Panel Discussion AI in Digital Public Infrastructure (DPI) India AI Impact Summit

economy. Saibal Chakraborty, Managing Director and Senior Partner, Boston Consulting Group. The moderator, C .V. Madhukar, Chief Executive Officer of CoDevelop. And Mr. Sangbi Kim, Vice President for Digital, World Bank, who will be joining us in a bit. Thank you. right I will let the moderator take it forward

thank you so much thank you Arjun and thank you all to the panelists the last session of the last day is a bit challenging so we will try and keep this focused and at the end if you have any questions we have time it would be great to have any questions if you want to have any discussions I think we have heard a lot about AI and DPI in the last four days I don’t want to belabor the point I think what we can celebrate for sure in India is that in terms of DPI and the thinking that we take to the world and to our own problems is amongst the best in the world.

And as we look at the journey on AI, which is just beginning for most of the world, what I see is if I look at the US, for instance, there is one spectrum of conversation, which is AGI and beyond. And on the other end, there is a despondency and worry and concern about jobs and privacy and a whole bunch of other important considerations. I think where India is, is somewhat different. India is saying, look, there will be a Chinese model, there will be an American model, there will be a whole bunch of other innovations that are going on. What do we do to embrace and use all of this for our benefit? I think that optimism and sense of can do and must do, I think is very exciting, and I think it’s been palpable in the last four or five days.

So to discuss the power of AI and DPI and AI in DPI, we have a wonderful panel as Arjun has introduced already. I will start with Dr. Hans, if you don’t mind. I think you represent a national government and you are living through these choices on a day -to -day basis. As a pragmatic government that has to think about sovereignty, inclusion, safeguards, what are the main considerations that are on top of your mind?

Thanks for the question. You’re right, I think those three words are very key. When you’re talking from a government perspective and therefore national, inclusion, integrity or the safety of the citizens. and also safeguards. I’ll just run through these three very quickly. Why inclusion? When it’s government and you introduce a new capability, you have to be sure that that capability does not increase divides, that it actually reduces divides. So you have to be very sensitive on the inclusion angle. And, for instance, AI, together with DPI, can stretch inclusion through its voice -first capabilities, we talked about cloud -first, API -first, et cetera. Now we can really seriously talk about voice -first and also the translation capabilities.

And accessibility has to be also broadened in terms of multi -modality and also, where necessary, include a human in the loop in the service delivery cycle as well. So some of the inclusion dimensions. Then more to integrity, which is a tougher one, And I think that’s a really good point. And I think that’s a really good point. An important point when you’re looking at integrity is to start from the premise that AI will not redefine DPI. AI would, or at least where we look at it from now, maybe I’ll be wrong in six months from today, but the DPI foundations must be in place first. DPI should be mature, and your approach to implementing DPI should be mature, and then you apply AI as a scaffolding on top of that foundation to accelerate your build and delivery.

So what are those foundations? Clean data and data maturity, maturity of your data architectures, clean registers, for instance, also reliable APIs, APIs which are not susceptible to cyber attacks, et cetera, and also the institutional capacity. The institutional capacity behind the DPI delivery. So these are foundations which should be in place. Now, on top of that, you apply AI and there are several unique features of AI which would deliver you super experience to the citizen. And I’ll come to that in a short time. The last one would be the safeguards, bias detection, and also the augmentation to consent. Because when you have AI systems in place, your consent could be AI generated as well. So you need to be careful about the augmentation you require, explainability, and human in the loop as well in terms of your safeguards.

