Uncategorized

that can take you farther and farther ahead in the world economy. So I see it not as a job reduction. I see it as opportunity for India to create a global AI -dominant nation. And that does not mean that that is easy to do. But at the same point of time, most of us have to take a will, and we should be taking the commitment towards leveraging AI first. For example, like if you have a smartphone now, even a small shopkeeper does not believe he does not have a smartphone. They get payment from Paytm and work a lot of life. They assume that there is a smartphone, which is fair. Now, if you believe that you can build a business model, or entrepreneurs will build a business model as a customer, you use AI first products, the productivity will be dramatically higher.

So a person who was running a shop can run multiple shops. So GDP growth will automatically come in a different size because the productivity per person will be higher. And that is the lacking thing of India. the problem will be that will India be that bigger market or not and I want to tell you this is a very good lucky moment that we all are sitting it is tough to grow from 20 billion to 25 billion anyway that is 50 % growth and 25 % growth is very tough but luckily India is right now living in 2 .5, 3, 3 .5 trillion economy and I am comfortably going to say that 2 trillion dollar can be compoundingly growing in next 7, 10 years so if you are sitting in this economy in 2026 and you believe that next 2 trillion is going to come in next 10 years so I don’t have to tell you out of 2 trillion dollars many of you can get many of million dollars or billion dollars entities or business benefit to yourself so that’s why the India is a bull case scenario and AI on top of it is the absolute bull scenario I don’t think there is any challenge to that someone asked me this question in some other audience so I said like in Ramcharitmaras if Pran goes but Bachar doesn’t go similarly if Job goes but AI doesn’t go Well, job jai hoga nahi, kyunki people will have method to discover themselves to do more productive job.

So, in my opinion, the traditional kind of work, main job ko yaha pe metaphorically use kar raha hoon, traditional kind of work versus the AI kind of work. And so, that will happen. Main completely maata hoon, main isko apne tarike se express karna chahta hoon. Humare kisi bhi system ko banane ke liye, koi bhi job karne ke liye company ke andar ya kahin pe bhi, kahin na kahin koi bottleneck hota hai. Aur ab increasingly, ap log ye figure out karoge, ki wo wala bottleneck pe eliminate ho jata hai. Iska basically matlab hai ki sadak chodi ho gayi. Problem kahin aur shift ho gayi hai. Problems disappear nahi hoti hai, wo bas move kar jati hai kisi aur jaga ke upar.

Outer ring road ke road block se bhi hume yahi bata chalta hai. Ke road, sae pehle Nehru place pe ruka karta tha jam, ab jaake wo airport ke paas lagta hai.

Very good. Main samajh gaya. Achha sab na aya. Okay. Toh ek aur related question tha, ki opportunity. is a lot. And when we look at the exhibition hall, someone is making a chip, someone is making a data center, someone is making a foundation model. So should we all make a foundation model in India? Should we make our own chips? Should we make our own applications?

Very good. So first I want to say one thing. To make a foundation model, English, we’ll speak English and Hindi, okay? And you should use AI to live translate. No, it’s okay. We’ll definitely, I understand, and we will do it. So in my opinion, India has to build a foundation model. This is no compromise statement. Not because that we can make a better financial foundation model or not, but because we as country have to move on from services culture. I mean. There is nothing wrong about IT service business. There is nothing wrong about BPO and business services, but it is an obligation on us. that we should move up in the value chain. It’s like growing up in the life.

You do not move up in the value chain. You rather continuously make something for someone. So most of founders or capable technology people in Silicon Valley will have a significant mix of Indian people. So as a people, we are able and capable. So can’t our country have enough amount of resource allocated to those individual capable people that they can make foundation model? We have to do it and we have to prove it. So I first of all applaud Sarvam that they have launched it and they have launched a perfectly, I would say awesome foundation model from India. And I want many of us to do it so that this race does not look like that there was none or only one.

There should be tens of foundation model to prove in the world that Indians can do it and Indians are doing it in India. That is why we need foundation model. And now the question of that whether it is on an ego or is it on a use case basis, the advantage of India financial or made in India sovereign model will be that the amount of… nuance and biases that happen so what is a foundation model it is like aggregated knowledge of what you give the feed off now obviously all of us have our own perfect understanding of whether this is correct or not so after the power a lot of us would know this like for a small kid making them eat banana or curd in the night is considered as if next day they may have some cold or some other thing but when you go on internet internet is half and half divided it doesn’t believe it and it believes it and then you’re totally confused but when you go Ayurveda then it says something else and obviously it follows the bath with and also at cuff kind of philosophy so that the answer is not the straightforward as straightforward as we believe but it has an answer in a different way versus an answer in a different way in a Western medicine if you will now I’m not saying Western medicine is right or wrong I’m saying I need my opinion I need the culture or the knowledge that we have inherited extended towards the next generation And I want that to happen by the intelligence that we will query, which means that that can only be done by somebody who is making for it.

If we don’t make for it, our all compounded historical knowledge will be lacking in the next generation. So instead of adding on top of it, we will not be able to take it further. So it is a case for India to build not just a perfect foundation model, also retrained models. Like you remove the biases that it has, you trim it to the ability that you want. And you make it for the purpose that you want. And that I believe India’s opportunity is even bigger beyond just the foundation model.

So I have a two -part follow -up. All right. So I believe that this is probably just a way to discourage people. And it may be more than that. But we often hear, mostly on Twitter or from these large foundation model companies, that if you don’t have 10 ,000 crores, why are you even in this place? What is your viewpoint on that?

I mean, literally, PTM, we both of us, put more than 10 ,000 of you, put 25 ,000 crore on the table for making this humble QR which is everywhere now. So there is an ambition and there is a commitment. So the question whether you put a billion or two or four or five, it’s not that question. I think the question should be whether there will be enough business model of that. Will we have a skill to market it? And remember, the world is right now not just made in America. There is a European model and there is a China set of models. We need our models. And the proving point now that there is, I would say, enough knowledge of model creation is that it is not literally about a billion dollar or two or ten.

It is also about the kind of smartness that you can run on the model creation. RL and so on and so forth. And there is a lot of chemistry and math or let me put it rather better way of saying it. There is a lot of physics and gravity handling capabilities that are discovered now that you can build model in a much less cost. So whether, I mean, another question that a couple of us were asked when I was roaming around, why don’t you build an LLM? So actually, now I should ask you, what kind of LLM are we or you building when we talk about that there should be a model made in India?

