we have to move when we are able to, and hopefully what we do can serve as a useful reference point. We don't know for a fact that the guardrails we put in place will be adequate, but we're very open to sharing it with our colleagues around the world, and then improving them, as well as to contribute to the conversation.

Dyer: A lot of the discussion about AI in recent months has focused on data centres and the infrastructure needed to support the industry. Singapore is a small country that is not over abundant in renewable energy resources. How are you going to manage to stay competitive in that part of the AI sphere?

Minister: You are absolutely right. We started thinking about this. One observation to make in the first place is that relative to our GDP and the size of our population, if you look at the amount of data centre capacity in Singapore, it is already one of the densest in Asia. So, you know, that is where we start from. But at the same time, we understand that as the AI workloads become more pervasive, then surely, in terms of data centre capacity, we will also need to have some head room for growth. That's why, from about 18 months ago, we decided to go on a pathway to put in place a road map towards greener data centres. Essentially, there are two things that you can do to make the data centres greener. They can use less energy or they can use greener energy. And we are pursuing both pathways. In terms of using less energy, data centres typically will have a 15-year cycle and there will always be a need for tech refreshes. So, when the cycle comes for existing data centres, there will be an opportunity for them to level up. In terms of new data centres that are coming on stream, then these higher thresholds and requirements can already apply. Now another area that is, of course, interesting is that we will continue to look at is how renewable energy can play a part in powering the data centers. So that is also an area of active investigation. For that, we work together with the companies and research community to try and identify pathways forward. So, it's something that we have thought about for quite some time, and we will continue to build on that.

Dyer: At strategic level, do you need to have the data centres within Singapore?

Minister: That is an absolutely interesting question. We had to deal with this question with manufacturing. Singapore is not big enough for the entire supply chain to be within the country. Some parts of the value chain can be hosted in Singapore because it plays to our advantage. So, for example, we have very efficient logistics; we are an international maritime centre; and we are also an international air hub. So, for high-value components that need to be shipped out or flown to other facilities, Singapore is a good place. Singapore is also a trusted hub where intellectual property is strictly observed and intellectual property rights can be defended. So from the company standpoint, it can be very advantageous to host certain activities in Singapore despite our cost structure. I believe that the same will happen for AI workloads. You will notice that in the region, and I think that FT has reported on this before too, that there are many data centre investments that are taking place in Malaysia, Thailand, Indonesia, and you could potentially see this as a grid operating together. And it will be very much up to hyperscalers and companies that have regional footprint to decide where they will distribute their AI workloads.

Dyer: A lot of the discussion of AI is often framed in terms of the competition between the two big beasts in America and China, the two big competitors. Where does that leave Singapore an does that leave you in a difficult position or is that an opportunity as one of the few countries that actually has good access to both of those countries?

Minister: We’ve worked well with both countries, and actually internationally, they're not the only two countries that we interact with. So, to give you an example, with the US, we have a dialogue on Critical and Emerging Technologies. It has gone well and allowed us to identify areas that we can strengthen our collaboration on. AI safety is one of those areas. In fact, later this month, there is an inaugural meeting of the International AI Safety Institutes taking place in San Francisco, and Singapore is part of that conversation.

With China, we have a digital policy dialogue that has also gone on quite well. Among other things, we are interested in how the Chinese look at AI needs to be regulated, especially when it comes to content that affects societal harmony and cohesion. That is an area that we both have an interest in. But I should also say that it is perhaps too narrow to just look at our relationship with these two countries.

There is a large community of interested stakeholders. For example, countries in the Global South, everyone wants to see how to benefit most from AI. In support of this effort, we came up with an AI Playbook for Small States. This is also in support of our role as the convener of the Forum of Small States within ASEAN. I should say that the digital ministers understand each other's priorities, and also have commitment building on what we have been able to do together in cyber security to advance on AI development. So, among other things, we, at the ASEAN level, have an agreement on what kinds of AI deployment should be encouraged in order to meet ethical standards. We have a data management framework, and we are working towards strengthening cooperation at the AI ecosystem growth level. There is, of course, a foundation of digital economy agreement that is already in quite advanced discussion, for a digital economy framework among ASEAN countries. I see this as a really important part of our engagements. After all, our stated ambition is AI for the Public Good for Singapore and the world. So, it means that we expect to make our own contributions just as much as we will benefit from global developments.