- 3 months ago
"As global powers compete in emerging technologies, regulation has become a strategic tool, shaping innovation, competitiveness, and global influence. From Europe’s precautionary approach to the U.S.’s market-driven model and China’s state-led system, competing frameworks are redefining the rules.
How do these choices impact innovation, investment, and the ability of companies to scale globally? This session explores the trade-offs between control, competitiveness, and openness, and how businesses can navigate an increasingly fragmented landscape. In a world of competing standards and systems, is convergence still possible or are we entering a new era of technological divide?"
How do these choices impact innovation, investment, and the ability of companies to scale globally? This session explores the trade-offs between control, competitiveness, and openness, and how businesses can navigate an increasingly fragmented landscape. In a world of competing standards and systems, is convergence still possible or are we entering a new era of technological divide?"
Category
🤖
TechTranscript
00:01.
00:01.
00:01.
00:01.
00:01.
00:01.
00:02.
00:32Good afternoon. Great to have you with us today. I'm joined by four wonderful panelists. I want to introduce them
00:41to you very quickly. The title of our session today is The Great Tech Race, Regulating Innovation in a Multipolar
00:48World.
00:48And with me I have, I'm going to go in my order, not the order they're sitting, Julie Huguette, Director
00:55of Mission French Tech, Ning Ning, VP of International Business at ZeroOne.ai, Sanjay Puri, President of Knowledge Networks,
01:05Norman Van Hacke, CEO of Cradle Fund. Very excited to have you all with me today. So I wanted to,
01:11if you will allow me, just a few moments to do a bit of an opening,
01:17to set sort of our stage for the conversation today.
01:20So there's this stereotypical story you've probably heard about. I'm not going to name it exactly, but it's about the
01:28global tech race. The different parts of the world each play a fixed role.
01:34Some parts are innovating, some are making the rules, and others are following them, or replicating them. This is a
01:42very out-of-date story, and I'd like for us to set it aside for good, especially in the context
01:49of this discussion.
01:50Because the real picture is actually quite interesting. Regulation has stopped being the thing that happens after innovation. In fact,
01:58it needs to happen in concert.
02:01It's become a strategic tool in its own right, shaping who gets to build, who gets to scale, and whose
02:08standards the rest of the world ends up living with.
02:11And different regions are making different bets. Some are leaning towards protecting data and people. Others are towards more market
02:18-driven growth. And some are more towards state coordination.
02:22And each is actually a strategic choice. And underneath that, there's the competition, and then there's the question that matters
02:30just as much.
02:31When the stack can fragment chips, cloud, data, models, all of the above, does sovereignty mean pulling apart?
02:41And we think of sovereignty, and sometimes we relate it to nationalism, and I don't want us to do that
02:47today. Sovereignty is not just flags and borders.
02:50It's so much more, such a much more nuanced conversation. So, thank you for allowing me to wax poetic for
02:57a minute.
02:58I wanted to dive right in. You know, this is a very interesting conversation. So, I'll begin with you, Julie.
03:07Europe often talks about regulation, but what does Europe need to do today to actually create the next generation of
03:15sort of these AI champions?
03:18Good afternoon, everyone. Thank you for this question. Maybe a few words about what we do at the French Tech
03:23Mission before I answer.
03:24French Tech Mission is a public organization, so we support the growth of French startups in France and abroad.
03:32We work with communities of entrepreneurs called the French Tech Communities everywhere in the world.
03:38And our common goal is to create our own French Tech Champions.
03:42So, what do we need to do that? We need three things. The first thing is, of course, investment.
03:47European startups have raised €60 billion last year in 2025. I think it's already a good signal.
03:57The startup AmiLab from YAM Lequin has raised €1 billion earlier this year.
04:04So, it's a good example of what we can do now from Europe.
04:07The second thing is public and private procurement. To give you another example, the public procurement in Europe is €800
04:17billion.
04:28When you know that they start the company with $22 billion of public procurement, you understand what we can do
04:42in Europe.
04:42We can also create our own champions like SpaceX only with the public procurement.
04:47And if you have the private procurement, of course, we have a good way to accelerate our startups.
04:55And then finally, the third thing is that we need to go global from day one.
05:00Why? Because the French market is still small. European market is fragmented.
05:07So, we need to accelerate. We need to open a lot of different countries very quickly.
05:13That's what some of our startups did. I'm thinking about Swild, for example,
05:18who started from day one in Brazil and now they are leaders in this country.
05:23Blablacar or Brut in India, and now they are leaders as well.