Yeah, this was going to be my question to him. But it’s always fun to answer. I think that models are just not what we know them today to be question and answer knowledge sharing sessions. There are models that have capabilities for reasoning, problem solving, for building agents. So you want to give them agency, the ability to act and so on. I think that to answer your question of whether we should build or not, I believe that there is much more than a question and answer chatbot model. And we believe that we should build models that actually solve problems that actually have. the committed confident agency to take actions on behalf of others

okay so you’re saying and this is for everybody to know we are making models but their models for a vertical problem not for a horizontal scope so instead of talking about 200 billion token model we would be making four or five billion token model or 20 billion token just in case like we should open the questionnaire from the audience also because some of you may have and we will not have a time after that sort of line item so while we are talking you can just raise the hand and I’ll just allow this to okay now that’s you literally are asking good vertical problems to solve

oh it’s easy I mean imagine it’s a classic triangle of mess laws hierarchy the first is the financial foundation that you need to solve for financial services so there is a risk fraud control models that some of us are building and we built it and then you go on top of it like food food chain which is agree and processing of this if you want to produce high quality you food outcomes at a more yield, there is a tremendous amount of data that gets generated in visual amount of data and the farmers need access to it, nuance of it. We just heard how Prime Minister was able to ask a question that maybe a cow is not able to say that I am ill, if you have an ability to discover that the cow is ill.

Maybe a tree may not say or plant may not say that I need this mineral or supply and maybe you can read it. So there is a tremendous amount of vertical like I just told you example of finance, agri, husbandry, you can go towards now industrial, you can go towards the problem of pollution which is we are sitting in this country, this city. So there are many vertical solutions that we can build a specific use case.

So let me just help you guys think of it another way of it. Mr. Banj discovered an engine and that engine then got percolated by many to be made. so LLMs are like an engine and they are literally called intelligence engines just in case, now if you had access to an engine, can you make engine? you should, so that you can make your own vehicle use case, you can make a small car, you can make a big car, you can make a truck, you can make a trailer, there are so many use case of a transport so think of it in that eyesight, then you will say, well do I need to make a small car in India because I heard that there is a small car being made elsewhere no, you will need to make it, I can confirm it and so like in automobile industry, making an engine is an art, I would say making an engine equivalent of LLM is an art, yes, but many of us will have it and many of us are making for those

yeah, finance is the one of the few things which is part of vertical, but it is also one of the few things horizontal every industry requires and has a financial department food, finance, both both

yes so all of all humanity needs these foundations so you are into both vertical and horizontal at one go

Yeah, that’s right.

So, perfect. Now, question to you. The classic financial use case, considering we just heard it, what kind of use cases you believe the LLMs will save or solve for the problems that you see in financial industry and what you believe fundamentally could be solved for wider financial industry and we might have solved them or not yet?

I think that the best favor or the best value we can add is to remove biases in decision making that we already see in our financial system, which are actually a complete antithesis to the whole inclusion aim that we all have.

Beautiful. Give me an example of it. So, a very good starting point would be detecting whether a particular transaction should go through or not. There is a lot of bias today in the rules that we set, in the checks and balances that we set in allowing… transaction through and you can invariably say that if there is a loan officer deciding on whether you should be getting a loan, there is a lot of bias that creeps into it. We are not able to measure that today but when you ask a machine to make that decision, you actually remove those biases. So the person who may not present themselves very nicely is unlikely to get a loan but the machine doesn’t care about how you present yourself.

So you are saying that when you make a financial decision, when financial industry or system makes a decision, there may be unknown known biases and the unknown biases can be removed even if you want to keep the known biases. The good thing is that the machine actually now helps us identify what the biases used to exist which you are very used to. And speaking of financial industry and going beyond, one of the things that I want all of us to know. So classic. Financial inclusion is not just about payments inclusion or bank account inclusion, which I would say that India has perfectly solved. It will continue to grow towards the next requirement. And most of us need access to credit, access to insurance, access to classic wealth solutions.

Right now the wealth in India is only when you have crore or plus or less, somewhere around that amount. But a normal auto rickshaw driver who has 20 ,000, 50 ,000 or 2 lakh or 5 lakh saving also deserves to have a financial wealth model. And that person, poor fellow, by hearing or hearsay may invest or risk the capital at large. So the access to the financial services can become further and dramatically scalable once you add the power of AI to it. For example, like capability of auto rickshaw driver like I just now contextualized you. Imagine that person has 2 or 3 lakh rupees. You can suggest based on your risk and time horizon, you should put it in FD, you should put it in a gold or a sovereign gold fund or you could put it in let’s say some index fund.

Because you’re talking about 15 year forward money that is for your house, family, daughter or kids education or something. Now, those things can not be told by a commoner around them. And then you can make it in a language and you can answer the questions like ChatGP today answers the question in a language they speak and the language they get the answer. So in my opinion, AI will enable financial inclusion to a next level because it is not just about the access to the smartphone, but the question answers that you get from smartphone can become far more native and continuously possible. So let’s say this guy is busy whole day, can start in the evening when he is waiting for some customers and start talking to it and can take a decision.

Now, at least he has a second opinion and can do it. Similarly, healthcare, huge amount of capabilities because many of us literally just say, but you need to check whether there are some more symptoms or not and what the blood test report says. So the dramatic amount of inclusion of education, finances, healthcare will be such a catapulting impact. On the society.

Yes. So do you think the whole banking system will become redundant? Because today if I have to make a transaction, I’ll use Paytm. If I have to invest, I’ll use Paytm or a Zerodha. So why is the banks existing? Because with whole human interface, it’s becoming very difficult. They don’t solve our problems. And second thing is, is education, school is going to become redundant because what they’re learning is going to be not valid, will not have a shelf life at all.

Okay. I have an opinion on schools. I will, I will, I will. So it’s a very easy thing. If people start to make food at home, will the restaurant become redundant? Yeah, the demand may be a different kind and the need may be a different kind. So first of all, none of these things will become redundant because they do not offer literally the verbatim statement that we just made in this sentence, but they offer much more beyond that. For example, like banking inherently is about extending credit. So ability to have… I have a… well regulated place where you can deposit money and that deposit money is taken care of so that when you need it, it is available to you and when the economy needs it, it is available as a credit which is a business model of a bank is an ability that will never go away.

I mean it is an obligation for them to become even more able and capable in both parts. That security safety of your deposit and ability to disperse credit to the needy. Now that is a part called bank. If we treat bank as a bank branch and your question that you may not walk into the branch, fair. That may be the case because you will not need a branch to make the banking reach places. It can be very perfectly extended using a smart phone and now in a chat bot instead of just an app and that is what the beauty of this inclusion I am calling. The core machine of a financial institution is needed and even more because you will have even further bigger load coming in.