So we need to be conscious that bias, opacity, at scale would mean harm at scale as well. So everything AI scales. So we need to be conscious that bias, opacity, at scale would mean harm at scale as well in terms of your safeguards. So we need to be conscious that bias, opacity, at scale would mean harm at scale as well in terms of your safeguards. Sovereignty is all about capability. It’s not isolation. it’s about building the capability in terms of being neutral and having a neutral capability in terms of vendors, in terms of cloud, in terms of technologies so that you have control, that the state has control over those building blocks and can choose across technology delivery parts as well as core technologies also data classification, protection, privacy basics, so I think if you go back to the foundations of protecting your citizens and their data you can’t go wrong and AI will only be an accelerator finally the experience angle, because a government would be very keen to deliver super experience and where does AI come in here I believe there are some specific strengths of AI when combined with DPIs and that is in terms of in particular API orchestration picking the right API to call at the right time also the fact that AI enables unconstrained scenarios so while linear methodologies would be constrained to say 4 or 5 scenarios when a citizen needs some help AI can do a billion scenarios and each scenario could be implemented using a unique set of API calls and unique set of DPI access and this can be orchestrated using AI so can AI make a big difference a leap forward in service delivery?

I believe so can governments and the sovereign use it or should they? Definitely but we need to be conscious of those 4 dimensions that I just mentioned of inclusion, integrity, safeguards, sovereignty all of these are important and we are still coming together to deliver a perfect experience

Thank you, Dr. Hans. I will give a bit of a breather to Sangbu. Thank you, Sangbu. Great to have you. I’ll go to Robert first and then… Robert, as you at UNDP have led a very important work on safeguards, worked on the Global Digital Compact, engaged a number of countries on sustainable development goals and how AI and DPI can be an accelerant to all of those outcomes if you want. From your vantage point as you look at the AI revolution that’s unfolding upon us, what captures, what’s top of mind for UNDP?

Yeah, no. So I think that the reason we have been so excited about digital public infrastructure as an approach overall is that it really does bring some very particular characteristics. And one of those, maybe the most important, is the population scale. and so it is something that can reach so many people so quickly if you get it right. We also have been learning that if you don’t get it right, then you can have problems and challenges at scale and so in the DPI space, one of the things that concern us the most is how do we ensure that as countries are building their DPI, how do we make sure that we are putting the safeguards in place?

And this is work that has been supported generously by Codevelop and Gates Foundation and others and has led over the last year and a half or so to the creation of a universal DPI safeguards framework which we’re now implementing or supporting a number of countries at national level to implement. But… But when we talk about that, what does it actually mean? And Dr. Hans referred to some of these important things. And so we are talking about the DPI as a whole. And so we are talking about the DPI as a whole. And so we are talking about the DPI as a whole. And it’s about what we’re learning is that the earlier in the process that you can start discussing the safeguards, the better off you’ll be down the road in terms of inclusion.

So if efficiency is your only metric, then you will probably rush ahead and leave people out. But if inclusion is your driving KPI, then you really need to make sure that you’re sitting down at the beginning and planning and designing with people in mind. When it comes to AI, then, it’s basically the same thing. And we need to be really careful, as Dr. Hans was saying, that we are putting the inclusion aspects, the safeguard aspects, at the center of our planning from the very beginning. And so that means, and you referred to a couple of these, but, you know, a multilingual platform, multimodal that can support people with disabilities, making sure that you are correcting for people.

So that’s one of the things that we’re doing. And I think that’s a really good thing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. And I think that’s one of the things that we’re doing. and then of course making sure that data sets are there’s bias detection or you’ve got some understanding of your accuracy representation all of those kinds of things because when you’re going to layer AI into DPI as an accelerator well then you want to be sure that you’re on the right track and that people are considered from the very beginning with that and so I think that’s we see all the

Thank you Rob Can I go to Sangbu now as you see the evolution of DPI plus AI from the World Bank standpoint one of the things that we’ve often looked at is is having open data sets that can train air engines would be an important way of advancing the benefits of AI but as you know governments build silos one silo after the other how are you seeing this from the World Bank’s standpoint what is the next journey for the next 3 -5 years looking like in terms of getting countries to become more AI ready any thoughts on that would be great

DPI has a lot of aspects and characteristics this is one of the very productive and efficient way to ensure the interoperability even though it cannot ensure everything but we make some more effort to clarify some more interoperability capability so but it is basically what it looked like. So that’s why in AEI era, in order to prevent some siloed approach, DPI can play a very significant role, I would say. If you think about the AEI era, what is the relationship between DPI and AEI? If I just compare the DPI from the previous version of DPI, I would say DPI is more helpful for AEI era compared to the previous mobile era in many reasons. If you look back on our history of computing, we started from the computer and PC and evolved to mobile and evolved to AEI.