05:28So, I think it's also something we need to develop more if we want to create global leaders from Europe.
05:34Julie, thank you for painting sort of the picture, not just in France, but also in Europe.
05:38I was telling her earlier how excited I am about the panel because it's really a multicultural panel.
05:44We're going to hit so many different regions, although not all yet.
05:48So, Ning, I wanted to go to you next from this foundation model company that you're deploying in the real
05:55world.
05:55What does the global AI map actually look like to you?
05:59Is it just one race because we're talking about the great tech race or is it multiple?
06:05Thank you. So, yes, we started with the models, but now I spend most of my time traveling around
06:13and working with governments and enterprises who is trying to really put AI into their real operations,
06:20whether it's a mining site, whether it's an agricultural farm, whether it's a factories or public services.
06:24So, I think from the point of the, the angle of the deployment is a little bit different.
06:29So, I think a few years ago, the model was the price, right?
06:34And then it takes a lot of money to train the model and from scratch.
06:38And then that was the mode.
06:40But today, we're seeing increasingly the model is moving from the price to a component.
06:45It's a very important component, but it's not the only place where you'll win.
06:49So, I think the layer, I think gradually you see is very, more and more important to, is a layer,
06:56see between the model and the real enterprise problem solving and the value creation.
07:01So, it's about building the ontology, it's about building the context, the institutional memories
07:06and the workflows and agents, right? All of that.
07:09And then, and more importantly, how you're going to surround AI in the center,
07:14you're going to redesign the products and services.
07:18Eventually, you design organizations.
07:21So, I think there will be two races.
07:23One is the frontier model race.
07:25So, that is very expensive, very concentrated, and few and few players are staying in the race.
07:31And the other race is the deployment race.
07:34I think for many of the companies and governments, it is open to every governments and enterprise for this race.
07:43I think that race is going to decide the next decade.
07:46So, if I'm trying to draw a map on the, I mean, for this, I'm not going to draw a
07:52line between the East and West
07:53or join a few big circles on the map.
07:56I think I'm going to draw a line.
07:59On one side, that would be the companies and organizations going to moving AI native.
08:05And on the other side, it's just bolting the new tools onto old habits.
08:11I think that choice matters most.
08:13I find that fascinating, the two tech races.
08:17Which one is the one that, you said the one that's going to decide is the second one,
08:21but which one is the one we're on right now, the frontiers?
08:25I think we are both.
08:27I mean, we are both.
08:28The frontier model race, the other one is the deployment of the models and technologies.
08:34So, Sanjay, you're sort of the next one on the hot seat.
08:37I wanted to go to you next.
08:39And we've talked about this before, actually, on a stage.
08:43From the governance point of view, what is the single biggest thing policymakers are getting wrong right now
08:49in this great tech race conversation?
08:53Well, thanks, Anna, for having us.
08:55And great panel to be here with everybody.
08:59So, I would say I'll just give you two.
09:02Since you asked me one, I'll give you a bonus one.
09:04I think the first thing that regulators are getting wrong is they're trying to regulate the model.
09:09But the problem lies actually in the outcomes of the model.
09:15So, when you look at what the models do, it is not the model, but let's say there is somebody
09:22who's been denied a job or a medical issue that has not been done.
09:28That's where the problem lies.
09:31But if we are going to be out there, you know, saying, okay, let's regulate this model, that's not going
09:38to work.
09:39Because from a governance standpoint, you can't do that.
09:42And I think the second thing that is there is speed.
09:48Legislation works at its own pace.
09:51Committees, hearings, and things of that nature that go on for years.
09:55AI is moving at an incredibly fast speed.
09:58You're having frontier models, as he talked about, every few months, sometimes in a few weeks that happen.
10:06So, legislators are legislating something that in many cases don't even exist anymore.
10:12Or they are legislating models that were there three generations ago.
10:18So, they need to be working not just on coming up with fast legislation, but on legislation that is going
10:24to be outcome-based.
10:26And that is adaptable to two, three generations forward.
10:30And then have the ability and the tools to be able to regulate it.
10:34Because you have regulators who themselves, unfortunately, don't have the background and the skills and the expertise.
10:41And many countries, unfortunately, don't have that kind of a skill or the background or the expertise to say,
10:48can I look into this frontier model as my colleague just talked about in the foundation model?
10:53Can somebody look into those foundation models and say, is my rule being applied or not?
11:00There's no way.
11:02Because they don't have the expertise to do that.
11:05So, those are the two major things.
11:07So, let's regulate the outcomes and then figure out a way that we can keep pace with this technology.