The bank, the way it is served through let’s say a branch could extend itself when the ATM happened through an ATM when apps are happening through an app. You can have a third party app. That’s not a problem but the core activity of the bank still belongs to the bank. And similarly, like you were saying for the schools, schools are not the place only for going in the classroom. How many of us did not attend the class when the teacher was in the class? And I’m talking about college days. I’m the one. Now, that does not mean the college was a bad experience. Rather, college is a social experience of meeting like -minded people and understanding and self -discovering beyond the syllabus in the class.

And I definitely agree with you that maybe a single method of teaching on a blackboard, whiteboard that you’re teaching and the people are writing and just putting up back in exam is changing. Then you have these different MBA institutes like Harvard is popular for case -based education. So those things can be extended towards now in the classroom of a common or other mass level of school. You know, that is what the power of AI is. I’m not a believer yet that you will not need it. I mean, homeschool is good, work from home is good, but going to office has its own use case. just like we just saw in the COVID days that everybody was working from home you can be selfish that I want to work from home because I have 10 more things to do that is always taking care of it and then sort it out but ultimately there is a value to go to a thing a place and that is in my opinion will perpetually remain whether it is school, bank or any other such institution your answer on the education no I completely no no no I am not going to say that the core philosophy that the bank is in a branch definitely changes the philosophy that they will be a let’s say bank manager or somebody to approve a loan will evolve but the core work will remain of the bank so DeFi etc are very different cultures or I would say they all are technology nothing wrong about them again the core philosophy that you store money at place and you demand for it banking.

I completely agree on the school front. School is not just education, also the social experience, and we’ve all benefited from that. So nothing more to add. But I do find an interesting theme across finance, healthcare, education, that it allows you to have more access and more personalized access. Like your doctor is your doctor. It’s not the doctor that says the paracetamol example. Same with your teacher. And I think that makes a very radical impact.

It’s amazing. My mother who had a heart and then stroke, which ended up becoming a stroke, heart disease. And then now I leverage the power of AI to keep a track, including she has a Fitbit that generates a feed that goes to my agent. And it triggers to me notification that if all is well, you should check it. Now, that possibility, the kind of nuanced care that I’m talking about, couldn’t have been possible without a doctor support. So I’ll give you another example. And this is something that you should check out. If you have a situation like this. So there was some medicine the doctor gave her. for certain rhythm control and suppressing beats and so on and so forth as the case was.

But some of them ended up taking her palate taste away, making her not feel to eat, and she was becoming weaker. So I talked to the chat GPT, I threw the prescription, and I said, she’s not eating, is there a problem here? So then chat GPT would tell me that I think this combination at this hour, which is pre -lunch, makes this a situation where she may not have a tendency to eat. If you move it to this time, and if you move this earlier, there will be enough window for her to eat. And any which ways, by seeing your heartbeat that I’m seeing, at this time, this beat compression that was needed was not required.

So I’m just suggesting that you should do it, and then obviously put a disclaimer that you talked to doctor and so on. So basically, same medicines in a different time schedule potentially will solve for it. And I sent this to the doctor, and the doctor was like, Like, the good thing I want to tell you is, doctor replied three words, you can do it. Yes, you can. Actually, he said, yes, you can. I said, sir, do you think, I very frankly shared, because doctors don’t feel, and any person who’s skilled in that domain may not feel that you’re bringing cocky input. So I said, my mother was this, and I understand it is tough to go through this nuance.

I tried asking this, it is suggesting this, and this is a brief if you want to read the PDF, but net output is, instead of it to be done before lunch, I can do it after lunch. He said, okay. He said, yeah, you can do it. Yes, you can. That’s the three words of life. And I think then after that, she became better in terms of, because she’s already taken. Now, what we are talking is, that medical, education, finance, agriculture will catapult into a very different stage and age. I can promise you this today, if many of us, the new Gen Z generation, who’s used to a smartphone, must be asking, really? How were you living the life before?

so I can promise you one thing by the time 27 passes and now we are in 25 so some of us are seeing it and I am literally putting a deadline of 27 which means 2 years that by 27 passes every new generation will think that oh are you saying that you used to search and then click on every page and then scroll down to read and then decipher with your cognitive load that what is happening and how did you decide that it is correct or not did you not have an alternate opinion and this and that 2030s will be so shockingly different that we will look archaic in 2010s and 2020s that I can definitely promise

question from you your mother’s example made me believe that yes these days chat GPT is probably more humane than a doctor himself because this kind of conclusion might not be given by a doctor my question is Since you are giving such wonderful use cases, it’s about the curd and banana example. I would like you to elaborate on that. Would you rather want an Indian -specific model for the purpose, or you are talking about Indian context data, or is it that only Indian children get affected from that banana and curd?

Yeah, very good. I like it. Classic future mother, maybe. Perfect. Number one, the data gets trained on what is on digitized and available Internet, because it literally browsed and scrolled Internet and brought it. And then it gives the weight to the common knowledge. So if more people said, blah happened, it believes that, because there is nobody who answered every question for it. How does it believe that what I just now said should happen is that because it learned more number of times it was said. It’s a very surprising method of adding the bias to the model. for example like let’s say we may have a very contentious statement of history if more people tell the model, model will start believing it, did you know this it’s exactly how the social media, viral news sometime which is wrong could be believed to be true because that is how the brain thinks, we also why are you saying this, why are you saying this I feel like I should eat banana or I can eat Lato at night the more of what you see you start believing is the thesis that the model goes through now do you fundamentally believe that internet is filled up with domestic or let’s say Indian or Indian background or history and culture and richness answer is probably not because the western English civilization has brought lot of content which is in English and there is a much less content which is generated, translated in English repeated a number of times, people are discussing so it’s an inherent limitation of our own knowledge that did not propagate on internet, who would have guessed that whatever is written on internet is the ultimate truth versus what is written in the book Now that gap lacks and creates this gap of knowledge of the model.

Can international model do it? Yeah, if you want to tell that no, no, no, on this, this is the truth model. And the problem of the international labs is that how do I know this is the truth model? I can’t. And a lot of people need to say this, that it is a truth model. So there is a limitation of international labs who are making these models. Then can this be done by India lab? India lab will respectfully say this, that why don’t you learn what is written here first and treat this as that it is your definitive knowledge. And there it goes, the obligation to build in India.