The trend is that it is from the very supplier -centric approach through the user -centric approach. We are evolving from the supplier mindset through the user mindset. The DPI is exactly just some of the really, really important tools to ensure that user -centricity. Because without identifying some good tools of users and some interoperability, it cannot really be achieved to fully support the user -customized things. So to me, the DPI and AI will be really important relationships. On the other hand, one opportunity… I also see through this trend… maybe DPI can be also very well upgraded quickly and efficiently upgraded through AI enabled technology so we used to collect all the data in a very manual way and then try very hard to streamline the governance of data sometimes very manually and sometimes by some intervention by the programmer but now we can see some old medical way so that we can quickly streamline all the DPI platforms so this is really a good opportunity for all of us from the World Bank perspective we really expect to see some more progress of this space

Wonderful Saibal as a senior partner at Boston Consulting Group you have a bird’s eye view of a lot of early thinking on AI led innovation around the world, especially in the private sector. And I’m also sure you’ve been observing closely the India trajectory of DPI in the last decade or so. What lessons from the DPI journey in India can we take to the AI era that might also propel private sector innovation to levels beyond if we just didn’t think about DPI as core infrastructure for AI?

So I think, firstly, India’s journey in DPIs has been a fascinating one. It makes me immensely proud that whichever country I go to, and I do work with quite a few South Asian and Southeast Asian countries, India is almost always seen as a benchmark in DPI, and now increasingly AI on top of DPI. I think, so maybe I’ll just answer your question in two parts, right? I mean, when we were building DPIs, starting with Aadhaar, you know, and then, you know, moving on to UPI, et cetera, the idea was always to build open population scale software, which can then trigger innovation, right? And that’s exactly what has happened over the last decade, right? The amount of innovation that has been built on these DPIs is mind -blowing.

I mean, India is now a country of 120 unicorns, and every unicorn, some way or the other, leverages the DPIs. So then, coming to the second part of your question, I mean, what lessons, right? So if you look at the way India is now thinking about AI, and as BCG, we have been privileged to be part of two of the very leading companies in the world, right? So we have done several seminal efforts, one with India AI Mission to build AI Coach, India’s national AI platform. and also with the state of Telangana to build the equivalent for the state of Telangana. Both of these happened last year. I think we are taking a very, very similar ethos, right?

So if you think about India, as I mentioned, there are 120 unicorns. There is no dearth of VC funding in India at all. However, 90 % of that VC funding actually goes into fintech and e -commerce. Very little goes into climate and sustainability. Very little goes into education. Very little goes into MSME relevant topics. So there is a gap, right? So what these platforms are trying to do? And then similarly, access to data within private sector, there is good quality data. But the biggest source of data in India is the government. The access to government data is still at a very nascent state. Quality of data and access to data. And then, of course, the biggest thing in AI, which is compute, right?

I mean, access to compute is what makes or breaks a startup. So the way in India, the way I see it, the way we have started thinking about AI platforms, and I’ll use the word platform, it treats AI as a shared public infrastructure. Just like DPI was a shared public infrastructure, it treats now AI as a shared public infrastructure. If you look at India AI mission, more than 38 ,000 GPUs are now available at, you know, less than rupees 60 per hour, which is less than a dollar per hour. So if you are a startup, very early stages, working in your garage, suddenly GPUs have become a bit more affordable for you, right? That’s genuine shared public infrastructure.

Government data, it’s early days, it’s very early days, but government data is being provided access to through platforms like AI Coach, or in Telangana through the TGDX. to solve the financing problem. How do you channel financing into socially sensitive sectors? The central government and the states are thinking of building fund of funds, which actually then, you know, encourage VCs to co -invest, focusing on those socially sensitive sectors where they would normally not invest in. So if you think about it, the ethos is very similar to when we were building DPIs. How do I create shared capabilities centrally, which then can trigger an entire new wave of startups, you know, and therefore a market ecosystem, just like the DPIs did.