11:14Because you could, I mean, think about it, the EU AI Act has been delayed.
11:19I mean, there are so many things that have happened since, well, just even in the past few months with
11:24mythos and things that have come in.
11:26So, we kind of need to be in the game from that standpoint, Anu.
11:31And a quick follow-up for you.
11:33When you say be in the game, do now, is now a new competency and skill of a bureaucrat or
11:40a politician or a policy maker to also have technical skills?
11:45Well, I wouldn't say it is technical skills, but it is more an awareness of what is happening.
11:51I mean, if you look in the US Congress, which I'm in DC and so are you, they're probably on
11:59my two hands out of 435 members and out of 100 senators that have that ability to really understand the
12:08consequences of what this technology and the transformative nature of this technology is.
12:13That's not to blame them.
12:15They are dealing with Ukraine.
12:17They are dealing with Gaza.
12:18They are dealing with border security, et cetera.
12:20But this needs the attention that it deserves.
12:24And I think our legislators all deserve to kind of take this and take this very seriously.
12:31Thank you for that.
12:32And Norman, last but not least, I wanted to go to you.
12:36You know, I've been hearing, especially the last couple of days, you know, Europe, China, the US are dominating sort
12:43of the airwaves when it comes to talking about frontier tech and AI.
12:47But you come from such an interesting background from Southeast Asia startup ecosystem.
12:53What does this race look like for a founder who isn't in any of these three top markets that I
12:59was talking about?
13:00Yeah.
13:02First off, Cradle, we are a government agency of Malaysia.
13:06So we provide support for early stage setups in terms of funding, capacity development, market access, and also going through
13:16scale-ups.
13:18The one key thing, because what's interesting is that over the years, especially in Southeast Asia as well, there's a
13:23lot of interest of wanting to be there with the rest of the world.
13:27But not everybody knows that our upbringing, every single country within the Southeast Asia community brought up differently.
13:36So we have different cultures, of course, different regulations, different populations, and different interests as well, different governance systems.
13:46But the drive is quite similar, per se, because we see how the rest of the world is advancing.
13:55And with the advent of technologies such as AI, and the access to those technologies should be quite easy for
14:02a small country who do not have the depth of capital like the West or China or in Europe as
14:09well, right?
14:10So it provides an opportunity for innovators and founders within the region to come up with something that is not
14:18just to serve locally within the region, but can also go global.
14:24So case in point, we actually, back in our earlier days, we funded a company called Grab.
14:31So it was at the point when Uber started to come into the region, and soon enough, Grab actually took
14:39over, bought over all of Uber within Southeast Asia, and became the region's first Decacorn.
14:45So they got listed in NASDAQ, right?
14:47So it's not to say that there aren't opportunities within the region.
14:51It's just that you have to navigate yourself a little bit more creatively, per se.
14:56But within the confines of the governments, the rules and regulations.
15:02Yeah.
15:02And so interesting that maybe I have a quick, you know, follow up for you.
15:06It's when we talk about the great tech race, we're so hyper focused on AI right now.
15:11But there's so many other frontier technologies, quantum being another one that's actually taken quite a bit of space here.
15:19Yeah, I mean, I would like to take the last word in your sentence actually, space.
15:23There's actually quite a lot of interest in the region as well to start to develop space technology ecosystem.
15:30So we're seeing, I mean, there's no launch pads or anything like that over there.
15:35I mean, the companies over in the region are actually launching their satellite via SpaceX or other service providers outside.
15:42But then demand is increasing.
15:46And the access to those technology is getting lower and lower that more and more corporates within the region are
15:52actually thinking about why don't we do this ourselves?
15:55Why don't we look into this?
15:57Because this is kind of like the next frontier of technology while everyone is still trying to figure it out.
16:02Why don't we take this opportunity to look into it and start building our own ecosystem to serve the region.
16:08But of course, we still keep ourselves very open working with everyone else outside the region because we're not the
16:14main experts of things.
16:15So, yes, I mean, we build ourselves in a way that we do not restrict to who is it that
16:23we're talking to when it comes to innovation and, you know, coming up with the new technologies.
16:28Yeah, I imagine it, you know, not to, you know, give you my nerdy credentials, but sort of like Star
16:34Trek in an opportunity where this kind of competition helps all ships rise and kind of get us to the
16:40next level.
16:41I know, Julie, you also mentioned that there is a European company that is directly doing what would SpaceX has
16:51done or it's competing directly with that.
16:53But for you, I had a different question. Do you believe Europe will still be sort of a tech builder
16:59or will it mainly become a technology user?