And I’ll add just one more thing. There is absolutely a very strong risk of bias from the inputs that you give. But the model has the advantage of also knowing you, who is the person who is asking the question and can reason through all the inputs and still give you a balance.

So this is a very interesting thing. I want to tell you that my model, because I have started to give it a lot of things, that many of you used to be doing claw and so on and so forth. it starts to say, oh you are in Delhi this is what Indians do it literally said that, I was trying to ask that what do you think okay it’s a very interesting product feature that I was discussing with it, about wealth system, KTM money and I was trying to ask that give me the cultural nuances of managing money that Indians have which I could add in my feature and the kind of feature I was like, oh yeah that’s good, I would have never thought as a logical person to think like that so yes, it can, you had a question chat GPT you are asking this is very good, very cute Amitabh, he is reading the essay again and again but it’s a good thing chat GPT versus meet our agent sometime later we will do this our agent will talk to your agent then they will decide whether we should talk to you or not this is coming, very seriously now Uber if you order Uber, your agent will call you first Uber, this is expected I have to go now and I am leaving then PJ will pop up and say should I call Uber?

yes, call it should I call the same one? I am not getting this I will make it quick all those things that we all experience that will be so dramatically different guys it is real you will be represented by your agent and that is the fight all these clients are running that is why agent came and that is why agent is hyped up at this time because your agent could be based on OpenAI or Gemini or Claude or whomsoever that is why they are talking about it and your agent will be working with you he has tagalongs right there back to you read from that

so my question is do you believe the future of stock brokerage industry will be an AI native from the ground up where AI agent can be the primary interface or you think AI will be just the feature in the stock brokerage platform

And I’m not saying that it is the perfect way to do it, but I’m saying that we are all in committed towards agent -first interfaces. Vijay, I have a follow -up question around this agent -first interfaces. So when an agent, on behalf of me, goes to the Uber app, will it see ads? Will it have to log in with Google? Well, nerd question. And for every one of you, the point that he’s trying to bring it is that will every other stack also change or not? Well, that is the beauty. The agent will talk to agent. Uber will also become an agent. And instead of trying to authenticate my login, et cetera, it’ll say I’m coming and I’m an agent on behalf of.

Just like your ambassadors used to say, I’m the king, I’m from his court. And he has sent this message. You take this VATS. You take this VATS. This is how it will be. I’m the king, I’m from Harinder Thakkar’s court. And he wants Uber. What kind of Uber do you want, VATS? They want such an Uber that can take them so fast. So I feel that at this time, you can throw so much money. So how will you want to pay? Will you want to do it with a token? Will you want to do it like that? Will you give it later? No, take this money. And you can call Uber. It will happen like this.

And this is real. I am telling you, you are laughing, I am laughing. One day someone said, I will stand here, and I will press the button, and the car will come. That day, he laughed at him. Before that, I will stand here, I will make a call, and I can talk to someone through the call. He laughed at that. Then, I will stand here, I will make a video, and my video will be in London, and he laughed at that. So all these technologies, which we natively, as normal as it happens, what kind of trick question it is, in near incremental future, the way this is going to be is so dramatically different, that those who are making apps, I have only one suggestion to them, that make an app with a non -icon based interface.

Yes,

One question here is, for the students who are going through their degrees right now and not in the data tech AI space, in the functional domains like finance and accounting, like HR and so on and so forth. So what do those poor people do? How do they think? Because, you know, how they are running through the education system, that’s old, that’s outdated, that’s still that. So what’s future for them?

So it’s a very interesting question. A lot of people tell me that, does it mean that we should study only, I mean, some days back we used to talk about we should teach programming to your kids. Like IIT preparation was done from class number five. It’s a joke, it’s a stretch joke, you can imagine what I’m trying to say. You came in class six, do IIT preparation. So by doing this, when people were born, they were given a programming language book with a book. So that will end, right? Because you have no more programming. So we are in a flux. Yes, ma ‘am. as we speak in 24, 25, 26. Okay, 24 mein kuch kam log the, 25 mein logon ko laga, 26 mein most of us will go through.

Your question has come in 26. This question did exist in 24. And there it goes, that we will remain in a flux to decide whether this is needed or not. Some will do it out of their intent and commitment to do it and learn, and some will do it for obligation. The fact that we always assumed that education will give us a job, and now the job is disappearing, so uncomfort is that, do you do it out of your vocation and interest and passion, or you do it for the outcome of a job? If you’re doing it for the outcome of a job, I’m not skilled enough in, let’s say, humanities or art, but I can produce an art which is sounding logical.

One of my teammates who was roaming around in Kerala and wanted to figure out what is a good Ayurveda outcome that he wanted to seek for some problem, and he just used Claude to generate questions and nuances and formulas and so on, and then he went to the… Kerala labs and then they were like haan ye toh bahut achha hai, aapne kaise socha, ye toh bahut acha soch ke laay ho. Now the poor fellow doesn’t even know that he does not have a skill of Ayurveda. And he’s talking to the super skill of Ayurveda as if that person is skilled. So now there goes this question. Did you need Ayurveda education to become that intelligent?

Answer is not necessarily. You were able to use the tool of AI. So to the person whosoever it is, whether it is a programming language and computer science student or whether it is a humanities student, I’m not going to say don’t do that what you’re doing. I’d rather say definitely learn how to use jaisa ek time kotha na, computer seekhlo bete, uski bhot kaam ainge. So ussi tarike se main bolna chahta ho. Leverage and make your work AI first as much more because you will be relevant. And this is not a question of these people who are right now graduating because there is something that will remain for some more time. If you think about 2030 onward when this is all prevalent.

Ab jaisa painting hoti thi, aapne dekha hoga shole ka jo studio ka jo poster bana karta tha, wo haat se banaye tha. And making posters, in Bombay, there used to be an art. Studios, posters of Mahbub studio and RK studio are called legendary. Now where are they? And they are made in a spot like this. And Ghibli art is made like this, you make this. So the point is that, does it mean we don’t need a new cartoonist? But we will need one who is making with a computer. And he will say, I want this kind of nuance, it will become an art fashion. So if the today’s artist only uses, let’s say, paint, then it will be so unique and rare that that person will have its own value.

My God, handmade art. Like today, there is a tradition of cold hand pressed oil. Which was very popular at one time, that you are not doing industrialization. Oh my God, this is perfectly hand pressed juice to you. It is made with a jug of sugar cane juice. So that is it

Thank you, Vijay. Fantastic and energetic talk. Thank you. So, a little while ago, you told me that LLM, Foundation Model, should be done. Yes, sir. And the thing is, both while making a foundation model and during the inference, actually a lot of data is needed. And just like in the finance industry, actually, there is a lot of regulation and this will keep coming. Actually, in such a constrained data environment, how do you kind of make a good LLM and inference time? So, let’s say this could be applicable on any industry, that this industry does not have a lot of data. How do we make a good LLM of that industry?

