Serena CESAREO, Senior Researcher, Tortoise Media Simon CHESTERMAN, Vice Provost, National University of Singapore and Senior Director of AI Governance, AI Singapore Shameek KUNDU, Member of Singapore Government AI Advisory Council Andrea PHUA, Senior Director, National AI Group, Director, Digital Economy Office, Ministry of Digital Development and Information, Singapore Moderator: Ellie AUSTIN, Deputy Editorial Director, FORTUNE
Category
🤖
TechTranscript
00:00Now if the first session with Clay and the Minister was about Singapore as a global AI powerhouse,
00:07as I mentioned in the introduction, we're now going to talk about the extent to which Singapore can be used as a test case
00:13for small and medium tier countries. And Simon, I'm going to start with you.
00:18We've heard mentioned that for such a small country, Singapore really does punch above its weight in terms of AI investment and innovation.
00:25What lessons should other small and mid-sized countries, maybe the ones we've just seen on the Global AI Index,
00:32take from Singapore's success in this area?
00:36I think it was pretty clear from the outset that a country like Singapore could never compete with China and the United States in terms of scale.
00:43So if you can't beat them on quantity, you can at least try and compete on quality.
00:47And so we've really invested in talent and in a regulatory environment that encourages innovation.
00:53And I think that's maybe something that Singapore is at least an interesting experiment,
00:57that in contrast with the kind of European approach, where it's often a kind of oppositional mentality,
01:02how do we hold back? How do we protect the rights that might be at stake? That's one approach.
01:07The American approach is very hands off. How do we sort of just fan the flames of the free market?
01:12Or the Chinese approach, where it's very much state controlled. Here it's very much intended as a partnership.
01:16And so you make a virtue of small size. You make a virtue of having a highly connected,
01:21highly educated digital population and a small environment where you can organise events like this,
01:26where you get the regulators, the industry and the think tanks around the same table to try and craft a whole of nation strategy
01:34that gets everyone aligned on how to take advantage of the digital economy while minimising or mitigating the risks.
01:40And Shamik, before I ask you a question, I believe you've got a new title as of yesterday that we want to congratulate you on.
01:46Thanks. Yeah, as of yesterday, I've just joined as the head of the AI Verify Foundation,
01:50which those of you from Singapore know. And I guess the relevant aspect, thank you,
01:54the relevant aspect to Simon's point is, I think, Simon, you mentioned about openness and about a regulatory environment
02:00that is encouraging of innovation. So the AI Verify Foundation, for example,
02:04takes the approach around AI testing and safety that it shouldn't just be something that people tick boxes against
02:11or something that people do because they're asked to do. Instead, how do we make it a forum, a toolkit that allows companies,
02:18particularly smaller and medium-sized companies, to adopt AI? So you go from one end, which is,
02:23hey, use AI safely or otherwise don't use it, to the other angle, which is, actually, how can we help you adopt AI safely through toolkits?
02:31So I think that's a continuation of that theme about encouraging regulatory environment.
02:36And, Andrew, I'm going to come to you in one second. But before I do, Serena, I wanted to ask you, from your research,
02:42are there any small, mid-sized companies that you expect to be shooting up the rankings soon to compete with Singapore?
02:50I would say, yeah, definitely. There are a lot of other countries that I've mentioned earlier also during the presentation
02:57that are sort of smaller, but also really dynamic AI hubs for different reasons.
03:03For example, when it comes to talent, Germany, France and Canada are quite high on the index.
03:10When it comes to investment, Israel and Canada, again, and Korea are also very strong performers.
03:17In infrastructure, Canada and India as well, we expect to rank a bit higher.
03:24So, yeah, it really varies. And I guess it depends on when you look.
03:28But, yeah, definitely there are other countries aside from Singapore.
03:32And, Andrew, we were talking backstage and Serena was saying, and correct me if I'm wrong,
03:36that Singapore has gone up six places in three years on the global ranking list.
03:41I wonder, I'm interested in your insight on how it has achieved that ascent really in a relatively short period of time.
03:48What can you pinpoint there?
03:51I think if I think about how we came up with the strategies across government,
03:55it's really thinking about AI from a systems level approach. It's not about using AI for AI's sake.
04:02It is about making our industries more competitive, delivering government services better,
04:07and helping our citizens also understand and interact with this technology in a way that builds confidence and trust.
04:13And what does that look like? When you talk about helping citizens understand tangibly, what are you doing to make that happen?
04:19So, we spend a lot of effort skilling our workforce, telling them about the new technologies and skills that are required to do better in the workplace.