17:02And I know it's a bit of a provocative question, but I've heard so much in the last few days
17:07about tech sovereignty here.
17:08So I wanted to hear from you.
17:11Of course, we need both. We need to build our own solution.
17:13It's key for our sovereignty and we need to adopt this solution very quickly as well.
17:18Since we organized last year the AI Summit in Paris, I realized that our AI ecosystem accelerates very quickly.
17:28Now Paris ranks third worldwide in AI after San Francisco and New York.
17:34And we have almost 2,000 AI startups in the whole AI chain value.
17:43It represents almost 50,000 people who are working in this field.
17:48So we have talent as well.
17:51What is interesting is that we cover the whole AI chain value from semiconductors to infrastructures to fundamental models like
18:01AmiLab, Mistral, H Company for computer use.
18:06And we also have obviously a lot of AI applications in different fields.
18:11The question we are asking ourselves now is we have the solution.
18:16We are building them.
18:18But what kind of ROI do we really want?
18:22And when we are talking about adoption, the question is more what do we want to do with that?
18:30We can have a social impact, a financial impact, and we are trying to think about that.
18:39It's actually quite fascinating what's happening in the European ecosystem.
18:43Thank you for that.
18:44So Ning, I wanted to go to you next because you've said fragmentation is happening across chips, data.
18:51It's happening across governance and data governance.
18:54But that sovereignty doesn't necessarily need to mean isolation.
18:58And there is one more ingredient to all of this story.
19:01And that's human capital and talent.
19:03And that's why we can't be isolated, right?
19:06So how do these open source models that you talk about actually can let a country that doesn't have the
19:11workforce and the human capital to build something that's truly sovereign?
19:16Yeah, so when talking about the sovereign AI, people tend to think it's building up, I mean, a large language
19:22model from scratch, spend billions of dollars, and then including an army of top-tier researchers.
19:28But I think, I mean, that probably true one or two years ago.
19:31But gradually we think AI, I mean, you tend to think AI is now, sovereign AI is a system, not
19:37just a model.
19:38And that system, a large part of the sovereignty can be contained in the layer above the models, not just
19:45the model itself, right?
19:47And then the layer, I mean, as I mentioned just now, including all the context, the guardrails, the governance, the
19:54workflows, I mean, all the institutional memories.
19:57So I think building the layer doesn't require a top- of the frontier research lab.
20:03It can be just a national-focused program, contain a group of people that really knows the country, know the
20:09language, knows the policies, and also a few AI experts.
20:13And back to the model itself, I think open source really lowers the barrier.
20:18So I think training, I mean, continue training a model, open source model, or fight to the model, maybe just
20:26a few million dollar project.
20:28And then you can use the fight tuning to customize a model to enhance the local languages, and also make
20:35sure the model is alive with the cultural values and local regulations.
20:39So I think that also doesn't require, I mean, frontier research lab.
20:43So, and you don't necessarily do it all along, you can just find a partner, not just do the tuning
20:49for you, but also do the knowledge transfer, so that you can reduce the dependency.
20:54And using open model probably bring another advantage that over the closed source models is the control, right?
21:01It runs on your machine, in your environment, under your law, and then your data never goes out, and then,
21:07I mean, for all the things you have done,
21:10I mean, you can send your people to do the inspection, and then that's especially important for those sectors that
21:17is data sensitive.
21:18So I think the sovereignty is the content more than just the model layer, and then the open source model
21:24already lowered the barrier,
21:25make it achievable at a lower cost, and then you can also find a partner to do it, I mean,
21:30and also do the knowledge transfer.
21:33And so, really, it's not that there's no race unless there is a little bit of cooperation, because no country
21:41or no region can do it alone.
21:43You know, I wanted to go back to you, Sanjay.
21:46You've described sort of this cod red urgency among policymakers.
21:50You talked about they're dealing with polycrisis, they're not dealing with just one thing, they're dealing with multiple things.
21:56And you've warned that some private tech companies may now hold more power than countries.
22:01Your words, not mine.
22:03If corporate decisions are becoming geopolitical, what does responsible governance look like in this kind of a situation?
22:11Well, we are also in the red stage.
22:13Remember, it's code red in the red stage.
22:16So, I mean, the reality is that you have a few companies that control basically the, you know, the compute,
22:26the chips,
22:26and the cloud on which all these applications run.
22:30And today, if some company in California decides to do a price increase, it can append an industry in Lagos
22:37or Jakarta.
22:38And that's not market.
22:40That's basically total domination.