So that’s how I see the journey.

That’s great insight. I think it seems from what you say, there’s a beginning of a new innovation cycle for the private sector cycle, and we’re looking forward to what comes out. I just wanted to, we have about 10 minutes left. have a somewhat common forward -looking question to all panelists. As the economists say, in the long run, we’re all dead. But what is long run in the AI ecosystem? Is it five years? Is it three years? Because it’s so hard to predict everything. Every day there’s something new happening. So I don’t know, Dr. Hans, if you’re okay going first on this forward -looking question. The question I would ask of you is, as a relatively smaller island nation, how do you expect to leverage this wave of AI innovation over the next three years, maybe five years?

What steps are you anticipating? How are you preparing yourselves to leverage this power that we have to advance our development outcomes?

So there’s a plus and a minus of being small. Let’s start with some of the challenges. Being small, you’re on the wrong side of the AI divide unless you’re economically in a very powerful position. Say for example, Singapore is an outlier, a small country with a lot of economic power and therefore attracts investors, attracts talent and the sighting of business. Country, one of the challenges we have in Sri Lanka is one, getting that minimum level of sovereign AI infrastructure in place, having the ecosystems around it, retaining our talent. Sri Lanka has very good talent but retaining and developing the talent for Sri Lanka is a challenge but one that we are confident that we can deliver.

So future of AI I think will depend on the market. We’ll also depend on the people. We’ll also depend on the trust in us. The institutions like the data protection, the institutions as well as the laws, and Sri Lanka is very mature on this front. So on the trust side, a smaller country can execute precisely with laser -sharp focus and therefore has a strength. Talent side, again, it’s something to focus on. Now when it comes to the marriage with DPI, I think it falls onto the positives of being a small country because the ability to implement modular systems in a neat and flexible way, in a way that these blocks themselves will evolve, on the confidence that you have a strong trust environment, gives you the ability to build in AI where, like I mentioned earlier, AI sits.

On top of a solid DPI, a mature DPI frame. So I feel the future, AI future will of course be very close, can add that extra piece of experience, lower cost, faster, and more flexible, meaning that it can address multiple scenarios through digital twin and other such AI constructs to deliver citizens a very customized, I’m from the service industry in the past, so I use the word customized, but citizen -specific experience. We’ve been tracking the learning about the focus of the new government and their leadership and the presence of AI in the world. We’re working with the government’s leadership, making big advances on DPA and AI, and looking forward to much exciting stuff in Sri Lanka in the next two to three years.

Thank you for those comments, Dr. Hans. Rob, could I come to you and think about, you know, you’ve gone through the process over the last couple of years with DPI safeguards. You have the Global Digital Compact. I know there’s a lot of work you’re doing on AI safeguards. Moving away from safeguards, I wanted to see how you are envisioning the developmental role of UNDP leveraging AI in the next three years. I guess three years is long term, but anything you can say that would be helpful for us.

No, absolutely. So I think there’s a couple levels. So there is an internal to the organization level, which is how do we ensure that UNDP itself has capabilities for leveraging AI to maximum effect. And so there’s a kind of a base level of work that we’ve done internally to the organization, upskilling programs, investing in some, you know, making sure foundation model capabilities are available, working in some SLMs, et cetera, et cetera. Then there’s the layer of working across the kind of verticals that UNDP has, whether it’s environmental action, governance programs, energy, et cetera, et cetera. And so how do we embed AI solutions and thinking and approaches into those verticals? And then there’s the picture of how are we going to support our country partners?