04:29We spend a lot of time engaging with industries and companies to talk about the job redesign that needs to happen
04:35so that people can use these technologies effectively.
04:38With our seniors and with our school-going children, we spend a lot of time, too,
04:42telling them about some of the ill effects of using AI poorly or using it wrongly.
04:49And that, I think, goes a long way into helping people understand what it should be used for and what it shouldn't be used for.
04:56And what are some of the things that you need to be aware of as you're dealing in the online space?
05:01So, these are some of the ways in which we try and educate, build awareness,
05:05and also put in place safeguards and guardrails that ultimately tries to promote the use and the investments in it
05:14and not stifle the innovation, even as we care about these important matters of safety and security.
05:21If I can add to Andrea's point, Singapore also really invested in its research ecosystem.
05:28If I remember right, from last year, Singapore's research and development spend per capita
05:35was something like 18 times higher than the one of the US, which I think says a lot.
05:42A few of you have mentioned talent and investment in talent being a key reason for Singapore's success.
05:48If any of you were advising the governments of France or Denmark, one of those countries in the middle of the list,
05:54what would you be telling them they should be doing with their talent investment?
05:58What should it be looking like?
05:59Shameek, let's start with you.
06:01Well, I think there's also questions of what kinds of talent.
06:03So, if I look at France, for example, I think in absolute terms,
06:06the number of kind of deep tech talent in France might well be higher already.
06:10I think one thing Singapore does well, Andrea, you can comment on this, is it looks at talent holistically.
06:15It's not only about the person, about the machine learning engineer or the data scientist.
06:19It's also about the Singapore Academy of Law, for example, Zeke will speak later today.
06:24They're talking about how lawyers can use generative AI more effectively.
06:28So, I think there's a whole of country approach to AI talent.
06:32It's not only about AI development talent or data science talent.
06:36It's also about how all of the economy, actually even students, how they use AI.
06:42So, that's one thing I would certainly advise any country, irrespective of how strong your core AI talent is.
06:48You should look at AI enabling the entire workforce and indeed students and elder people even.
06:53That's very time intensive, I imagine, and very resource intensive.
06:59Simon, it seems presumptuous in a way to imagine that every government of a small country
07:05might have those available resources to funnel into AI integration, you know, society wide.
07:12If those resources aren't available, how does a country approach this rollout?
07:17Yeah, well, again, I think if you're trying to compete on scale, you're going to lose.
07:20To do well in AI, it's well known you need talent, you need data and you need compute or computational power.
07:26And not everyone's going to have the computational power, not everyone's going to have the data.
07:29But the talent is where you might be able to make a bit of a difference.
07:32And so, I think the challenge, particularly for developing countries,
07:35is whether they want to invest in the hardware and the model development
07:39or invest in the talent that can take advantage of that.
07:42And I think that goes on to Shamik's point, which is you don't need everyone to be a computer scientist.
07:47Indeed, France, I think, is a good and bad example.
07:49Good example with initiatives like Mistral is a great new company.
07:53But I think about five years ago, they said, right,
07:55we're just going to train everyone in computer science or 20,000 computer scientists.
07:59Computer science is important, but actually more important than just training up AI models for AI model's sake
08:04is training the entire economy so that people can look out for opportunities
08:09because the large language models are pretty good at producing human-like text.
08:13They're very good at coding.
08:14So, the idea that you just need coders, I think, misunderstands the opportunity of AI.
08:18It's not just going to be a computer science regime.
08:20It's going to be an economy-wide factor that's going to be a bit more like electricity,
08:25which transforms whole economies.
08:27So, not everyone needs to be an electrician,
08:29but you need to know how to develop tools that will use electricity or apply electricity in new ways.
08:34So, at what point should we be talking to children in schools about AI coding?
08:39So, we already are.
08:40So, AI Singapore, there's free educational opportunities through AI Singapore for everyone.
08:45I think, candidly, I'm a little bit wary about getting kids too young
08:48because I think they need their screen-free time.
08:50But throughout the education lifecycle, certainly at the university level,
08:55we're training students to understand the opportunities and the limitations of AI,
08:59not because everyone needs to be a computer scientist, although we have great computer scientists,
09:03but because the lawyers, like Shameik said, the engineers,
09:07everyone's going to need to operate in an environment
09:09where AI is hopefully going to enhance their ability to do their job in new ways
09:14and hopefully to do new things that they never could have done in the past.
09:17I've got a couple more questions, and then I'm going to open it up to you in the room.
09:21Andrew, you talked a bit about strategy earlier,
09:23and Singapore has remained impressively neutral when it comes to the big AI power blocks.