22:43And so, when these companies put up a product strategy, that's basically foreign policy today.
22:50In today's environment, and look what's happened in the past month or so, where we have a company that really
22:58decided that it would not release its model,
23:00one of the leading frontier model, to the rest of the world.
23:03And it picked and, you know, decided which countries, which companies it would work with.
23:08And that's unbelievable power that you've given to a company.
23:15So, what should we do about this?
23:18Well, there are three basic things, Anna, that I think one needs to start looking at.
23:23It is, you need to govern the choke points, which is the chips, you know, the compute, and the cloud.
23:29You need to get more transparency from these providers.
23:32Basically, to really get true transparency, not what self-attestation that they provide.
23:39And third is something he talked about.
23:41Build alternatives, open-weight, open-source solutions need to be provided.
23:48And finally, if you are a country, small country, that is trying to fine some of these companies,
23:55well, their market cap might be more than your GDP.
23:58So, your fines are not going to work.
24:01What will work is become a customer that they cannot afford to lose.
24:05And that, I think, is the trick.
24:08Because what has worked for EU, most people don't realize, is not the fines.
24:14It is the market access that people want to this big market.
24:19And that's what has worked for them.
24:21And I think that's what companies need to do.
24:25That's what countries need to do.
24:27And there has also been talk about these mid-market or mid-tier countries coming together.
24:34Because procurement is one of the most underutilized regulatory weapon that people have,
24:40which has never been used.
24:42And these companies, which live quarter by quarter by quarter, understand procurement and understand sales.
24:50So, to me, that would be the solution.
24:53It's an important and very powerful lever that you just described.
24:58You know, Norman, I wanted to get back to you.
25:00In sectors like banking, you've pointed to the need to control AI adoption for safety without crippling innovation, of course,
25:09or startups.
25:10Where exactly should the guardrail actually sit?
25:13And who should set it?
25:15Is it regulators?
25:16Is it the market?
25:18Or the ecosystem itself?
25:20And that's a much broader sort of space.
25:23It's a very interesting question.
25:24Because in Malaysia, in terms of the financial systems is quite, well, one could see it as being highly regulated.
25:34But another way as well, there's regulations that allow innovation.
25:38So, in Malaysia, we actually are allowing tokenized asset ownership and trading.
25:46So, there's not a lot of country that's actually gone to that level just yet.
25:50And already, we're looking at ways on how to further liberalize investments or ownership of assets beyond just the physical
26:01asset.
26:01So, with this, it opens up plenty more opportunities for new innovation.
26:07So, we're seeing new startups coming in at every level of every financial transaction.
26:14There's always some innovation at every stage.
26:18And the latest, of course, everything is AI infused.
26:22And then, you have this thing where agents are transacting with agents.
26:26How do you control that, right?
26:28So, the approach that the country is looking at is that not to put in too much regulation that it
26:35would impede innovation.
26:37Because innovation is still very critical.
26:40And you don't want to put in too much regulation that you'll fall behind.
26:46Because Southeast Asia, Malaysia, we're already, you know, we're not as advanced as all the other countries in the world.
26:52But this is our opportunity to actually get ourselves up there.
26:56So, we can't be setting out regulations that impede innovation.
27:02I mean, that's the real key thing here.
27:04But having said that, security, you know, making sure that we don't expose ourselves too much, you know, in terms
27:11of capital risk or any fraud risk and things like that.
27:15So, that is still very, very much in our attention.
27:18And I do have to say that, I mean, our Securities Commission, when you look at all these things, they
27:22are quite advanced at that.
27:23And I'm very sure Singapore is as advanced as well.
27:25And every other countries within the Southeast Asian nations as well are really, really looking into this as well.
27:30Because, like I said, those kinds of technology are quite accessible.
27:34And we're seeing a lot of innovations coming out.
27:36And the governments within the region are actually being very supportive of that.
27:40Not just about regulation, but we also provide support, funding, access, opportunities to work.
27:46And actually develop, we do sandboxes for all this, right?
27:51So, to allow the development and implementation of these new technologies.
27:56Well, and, you know, that's always the big question we're all trying to answer.
28:01You know, can regulation and innovation kind of go hand in hand in a way that one is not lopsided
28:07and it doesn't stifle progress?
28:10And, you know, at the end of the day, the big promise of technology and AI was that not only
28:15was it going to democratize access and opportunity, but it was going to solve big things like big healthcare issues,
28:24cancer, all of those things.
28:25We can't go through discoveries if we are...
28:29Don't try it.