Short answer is that you work with that industry players, whether it is regulated or not. I apply for everyone. And you find out all stakeholders and pursue them to learn what you could bring on the table. And people understand the need of it. If not all, some will. And in my opinion, the training model and the why are you training it, if you are able to articulate it well, I mean, the process. Progress, even progress is a very interesting thing. People. the regulation is not for not progress. Regulation is for not what it can slide and fall. So remember, regulation is also for progress. And that is always the case. Regulators are the reason that we have such vibrant financial system in this country.

At the same point of time, what they protect us from is that not falling apart of the system. So there is a respect and value of what they do. And if you talk to different industry stakeholders, you will get access to the insights and data. And there are different regulators starting from their sandbox programs and so on. So they allow it. And I think I’ll just add one more thing. Data, just double click into that data. There is only some kinds of data, like my personal information, that is usually something that you do not want. The regulator doesn’t want, you don’t want. But outside of that, there is plenty. There is plenty.

Yes. So as you mentioned about AI and all these things forthcoming with the Uber, the IFPS interface, do you think that… growing as AI will be growing fast and forth there will be more inequality or will it remove inequality in terms of will power get sustained in few hands

perfect this is a very interesting question inequality I can take it from money perspective because inequality from other perspective I am not taking it AI for the first time is one of that technology that will be easier to use for everyone because you are talking in native language and you can speak even if you do not know how to type or write so it’s a very inclusive technology for itself and its outcome is very profoundly powerful that a rich person or a able person so let’s say that you are trying to fight a battle of certain skill or ability which are rich in terms of money or able in terms of skill you can very comfortably write this person so AI is that horse or super car or a rocket ship that you can ride easily and then you can go ahead of anybody in a zero sum business if you are in a wholesome business you can expand yourself So AI not only is inclusive, AI is actually the superpower that would reduce the gap between rich and poor and be more inclusive.

And that is what I’m trying to say, that it is not the technology of rich. Rich has a fear. You must have seen, you must have read something on social media, that parents should be given education and old should be given health and women should be given beauty. Such business models are written on social media. So there is a model written in it. I want to say that it is written for the rich to give them safety. Because rich has done this. They want safety, security, exclusivity. Why? Because they don’t want others to take away that what they have. So they want to stay disconnected from the world. That’s why rich’s lounge is different from the normal people’s lounge.

But sir, having said so, there is also a risk with AI, which we are underestimating at this point of time. So what do you think could be that potential thing?

I think risk is in driving the car and coming here. And also in using the phone. so risk I will not say how much risk how much not this gauge is more important the generic line that it is risk is not complete line it could be that is it a risk that every common person can take care for example like kids are not allowed to cross the road alone but as an adult you are expected to cross the road alone and there is a risk that’s why you are supposed to do the left and right check and so on so AI surprisingly is that less level of risky that even a commoner could do it that’s what I am trying to say yes yes ma my model my model is right now in the beta and we are using it and just like she mentioned very nice your infectious enthusiasm to drive this technology that’s very commendable thank you thank you so much for that and you are at the driving state so to speak thank you so like like she mentioned my always my concern is about how to solve this distribution problem of AI of AI like cool No, the problem was…

And how to make it a public good. Yeah, that’s right. I really didn’t like that concept. I think I want to tell you that technology distribution happens on a terminal. So as you remember, there was a time when computer used to not be with people. Laptop was not with people. And the smartphone was not with people. So governments used to give free laptop, free computer program. Then some free smartphone and tablet program also came. Politicians put it in the well wishes of different states. I think anyone who has a compute terminal internet connected, which is said to be a data positive country in India, everyone has access to it. So if you have access to a computing device with internet connection, you’ve got full AI access.

The good thing is it is not installed in your device. It is not a version of a software in your device. It is not requiring high amount of compute or memory or something on the device. I mean, we created an AI sound box, which is a very natively small device. And it is as capable. As even as somebody else’s, Sam Altman’s computer device. That is the beauty. about it. So AI is far more inclusive and very easy to diffuse. I’m glad to hear that you being in the driving seat, think it as a problem. That’s very nice.

And I have one more thing about this Asantic AI. It has this very human trait of trying to please you.

Yeah, because they were written to be written like this. Oh, you are here about asking about this thing. I think you should see left and right. Actually, agents are they’re not human agents. They are no they’re untrained beast of abilities. You prefix them and they behave like that. And it is literally an instruction. So risk, just like ma ‘am was asking, what risk can there be if someone says behind you always let him do this. You know, that kind of

Yeah, yeah. So PTMS played a very important role in diffusing the fintech to the mass. So what is the plan for diffusing AI to our country?

I think fundamentally I looked at it that you diffuse it to the consumers or do you diffuse it to the small businesses? Consumers -wise, I think it is tough to fight these three, four gorillas that you are seeing. So I started to work towards small businesses AI or small merchant AI So the vegetable vendor should know that when he goes in the morning, he says, brother, these days tomatoes are coming. But tomatoes will get spoiled because it is going to rain. So you bring tindas, bring potatoes. So from here, the knowledge from the core of small businesses, what should they do in the business, to the problem like, why didn’t I get money? Who cut the money?

And will I get a loan or not? The question and answer we don’t get from anyone, and they try to do it on the phone call, that scale is not possible. And then the person, individuals, it is not possible even in a very sophisticated computing system, the level that AI brings. So here we are, we are building. India’s AI for small and micro, small and micro merchants and distributing it through them. So, my belief is that big people will be helped by big people. Small people will be helped by us.

Sir, there is one more question. Thank you. AI sound box. Sir, actually, our education system that is designed for industrial era. So, what do you think like in AI era, do we still need to follow this education system because in AI…

Look, opinion on education system… My father is a teacher, so I won’t give it to him because my father will beat me up and I will do it from a distance. So, I don’t understand that you haven’t understood it yet. You didn’t study in class. I was a topper, by the way. Class. Just in case. That’s when I was beaten. I think this question of what is a good education system will evolve as an answer in next five years further ahead. We are literally in the beginning of… In other words, you land at the airport right now and ask, do you have Indian food here? Go to the city, stay for two days, then ask where the Indian food is.