09:29I wonder, is it realistic to think that a country of Singapore's size,
09:32an equivalent country, can continue in that way going forward?
09:37Or is that ultimately going to be a point where they have to express some sort of loyalty?
09:42I think geographical diversity is absolutely critical as we build AI products and tools for the world.
09:48So building on the interdisciplinary nature of AI,
09:51as well as the fact that AI understanding can start from any age and anywhere,
09:55I do think that it is more powerful and more beneficial when we build AI products
10:02that are built by teams of people that are based all over the world,
10:06that have specific local and regional contacts,
10:09that can help us to translate things that will suit that context better,
10:14and that can help us to identify safety and security risks
10:17that may occur differently in different parts of the world too.
10:20So I think the usual binaries is really an artificial construct that we have to constantly work against.
10:28I'll give you a couple of examples of how Singapore does this.
10:30We obviously collaborate very closely with the US government and with US industry
10:34because they are tremendous partners for the Singapore ecosystem.
10:37But likewise, we also have conversations with the Chinese government on AI governance policies
10:42to build that mutual understanding.
10:44There's also a tremendous amount of research that goes on at all levels.
10:48Besides that, at ASEAN and at the UN, we are the leaders of a digital forum of small states.
10:55That's 108 small countries that are also thinking about the capacity building,
10:59the investments that they need in infrastructure,
11:01the type of education and upskilling that's required for their workforce.
11:05And we use these platforms to talk about really important subjects
11:09as to how you might want to govern AI well, what sort of capacity is needed,
11:14and how can we learn from each other through all of these various platforms.
11:19It's a small start, but I personally think that it's a really important investment
11:23in this international collaboration because the more we understand what is at stake,
11:29and know that we have friends and partners to guide us through that journey,
11:33I think we are a lot better for it.
11:35It's a small start, but it's a mighty start.
11:37Shamit, so earlier this year we saw the publication of the ASEAN AI Guide,
11:42which appears notably more lenient, I would say, than the EU's AI Act.
11:47And I wonder, has Singapore had any response from EU governments?
11:50Has there been any pushback to it?
11:52I mean, I'm probably not qualified to speak about whether there's a pushback or not.
11:57But I think the point about lenient, I mean, I don't know.
12:00There are some aspects of the EU AI law
12:02which appear to be restrictive to many people in the industry.
12:05But I think a lot of this debate about whether something is lenient
12:09versus something is more strict, it's somewhat superficial,
12:13with a small number of exceptions around the governance
12:16of large language models and foundation models.
12:19There's actually a wide degree of agreement around the world.
12:22So I would not characterize ourselves as lenient.
12:24I think pragmatic would be a better word.
12:27And that is true for Singapore.
12:28I think that is true for ASEAN and, frankly, for the whole region.
12:31And, you know, people often...
12:32Just to continue on the example,
12:34I often have a lot of colleagues in the West, in my private sector days,
12:38who'd say, well, Asia can go out because they're very liberal
12:43and it's Wild West out there in terms of...
12:46Actually, I'm sure you know this,
12:47but some of the strictest rules around AI use are actually in China,
12:51which is often quoted as the Wild West.
12:53So I think this, again, not going by the binary,
12:56I think Singapore's approach and, indeed, the region's approach
12:58is more pragmatic rather than being...
13:01And pragmatism can sometimes be very strict
13:03and sometimes it can be very liberal.
13:06That said, I think one of the things that Singapore is really trying to do
13:09is navigate between two sets of risks.
13:11And there is a risk of under-regulation.
13:13If you expose your citizens to risk,
13:15if the Wild West comes, then it can be a really dangerous environment.
13:17So we've got to make sure there's a baseline of support.
13:19But there's also a risk of over-regulation,
13:22particularly for a small economy like Singapore.
13:24Over-regulation could drive innovation elsewhere, could constrain it.
13:27And I think that's one of the things that this part of the world
13:30has really tried to navigate between.
13:32And one thing that Singapore has done and that the AI ethics framework
13:36that ASEAN adopted also reflects
13:38is a realisation that we've got to explore this new territory
13:40in partnership with industry.
13:42And so that's why initiatives like AI Verify are so important
13:45because they're not about telling companies,
13:47don't do this.
13:49It's about telling companies, OK, how can you use AI responsibly?
13:52How can you live up to the standards that you all espouse?
13:55How can we navigate this new territory together?
13:58What isn't Singapore doing well that it could be doing better?
14:04All of the above.
14:06More money for research.
14:07Sorry, my colleagues at university would kill me.