28:29Exactly.
28:30Exactly.
28:32You know, I had a question.
28:33I wanted to ask all of you sort of a similar question, but I wanted, Julie, to go one more
28:38time back to you.
28:40The real question may not be only or quite who builds AI, but who deploys it.
28:47And that's something Ning talked about.
28:51And he said these two races are kind of parallel right now.
28:55To you, the question is, how can Europe accelerate AI adoption at this crucial time to just get you to
29:03that competitive advantage?
29:05Yeah, I think adoption is key for large corporations, but for SMBs as well.
29:11First, we need to add the question, what kind of ROI do we want?
29:16For example, hospitals can use AI to reduce waiting time in emergency.
29:21We have some companies like Ikemia who use AI to accelerate drug discovery.
29:27So what do we want to do with AI?
29:29And when we choose, what we want to do is how we can find the right solution to do it.
29:35How we can, what kind of startups or innovation, what can we accelerate the adoption.
29:45And that's why in France first, we launched a program three years ago now called I Choose French Tech.
29:52So the objective was to accelerate the relationship between large corporations, public organizations, and startups.
30:02So it works.
30:04We got a very good result.
30:06The relationship between large corporations and startups have been multiplied by 10 the last four years.
30:13That's why in this stage, two days ago, we launched a European initiative, a similar initiative called Choose European Tech.
30:22With our Minister, Benjamin Haddad, Minister for Europe, and Anne Le Hénonf, Minister for AI and Digital Affairs.
30:30And with some other countries, eight countries like Germany, like Serbia, like Belgium, we decided to accelerate the adoption, the
30:41European adoption.
30:43So a lot of large corporations, public institutions are joining us in this program.
30:49The objective is to help them to find the right solution and to create more bridges between these startups and
30:59the companies who need to adapt.
31:03I've heard from all of you, from your vantage points, the regional and the countries that you're representing or are
31:11located in.
31:11And I'd like to open it up, get a bit of an aperture here to more of the rest of
31:19the world.
31:20Just walking in, I was speaking to two young entrepreneurs from Africa, such a young continent with a real demographic
31:29dividend.
31:31And they're here and they're very excited and they're talking about their companies.
31:34So the question that I wanted to ask to all of you, maybe we can start from Sanjay and come
31:38down.
31:42What about protecting languages and cultures, including oral-only languages in the Global South and Africa?
31:50That's a recurring worry.
31:52Is that a problem the market will kind of solve by itself, will trust it that it will do?
31:57Or is that one that we need more deliberate investment and policy to add and inject into it?
32:03And we can all maybe take a turn and answer that.
32:08So that's a very interesting question.
32:11We have about 7,000 languages in the world, out of which I think, I could be wrong, about 2
32:17,000 are based in Africa.
32:19But if you look at the prominent languages in AI, it's about 100.
32:24And 50% of the languages in AI are English based.
32:29So if you think market is going to solve this problem, I don't think so.
32:34Because the market basically focuses on what's going to give it the best clicks, the best returns and things of
32:42that nature.
32:42And if you look at Africa's, you know, it's maybe one-tenth of 1% is what is on the,
32:51in AI right now.
32:53So my perspective is languages like any other utility.
32:58It's like a library.
33:00We need to protect libraries.
33:03We need to protect water.
33:04Because if languages go with cultures, they go with history and things of that nature.
33:11I mean, from what I understand, even you have some, you know, Star Trek based languages on AI, but you
33:22don't have a lot of African languages out there.
33:25But if we put some attention to it and think of it as a public utility, that's going to really
33:32help create and protect cultures.
33:36I think that might be the solution, not just leave it to market, Ana.
33:41What are you thinking?
33:42Yeah, so I totally agree with Sanjay.
33:44I think the market itself doesn't really solve the problem.
33:47Because a language is spelled by tens of millions of people.
33:51It's going to, it will be served.
33:52But I think a language, it just speak by tens of thousands of people.
33:57It doesn't actually have the commercial value for doing that.
34:01But I think the cost of doing nothing, the cost is real.
34:06Because as AI has become more and more the way we learn and we work.
34:10So if the AI doesn't speak that language, I think it's going to disappear in the next generation, right?
34:17So, but the good news is that because intelligence becomes so cheap, right?
34:22So saving the language is not a mega super project, right?
34:26So I think, for example, basically, if there's already written, we just collect the corpus and then, I mean, we
34:32can use the continued training of the fine tuning to kind of enhance the language of the capability.
34:37And for those spoken only languages, I think the way we do probably is to build a better ASR, right?