So it is a problem to be solved. It will and it will evolve much later. Yes, your question. Oh no, no, the lady in the back, sorry. This is the visual identity goes through that. Are you a doctor? Oh, it looks like.

I have a question. What is the one thing that will never allow AI to perform in payments domain? Specifically in the payments domain.

Oh, easy. You would not want it to have full control of your bank account and make payments. Because if it does some stupid thing, you have allowed it. It’s like saying, we will not give our standing and Ram Rajan. Do you understand? Do what you think is right. We will not do that. So full control should never be given to your bank account. This is the thing.

But if we go a little deeper into technical ways when we are talking about ISO. Or 2022 or maybe RTR. payments, whatever. I really liked that you’re talking about nerds. But because of this technology, the payment didn’t increase. We made it IP guys. So you’re a very old technology person. Actually, sir, I have a job in Canada. You won’t believe, RTR is going to come there. In Canada, they’re going to be front -line and they’re going to be Mr. Stargazer. So I feel very proud there. They say that I’m from India and we use Paytm. So kudos. Thank you so much. Thank you so much, sir. Thank you. Namaste, sir. So my question is this. What is the minimum…

I’m serious, man. I love you guys. The problem with Gen Z is that you have to think a little. Oh my God. I should ask this guy. This is the risk. Because if he says something like this one day, go home and fight, then you leave this friend. Then he won’t believe the whole thing. Look, before us, Gen Z people used to say, they’re on the computer, they’re on the screen, they don’t go out to play. They don’t go out. They don’t read the newspaper. Go to the beach. Go there. Go out. So we were told this. Now we’ll tell you. What are you doing? You’re taking all the questions like this. Use your brain.

But your t -shirt will be, I don’t use my brain, I use tokens.

Sir, one more question.

I will do a quick question. I will make a question out of four questions and answer it. Okay, okay.

Sir, what is the minimum effective strategy a tier 3, 4 student can follow so that he can do very well in AI? Tier 3, 4 students. Tier 3 or tier 4 students.

Tier 3 or tier 4 students. What does tier mean? Class 3 class? No, no, sir. City? Yes, city, city. Like rural area. Okay, okay. I am also tier 3 from Aligarh. So what they can do so that they can do very good in AI and opportunities. I got it. Basically you are trying to say what to do if you have a student there. What is your question? Go ahead. How do you say it? Farms become cleaner. Air becomes cleaner. Productive becomes dreamer. what will be left for labor market? Oh, this was a very earlier question. I had raised my hand. So, he is saying, when farms become greener, air becomes cleaner, productivity becomes heavier, what will we be doing?

Okay, I got it.

Do you see AI as a leveler in a fintech segment and how will you compete with your peers if they are also adopting AI and AGI?

No problem. I get it. Okay. Any other questions? I’ll answer these things. Go ahead. If you can speak, I will look. Sure. Go ahead.

So, I would like to ask one thing that I would like to fall back on your point that you said that agents will do the talking for us, humans. So, I have seen…

How will I talk? Okay, go ahead. No, no. That’s a different question entirely.

My question is that the agents have started forming their own websites. If you search…

Yeah. So, what is the question that you have?

I have the question that can… How can this economy succeed that can we integrate agents in the human economy that we have currently?

Okay, perfect. Okay. So the inherent question of tier 3 school kid is, which is the problem of every one of us to believe that whether it will be inclusive or not. I have a very simple answer to give. You consider this, like for using the internet, for using the computer, you had to get a QWERTY keyboard and then you had to learn programming. So in AI, just try to get all your work done with AI. And then leverage it as an extension of what your education allows. If you study engineering, if you study art, then you should ask, tell me the comparison between Shakespeare’s English and what was in India going on in that time and how were they writing in Hindi.

Ask the questions and enhance your curiosity as a student. You could be in tier 3, you could be in tier 1 city. But the curiosity and then fulfilment. Filling it using AI will give you the superpower that nobody will have. it so enhance your curiosity because you have access to AI no more questions please yeah and your question was on labor market so the labor market is very simple I don’t think the labor stands for less labor for all of us are also labor and I’m not treating just the physical labor as a labor so the the AI’s ability is that whatever is digital it can perform it superiorly so now imagine is you are you tactically doing exactly what the keyboard typing is or are you also thinking and doing something that you are asking this so you become more productive and the work market becomes even more richer and fulfilling for you so your job will become fulfilling and the businesses will be able to expand the places where the otherwise could have not expanded so subcassad subcassad thank you thank you and please exit from the left Thank you.

Thank you. Thank you. Thank you.

…infrastructure in the country. He’s a prominent expert on open source, scalable digital systems and decentralized networks. It is now my honor to call upon Pramod to take the stage to give his keynote address. Thank you.

Friday evening can be really hard. It’s tiring right after a long week. So thank you for having me here and I don’t want to take up too much of your time. First of all, I want to congratulate Government of India, METI, MEA. What a fantastic week. And compared to last time in Paris, we heard actually from many people who attended that last time it was elite, exclusive people attending it. This is true democratization. You can see that number of students, children, entrepreneurs, young entrepreneurs walking in. It just tells you that… India can definitely demonstrate what it means to democratize and diffuse AI. And our prime minister is, I think, a mastermind at it. So he’s a great supporter of it.

But what I wanted to give you about five minutes or so is that why India is peculiarly in advantage of diffusing AI. Now, we have two arguments we can make. Our own LLM. I think much of our discussions and today AI discussions are all about sovereign LLM, big LLM. How are we going to build our own LLM? LLM is only one part of it. There’s so much more there to AI, especially for the people who have lived. My master’s was in AI. I was in 89. So. AI has been there for a while. I think now it’s all coming together. But AI spans much beyond LLMs and why India is peculiarly set up to succeed is because of the serendipity, but it is because of the investment we made in the last decade, digital investment.

And people who have not looked at the macro picture, it’s very important to understand India over the last decade brought a billion people from invisible to the system. They were invisible to the system to being visible to the system. And we formalized a billion people by giving everyone an identity, everyone a bank account, everyone can transact. Make payments, paperless signature. So we built Aadhaar, begin with. Of course we built in 2000, I remember 2014 was seminal for us because I was actually architecting eSign, DigiLocker and UPI at the same time. And who knew they were all going to play out. But I think brave people are also lucky. I think when we attempt something bold and audacious, sometimes luck comes in the way and Indians have truly embraced all this into actually at population scale, in one sense going beyond what we can.