14:09No, that doesn't count.
14:11That's a cop-out, that answer.
14:13Any of you?
14:14Any? I mean, OK, we can frame it differently.
14:17How do you hope Singapore's AI strategy to evolve in the years to come?
14:21If we're having this conversation in five years' time,
14:23what do you think has happened by that point that we're not witnessing now?
14:26So I think if I can respond to this,
14:29what I'm seeing, and I believe these are the three sort of core pillars
14:34of Singapore's new AI strategy,
14:36are infrastructure, talent and adoption.
14:39And I see these three things as being the three things
14:42that most new strategies of smaller but quite advanced AI hubs
14:48are also focusing on.
14:49So infrastructure, I mean, compute was mentioned earlier.
14:53So high-performance compute that can sustain AI systems is really, really key.
14:59Canada, I think, secured $2.5 billion recently
15:03for high-performance compute infrastructure.
15:06And I think this will be something that we will see grow in Singapore
15:10as well in other countries.
15:12And same for talent retention, high advanced talent
15:15and adoption of AI across different industry and different sectors.
15:19India also is really, really banking on this,
15:22specifically for sectors like education, healthcare, agriculture and other ones.
15:28I think one thing that really will pose a challenge in the coming years,
15:31and it came up in the interview with Minister Teo earlier,
15:33is the tension between two big challenges of the 21st century,
15:36which is technology and climate,
15:38and how we get access to the energy we're going to need
15:41in a way that still maintains Singapore's environmental commitments.
15:44And I think that's going to be a really tough challenge,
15:47but it was good to see the minister,
15:48and the government is very open to the need to confront that directly.
15:53I want to open up the room to see if there are any questions.
15:57If you have a question, please raise your hand.
16:01Yes, I can see one straight away at the back,
16:02and someone will come to you with a mic.
16:04So, and if you could just stand up, say your name,
16:07the company you're from, and then ask your question.
16:10That would be great. Thank you.
16:11Thanks. Roy Loke from PayPal.
16:14Thanks for the wonderful insights.
16:16I just was trying to tie this conversation back to the data
16:20that we were looking at prior to this.
16:23I guess what I'm struggling to reconcile is,
16:26I think this one fact that you shared,
16:28which was that outside of the China and the US,
16:32Asia in general has 2.6% of private investments in AI.
16:37And they use private investments as a proxy for where there's value,
16:42i.e., financial gain, financial benefit.
16:44And the 2.6% feels very off, feels unnaturally low,
16:49especially for such a large economy.
16:51So, I'm wondering what the insights are from the panel on what might be the cause.
16:54Is it that the value doesn't exist,
16:56or the barriers and the foundations to getting to that value are way too high
17:01that it doesn't justify the investments?
17:03But it just feels very counterintuitive to me.
17:04So, I'd love to get your thoughts. Thank you.
17:07Who wants to take that?
17:08Yeah, sure. I feel like I should.
17:10Yeah, you start with them.
17:12Thank you so much for the question.
17:13So, I think actually that 2.6, I wouldn't consider so low,
17:19considering that the US makes up over 60% of global private investment in AI,
17:24and China, which is also the second biggest AI player in the world,
17:28makes around 9%.
17:30So, 2.6% I wouldn't consider is that low.
17:35And within it, I think Singapore in 2023 was something like 0.7 billion Singaporean dollars,
17:45which I feel is quite high.
17:49And I think your point about AI private investment not being, I guess, the main metric,
17:57I think AI government spend is also something that we look at in the index,
18:02and that is very important when we look at how much AI has been invested,
18:07or how much has been invested on AI in the country.
18:11I don't know if I'm responding to this right, but...
18:14Maybe I'll try and do a little bit of squaring the circle here,
18:18which is one reason you see this, of course,
18:20and I'm speaking now more from my previous private sector angle,
18:24is that a lot of the investment numbers are probably more into the fundamental research
18:29and or compute side of it.
18:30And not just for Singapore, but a lot of Asia's approach to AI is actually about value.
18:36Where's the value?
18:37Now, you might say, well, that creates some risks in the long term
18:40if we are not the sources of fundamental research.
18:44Yes, but in the near term, in the short term,
18:47I think the cost of using AI and adopting AI is far lower than the cost of developing
18:52brand new large language models, etc.
18:54So, some of that number you are seeing in that.
18:56The other thing I say only partly tongue-in-cheek is I did work for a global bank
19:00with operations across from US to Asia.
19:04We just get more bang for the buck in Asia.
19:06That is a reality.
19:07I mean, we spend less compared to some of the more expensive centers.