34:44Or a better TTS.
34:46So I think with that, I think it's really the technology is there.
34:52I think the key decision is that who is going to pay the bill.
34:57So you made some great points.
34:58And I think whoever's listening is a young entrepreneur.
35:01That's a business idea right there.
35:03A free one.
35:05Julia Norman, would you like to also comment?
35:08No, I think everything was said.
35:11Maybe a few months ago, I was in India for the AI summit.
35:14And it was the first thing they were talking about because there are so many different languages.
35:21And it's a real problem, especially in the hospital.
35:24So, yeah, they're working on different solutions with local AI and translation.
35:28And, yeah, very interesting.
35:31Norman.
35:32Norman.
35:32Well, yeah.
35:33Coming from Southeast Asia, 11 countries, hundreds of languages, thousands of dialects.
35:40So it's like the permutations to that is actually quite a lot.
35:43So then how do you decide which to save?
35:48So, like, is that a question even?
35:50I don't know, right?
35:51So extending from that, I mean, we were just speaking just before this, right?
35:54We were talking about the disappearing art of calligraphy.
35:57So it's like writing, right?
35:59So everyone types on keyboards now.
36:00So I can't even write on a piece of paper.
36:03It will just come out squiggly, right?
36:05So same thing with languages as well.
36:07So if everybody is just learning how to prompt, would they know how to speak properly?
36:12Would they know how to think, you know, how to get information?
36:16Because prompting is so easy.
36:18So the art is just about prompting, you know?
36:22But I think you can't be too fearful of that.
36:26Because if you do that, then you won't progress.
36:28You might as well regress, right?
36:30So the best thing to do is actually, okay, this is what you have at the moon.
36:34How do you make yourself better with the technologies that you have today?
36:38So the part about, you know, saving the language, saving the art, the cultures and everything,
36:43I think that's a responsibility for everybody, right?
36:46So it's not just the market, not just the government.
36:48I mean, that's also even probably a business opportunity as well.
36:53You know, we never thought about digitalizing art and trading it and things like that, right?
36:57So that can happen.
36:58So I'm sure that it will always be there.
37:02And I'm sure there will always be advancements.
37:04And I'm sure there will always be someone somewhere that will look into preserving languages,
37:10art and techniques and everything like that.
37:12Because who knows, one day a different type of language may just be the source for a future solution.
37:19We don't even know that.
37:20Maybe writing will help you avoid hard complications in the future.
37:24And everybody will start writing again.
37:26So we don't know that, right?
37:27But my point here is that don't be afraid of it.
37:30You know, just embrace it.
37:32But, you know, within the right constraint.
37:36Yeah.
37:36You know, at first when you were all describing the languages, the breadth and depth of the, you know, solving
37:44this,
37:44you know, to me it sounded daunting.
37:46But then I said, everything is possible with AI.
37:50This shouldn't be very difficult.
37:51I think the question should be, is it worth it for us to save it from a civilizational point of
37:56view?
37:57And going back to what Sanjay said, sort of like libraries.
38:00This will be the new thing that we need to save.
38:03We all agree, right?
38:04That we want to save this.
38:06Thank you so much for these such insightful comments.
38:10I wanted to kind of, you know, we're coming towards the end.
38:13But I had a question for all of you again.
38:17In this world where competing standards, is convergence and collaboration even possible?
38:24I came from a panel earlier that that didn't sound like it could happen.
38:30And I feel like this is a whole different vibe here.
38:34Maybe this is a question I'd like to ask all of you.
38:36We can go right back to Sanjay and do the whole round again.
38:41I think convergence is not happening.
38:44And I think we should not focus on that.
38:47Because we'll waste a lot of time in there.
38:49But focus on standards.
38:51I think that's what the real issue for us is.
38:55The plumbing, you know, content provenance, evaluation standards.
39:01The important things that are there.
39:03I think that's what we need to really focus in on.
39:08I mean, I look at, you know, people flying planes.
39:12Planes, I mean, he just traveled from, I don't know, Melbourne or stuff.
39:17Going across the world.
39:20And the countries who agree on nothing, agree on aviation standards.
39:25Because they're worried that planes will fall from the sky into the ground.
39:31AI is traveling borders too.
39:34Why can't we at least agree on some basic standards?
39:37So that's, to me, is where I really think it is.
39:42Because if you make, you know, values the price of admission.
39:47For some joint thing, it's never going to happen.
39:50Because U.S. is going to say we are, you know, free market or innovation.
39:55Europe is precautionary, et cetera.