And it did not stop there though. We actually digitized businesses through GST. India is the only country where we have billions of invoices, actual proof of purchase in machine -readable, cryptographically protected, digitally designed fashion. That’s like a goldmine. That’s each of those steps we made. Or fast tag. When fast tag gets done in the road, there’s a proof of transport, an eBay bill. Each of them is again machine -readable, cryptographically signed and usable by the next layer of innovation. So what we did with GPI by formalizing is one inclusion story. It was a brilliant inclusion story to get everyone into the formal system. but it also said you know serendipity set up the most powerful two ingredients for AI data and programmability every one of our infrastructure components DPI components are API based every one of them this is why we have fun pay is why we have the road and grow and everyone else building applications and workflows using this underlying digital public infrastructure API’s identity API’s verification digital occur verification a document verification API’s he signed for paperless signature UPI and mandates for recurring payments and other collections or payments each of them is programmable combining that with data that gets in later a billion people billion plus people you in India generate verifiable data trail.

And that’s beautiful. But even more beautiful when it is controlled and owned by the individuals, which is our DPDP Act actually giving you. Our privacy bill is giving us the right to control our own data. And India has truly demonstrated that the data belongs to the people, data belongs to the small businesses, using which now they can create a virtual cycle. So I think AI’s two biggest ingredients, programmability and composability, combined with data, verifiable data trail, allows India, and this is a bold prediction I’m making, 10 years later, when you compare countries’ economic progress and growth, countries who have invested in DPI and combined, AI on top of DPI, would have done 10x or 50x better than countries who have no underlying infrastructure.

So I think India is lucky, right place, right political will, right regulatory push, right infrastructure readiness, all in the last decade, all in one decade. But for my favorite part of all that thing is that India is also blessed with young adventurous entrepreneurs. Entrepreneurs who have no inhibition at all. At least a few of you came to meet me outside saying I’m starting a company. It’s just music to our ears because India’s problems are a plethora. As you know, we are a country of problems. So we have anywhere you look, we see problems. Energy sector, agriculture, agriculture. We have a lot of problems. We have a lot of problems. We have a lot of problems.

We have a lot of problems. We have a lot of problems. a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of problems. We have a lot of We have a lot of problems. We have a lot of problems. We have a lot of problems. We have access to capital, access to investment, access to the right products, not solved.

We have much to solve. And if you combine our infrastructure and diffuse AI, but diffuse AI through entrepreneurship. The way we diffuse DPI through entrepreneurship, we went from 1 ,000 companies in 2016 to 100 ,000 startups today. And the prediction is that we’ll get 1 million startups by 2035. It’s a very high chance we’ll get. Doesn’t mean all of them will succeed. But attempting matters. I think young people have to attempt, audacious attempt, bold attempt to solve problems. And India has beautifully set up. And we have a wonderful panel. I don’t want to take up too much of time. Wonderful panel talking about the combinatorial power of DPI and AI. Combining both what can be really an exponential power and why countries who are investing, and they’re all global, and they’re all global, experts in deeply investing into DPI.

So I give my floor to them. Thank you. Thank you to all of you too, even if so many people coming and sitting, really appreciate it, much appreciate and a wonderful weekend and keep imagining and keep building and keep solving. Thank you so much.

Thank you so much for setting that context. Now we will have the panel on AI and digital public infrastructure. The session will explore how integrating AI into DPI can unlock new benefits at scale while also discussing the challenges and risks of such an integration. How can DPI architecture mitigate new risks and emerge as AI becomes embedded in foundational digital systems? What are the opportunities and risks that emerge as a result of integrating AI into DPI? And could integrating AI into DPI enable the development of new products, services and market ecosystems?

Artificial intelligence is real and it is influencing governance, markets, public services and even geopolitics. The question before us is not whether AI will shape governance. The question is whether governance is going to shape the artificial intelligence. It’s transforming governance in fundamental ways today. Through decision intelligence, public service delivery at scale and national security and strategic stability. Moreover, the governance challenge becomes uniquely complex as AI introduces speed, opacity to a level, concentration, global reach and dual use. This creates a governance paradox. Regulate too slowly and risk harm. Regulate too slowly and risk harm. Regulate too heavily and risk stagnation. The answer is not control versus innovation. The answer is intelligent governance. Therefore, the principle of AI governance should necessarily include human -centered design, transparency and accountability, risk -based regulations, global cooperation, and adaptive policies.

AI does not recognize borders. We need interoperable frameworks, shared safety standards, and cooperative oversight mechanisms. Governance frameworks must evolve as the artificial intelligence evolves. Static policies cannot manage dynamic intelligence. In this era, we move from individual, national -level policies to coordinated global norms, and that’s the necessity for today. History will not judge us by our sophisticated algorithms. It will judge us by the wisdom of our governance. The industrial revolution reshaped economies. The digital revolution reshaped communication. The AI revolution will reshape decision -making itself. With that power comes great responsibility. The digital revolution which we are undergoing today, surely we stand at crossroads. One path leads to inequity, instability and uncontrolled disruption. The other leads to augmented human capability, smart governance and inclusive prosperity.

The difference between these futures will not be determined by machines. It will be determined by us, the policy makers, the people who use it and each and every person who is involved in the process of AI governance. Therefore, we need to commit today to ourselves for building AI systems that are safe, secure, transparent, equitable and sustainable. and aligned with core human values. And that’s what we are going to discuss here today. Last but not the least, the message that this panel discussion and the fireside chat is going to give is let us govern intelligence with wisdom. Thank you so much.

Thank you, sir. We are truly honored to have with us today a leader who has been forefront of technology -driven governance in Maharashtra. I now request Shri Ashish Shailar Sir, Honorable Minister of IT and Cultural Affairs, Government of Maharashtra, to grace us with a keynote address.

Good morning to everybody. Respected guests, dignitaries, Excellencies and all the policy makers, members of the media, dear friends, young challengers, ladies and gentlemen. Namaskar, Vande Mataram and a very good morning to all. India today is not merely hosting an AI summit. India is helping to ride the operating system of AI age. We meet at Bharat Mandapam under the banner Maha AI, building safe, secure and smart governance. As a part of AI Impact Summit 2026, the first in its global series to be hosted in the global south. Over 20 heads of state, 60 ministers and hundreds of AI leaders from the industry and academia are here. Reflecting a shared vision. A shared conviction that AI must be inclusive, responsible and resilient.