19:11And that's not necessarily true for Singapore, but certainly for the region overall,
19:15you will get more value.
19:16So, I think the link between value and investment is probably better here,
19:20if I might say so, compared to...
19:21Yeah, there's also the question of whether you rent or bill.
19:24Yes, that's true.
19:25And a different kind of angle is I draw the analogy with legal services.
19:29So, one of the things that's been fundamental to Singapore's success
19:32is embrace of the rule of law.
19:34Now, legal services is only 0.55% of GDP.
19:37So, it's a tiny sliver of GDP, but it enables everything else
19:41because only with a trusted legal system that enables you to have finance,
19:45a banking sector, which is more like 14, 15% of GDP.
19:48AI is going to be the same.
19:49You need core investment, but to take advantage of it,
19:52that doesn't require you to have all of the data centers in Singapore or indeed in Asia.
19:58I would say that it's also a time dimension to a lot of these things that we're talking about.
20:03I mean, generative AI has maybe caught the world by storm for only about 20 months.
20:08I think many industries and businesses are seriously thinking about
20:12how to make their investment worth their time.
20:14And as we have seen, as it relates to, say, infrastructure,
20:17sometimes it does get cheaper as you wait.
20:21It's easier to buy as opposed to build.
20:25And I do think that companies and organizations do need to think about
20:29the kind of capacity that they have within the organization and outside of it
20:33to make that investment worth their while.
20:36I also would think that there is a certain dimension of, you know,
20:39getting over this hype cycle and making sure that we are building for the long haul,
20:43which is why I think some strategic patience in this arena is really, really critical
20:50because you have to make these investments not too different
20:53from when we were doing digitalization about 10 years ago.
20:56Many people thought that it would pass.
20:57Many people thought that, you know, I could sort of hold on and wait for it.
21:01But it is the businesses that do see the value and make those investments over time
21:07that will eventually see the benefits.
21:08So I would take that 2.6% with a slight pinch of salt
21:12because I think that's just a snapshot of where people are at today.
21:16I hope that in 5, 10 years time, that number will be significantly different.
21:20We've got time for one more question, if it's quick.
21:25Anyone got a question? Okay.
21:27Yes, one at the back.
21:31I'm Eugene from Tenstoren.
21:32So I have two questions.
21:35So firstly is many nations, right, they're actually investing in RISC-V
21:41It's an open source version of ARM so that they could really protect their supply chain.
21:45When they develop RISC-V, it's like an intellectual property to make their own chips.
21:50That's number one.
21:51So what's Singapore's strategy around that?
21:53And number two is, I believe, you know, AI compute is extremely important.
21:57But my question is, what alternatives are Singapore looking into
22:01because it's not really good to invest all AI and compute into a certain particular vendor?
22:08I would say that diversity is really important for us also on this particular piece of technology.
22:13Just to give some statistics, Singapore is the...
22:18I think we contribute 20% of semiconductor equipment and about 10% of global chip production.
22:25But that's not to say that we do all kinds of chips.
22:27We have strengthened some of them, like memory, for instance.
22:31And when it comes to AI, I think while there are certain players that are dominating,
22:36we know that it's still a space that will evolve very quickly as the technology evolves too.
22:41So I would say in response, the strategy here is to make sure that you have the right chips to do the right things.
22:48Our goal is to make sure that organisations all over here have access to what they need.
22:56And from a government standpoint, I think our investments will come across a range of options,
23:03from stuff that requires to be on-prem because of our security needs,
23:07but where there can be cost efficiencies and technological advantages,
23:12and being in a cloud, we'll also do that too.
23:14And then going back to the question of sustainability,
23:17how do we get all of this compute power for much less energy?
23:21I think that, again, would be one metric of deciding.
23:24Simon, quick-fire answer from you to Ant.
23:26Very briefly, the objective for a country like Singapore can't be to be the end.
23:30We're not going to be the dominant player, but we want to be an indispensable partner.
23:33So Singapore's history can be summed up in three sentences.
23:3719th century, a port for goods.
23:3820th century, a port for services.
23:4021st century, a port for data and ideas.
23:43So we've got to be part of that information chain, that flow,
23:46which requires us to have diverse partners and diverse interests.
23:51I think a last one from me is I think what will be key is not just investing in high-performance compute infrastructure,
23:57but also making sure that that is available to industry and researchers to use.
24:03And in the same way, we will need public data set platforms with data that can be made available,
24:09again, for both industry and researchers to train their AI systems.
24:14Serena, Simon, Shameek, and Andrea, thank you so much for your time.