39:58And we will get hung up on that.
40:01But let's at least talk about some basic standards.
40:04Like we do on some critical important things.
40:07Whether it's aviation, nuclear, whatever it is.
40:09Because I think we can bring some adults into the, at least we hope we can bring some adults into
40:14the room to come up with those things.
40:16So focus on standards rather than values.
40:19That would be my point.
40:20That's so practical of you.
40:22Thank you, Sanjay.
40:22Ning?
40:23I think, yeah, I agree that some better standards are needed.
40:26But I think every country might have a different AI rule book.
40:29I think that is a sovereign choice.
40:31Because every society is going to price different risks differently.
40:35Right?
40:36For example, the risk of being falling behind, the risk of being the privacy infringements.
40:41I mean, the risk of being social instabilities.
40:45Things like that.
40:46So I think it's very difficult to getting a universal AI law and then agree by United States, China, and
40:54Europe.
40:54I think that's difficult.
40:55But I think underneath the zoo, the zoo is going to be split and then keep splitting.
41:01But underneath the zoo, something is going to be shared.
41:05For example, the science, the engineering practice, they are going to be converged.
41:08Because I think the paper is still being published.
41:12And then, I mean, the open source models are still being published.
41:16So as the agent frameworks, things like that.
41:19So those things travel.
41:21I think, for example, the AI banking, the digital banking practice in Singapore can apply to the AI practice in
41:28Sao Paulo.
41:29And then the AI practice in using AI to control the robots in the autonomous factory in Shanghai
41:35It may be something useful for the Paris factory as well, right?
41:40So those things, I think, are going to be converging at some point.
41:44But, I mean, as we say, I think we are not going to have the same playbook.
41:50But I think a few things I think we can cooperate on.
41:53One thing is the AI safety testing.
41:55We do not want to see the catastrophic event caused by AI, right?
42:00That's the first thing.
42:02And second, I think, further strengthen the open source comments.
42:07The third one is the language and cultures as we speak just now.
42:11Yeah, yeah.
42:12Julie?
42:12Julie?
42:14Julie?
42:16Yes, sure.
42:17I think we can build something all together.
42:21We have to, we need to have a strategy.
42:23I mean, for example, I was in Japan and I saw that a researcher was working with Japan researchers
42:30on the future of quantum computing, robotics, AI.
42:34Each country are different.
42:36And I'm pretty sure that we can find how one plus one equals three with every different country.
42:45So, yeah.
42:46I love that one plus one plus equals three.
42:49Norman, do you want to chime in?
42:51I'm a bit optimistic.
42:52One plus one equals five, probably.
42:54But, yeah.
42:55Anyway, standards, definitely.
42:58I want to add on to that, infrastructure.
43:01So, I mean, like I said, every country in the world, I mean, especially in Southeast
43:05Asia, we have different upbringing.
43:07So, where we are today, probably not the same infrastructure-wise compared to other
43:10countries all over the world.
43:12But having said that, the uptake of technology is probably quite high up there as well.
43:18So, if you look at Southeast Asia, the internet penetration is nearly 100%.
43:22Mobile penetration is more than 100%.
43:25So, the market there is ready for new technologies and they embrace it.
43:29You know, back home, we pay with QR codes.
43:32We don't take out our card.
43:34We don't take out our cash anymore.
43:36I don't go to ATM anymore.
43:37You know, in China, I think they did this even earlier, way earlier.
43:41But, I mean, we're embracing and we're adopting and it's very much imprinted in our daily lives
43:47already.
43:47And we are very much ready for what's next.
43:51And, of course, when we step out of our homes and suddenly it's a different system out there.
43:57So, I can't, oh, I have to download a new app to be able to, you know, get an Uber
44:01or get a ride or even to pay.
44:04So, you know, it'd be great, you know, if we can have like one standard or something that can really
44:09be global, universal that we can apply.
44:13As well as set of rules as well that can apply for everyone to really level the playing field so
44:18that, you know, when it comes to innovating, I mean,
44:21it will never stop a founder from Malaysia to talk to a founder from France to a founder from the
44:27States or in Africa
44:29and come together and create something that can actually benefit everyone.
44:33What a great closing to all of this.
44:36And I'm reminded a former U.S. president once said you can have the 60-40 rule.
44:43The 60-40 is that you will definitely disagree 60% of the time on most issues with people.
44:50But there is that 40% and you got to start with what you agree with and maybe that's where
44:55we start with standards.
44:56It can be done.
44:57Thank you so much to my wonderful panel and to all of you for your attention.
Comments