And as you said, we’re engaged in quite a few countries already on AI transformation support, and it’s kind of looking at ecosystem pieces. Do countries have that mix of elements that Dr. Hans was referring to? Do you have the compute accessibility? Do you have talent? Do you have the data available? And so on and so forth. but the one thing that we’ve announced in this summit, during this summit is an exciting partnership with Xstep and a number of other players and we’re one of those players on something we’re calling 100 Pathways or Diffusion Pathways and that is looking kind of like more of a use case driven approach and finding over the next few years 100 different pathways to scaling responsible use of AI along different use cases and it’s something that we’re really excited about because I think we need that to complement the ecosystem support

That’s so exciting to hear because I think there will be a lot of iteration discovery and innovation to discover those 100 pathways to actually add value to people on the ground looking forward to what comes out of that Thank you if you were to look at the last decade of how development banks have funded and looked at digitization, and now if you look at the three years ahead of you, what might the World Bank and MDBs do differently to be ready, help countries become more ready for the AI era that you haven’t been doing in the last decade or so? Any thoughts would be great.

So good news is that from the Internet network point of view, the coverage is pretty good. So even in sub -Saharan Africa, more than 90 % of area of sub -Saharan Africa is covered by three -plus generation mobile tower. But the issue is that we are really struggling with lack of demand. How are we going to fully utilize this by creating some more value and profit? So now we are modifying some approach to really think about creating demand through government program, through developing some use cases. That’s why we are just keep highlighting the importance of small AI. Small AI is not really a small thing. So it is really about how we can really change the lives of our people.

So our approach is a little tweaking to the user -centric and demand -driven things. That’s our approach.

That’s great. And I think MDBs and the relationships that they have with countries can make a big difference in how the evolution and the benefits of AI will come to people. So looking forward to what comes out. Saibal, I know I started off by bucketing you as a private sector guy. But I also know you’ve been thinking deeply about government policy that enables or doesn’t enable innovation and growth. So as you look at the next few years, even from private sector innovation lens, what government policies might propel the innovation ecosystem? to serve the underserved populations around the world. Any quick thoughts, Saibal?

So, see, I think, you know, the government has a very tricky balancing act, right? I mean, when we were going through this entire experience of building AI Kosh, there’s obviously, for very good reasons, a lot of sensitivity around sovereign data, what data to expose, and if exposed to whom, right? Equally, without sharing of data, I mean, the reason why AI has taken off in a big way in some sectors is because internet came into picture 30 years back, and internet has been pumping, you know, billions and billions and gigabytes of data, right? So AI has something to chew upon. Now, if we have to do the same with government data, then that data, you know, needs to be exposed in a controlled manner.

So my sense is that from a policy standpoint, how do you actually provide that access to data? I mean, walking that tightrope where valuable data is made available to the innovators while not compromising on sovereignty or safety. I think that is one of the policy areas the government has to look at. Specifically for a country like India or similar countries which operate in a federated model, the center can do only so much. The real action, as we know, in a country like India happens at the state level and we have 30 plus of them combining states and union territories. So at the state level also, similar policies and institutions have to be set up. So Telangana, for example, has set up a Section 8 public sector undertaking to drive AI, right?

That creates the kind of focus and the agility that you will need to keep pace with this technology and do some real work at the grassroots. So my suggestion would be create those. Accountable institutions. who can anchor and drive AI and amend policies around data to make sure that people get access, the innovators get access while not compromising on sovereignty. It’s not an easy thing to do, but yeah, those would be my words.

Thank you, Saibal. I think the role of institutions, both for safeguards, but also to enable the innovation ecosystem haven’t been more important than it’s now more important than ever before. I think it’s hard to summarize this conversation, but I will say that I think we’re at the cusp of something extremely important and something very potent in some ways and can unlock a lot of opportunity for innovation. billions of people. Especially, I think, the segment of the population that was left out of the digital revolution because voice was not the predominant way of interacting. I think AI opens up that window and hopefully will drive much more widespread adoption and usage by common people around the world.

Looking forward to the next few years and thank you very much to our panelists for a wonderful discussion. Thank you all. Thank you.

Thank you so much to all the speakers. We will just have one memento being given by the organizing team. We made it. Thank you so much for being a part of the India AI Impact Summit. Just to tell you that the expo will be open tomorrow. People still want to come in and people are still not tired, but we are done for today and the sessions are done. Thank you so much. Thank you. Thank you.