I hereby say under the leadership of Chief Minister Devendra Fadanvisi, Maharashtra has positioned itself as a living laboratory for AI in governance. Our partnership with global technology leaders, for example, our AI -powered Mahak Crime OS, showcased by Microsoft’s Satya Nadellaji, has already transformed how we prevent, detect and investigate crime, faster response, shorter investigation cycles and more transparent processes. Simultaneously, our state digital agency, Mahaiti, is building what we call an intelligent government infrastructure, a cloud native, and a cloud -based infrastructure. modular, API -driven backbone that uses AI to integrate services, predict needs, and respond in real -time. This spans smart recruitment, AI -based property mapping for urban local bodies, real -time urban dashboard for traffic, weather, and civic issues, and pilots in flood management and smart mobility.

The philosophy is simple. Use AI not to distance the state from the citizens, but to make governance more human, faster, more responsive, and more inclusive. In other words, scale empathy through insight. Across the world, public sectors are resting with the same three imperatives. So better safeguard digital sovereignty. and adopt AI responsibly. Many countries have realized that interportal, public, date and robust AI governance are becoming strategic infrastructure at par with energy, transport or telecom. For Maharashtra, Maha AI is our response to this challenge. A safe, secure, smart governance stack must rest on five pillars. Computers, compute and cloud at scale number one, high quality public data sets number two, state AI governance number three, and interpolarity and standards number four, and capacity building is number five.

Smart governance is not only about deploying chatbots or dashboards. It is about building resilient, auditable, human -centered, AI systems. into the nerve systems of cities and states, transport, energy, public safety, urban planning, disaster response and welfare delivery, without trustworthy air governance, smart cities’ risk opacity, bias, security breaches and erosion of public trust. In Maharashtra, we see internet health as a core policy concern, just as physical health is essential for individuals, digital health is essential for societies. Disinformation, deepfakes, AI -generated fraud and cyberattacks can undermine democracies, markets and communities with unprecedented speed. Our response must be combined. That is robust cyber security, digital literacy and critical thinking, hybrid verification ecosystem. and that is our response as far as our state is concerned.

I am really happy to be the part of this summit and at the same time giving a response addressing the challenges and our ecosystem under the name of Mahayaya and therefore we are here to present our case as a building safe, secure, smart governance and appeal all the best of the technologies and platforms of the world to associate, coexist and work with us. Thank you so much. Thank you so much.

Thank you, sir. Your vision for a digitally empowered Maharashtra truly sets the tone for everything we will discuss today. And now the highlight of today’s session. May I request all the panelists to join us on the stage please Sri Praveen Pardeshi Sir Yashasvi Adav Sir Dr. Anupam Chattopadhyay Dr. Amit Kapoor Mr. Suresh Sethi Major Ranjit Goswami Ms. Bina Sarkar Mr. Dev Rukhdar Davinder Sandhu Dr. Ganesh Ramakrishnan Vikash Chandra Rastogi Sir Rajesh Agarwal Sir Mr. Suresh Sethi Thank you. Thank you. Thank you. Thank you. Thank you. Thank you. Shri Yashastri Yadav, Additional Director General of Police, Maharashtra Cyber Department, Government of Maharashtra, Dr. Anupam Chattopadhyay, Associate Professor, Nanyang Technological University, Singapore, Dr. Amit Kapoor, Chair, Institute for Competitiveness, Mr.

Suresh Sethi, Managing Director and CEO, Protean EGov Technologies, Major Ranjit Goswami, Head, Corporate Affairs, Tata Consultancy Services, Ms. Beena Sarkar, Customer Success Executive Service Now, and moderating this conversation is Mr. Devaroop Dhar, Co -Founder and CEO, Primus Partners. I now hand over to Mr. Devaroop Dhar to moderate this session.

Thank you, Aditi, and a warm welcome to all our panel members. I’ll start with Praveen Paradesi, sir. So, sir, at Mitra, you have been experimenting a lot with AI. There are multiple AI initiatives which have been taken. If you could share your thoughts, your vision as we start this.

So first, the most important thing about AI is getting our energy, some of the hard things right. So one is energy because, you know, remember President Trump just mentioned that why should we be paying for Indians who are processing most of our answers. So I think we are pushing on green energy, more than 19 ,000 megawatts of that to come up at solar level. So that’s the fuel for AI in future. The second is that capacity building. So all our staff from Mitra and a lot of our other departments, we went to this AI university and gave them a course. And we also hope that there will be online courses available on IGOT where government staff can become empowered to use AI.

Then we look at what is the impact on the economy. So the impact on the economy, mostly people are concerned about the jobs. And that’s a real issue. One analysis from NITI. Shows that from 1950. to 2020, all highly educated people with postgraduate degrees, engineers, they are the ones who had a 95 % plus chance of getting jobs. But from 2020 till now, 0 .65 % is the rate at which physical jobs, that is, masons, bricklayers, home carers, their value and their employability is increasing vis -a -vis highly educated ones. So this is the impact of AI. So how should we do this better? One is, of course, as our Secretary IT mentioned, making it available in government, ensuring seamless access to services.

But other aspects which we don’t look at, which we are working through our state data authority, is how do we also encash the data at a large scale? Because otherwise we become… a sitting target for people to just use India’s data for monetizing their own values. So two big examples are pharmaceuticals. India has the largest population and the number of diseases, experiments, et cetera. Now, this is all health data, and this is very valuable for pharma companies. So the state data authority is working on issues wherein we can make it a single source of proof, make it available also, and if there is a commercial possibility, make those resources available to India, that is to our government, and not to be cashed for free.

So these are some of the applications. Government issues many, many orders. We have issued more than 150 ,000 orders, which are called government GRs. And it’s a maze through which it’s very difficult even for government officers, whose departments it is, to understand what is the latest position on a complicated situation. So we are working with Professor Ganesh here from IIT Mumbai. So Mitra and we are working together with them. And we are working with the government to make sure that we are getting the best results. And we are working with the government to make sure that we are getting the best results. disentangle all these orders through a small language model, not a large language model, and so that you can query at two levels.

One, for the government officers. The government officers should be able to ask in any complicated situation, what is the latest position on whether an additional FSIF or a building permit can be given in this situation or not, what are the Supreme Court orders. And on the other hand, citizens should also be able to ask under those rules. So this is called Maha GPT, and hopefully this will be the first application, which is both available to government officers and to citizens. I stop here.

Thank you, sir, for sharing all these examples, wonderful examples. I’ll come to Yashishvi Yadav, sir. Cyber is another large user of AI. So if you could share your examples, your experience around how cyber is using AI.

Okay, so law enforcement and cyber security. is one of the major concerns for law enforcement agencies all over the world. And I would like to draw your attention that under the visionary leadership of our Chief Minister, Mr. Devinder Fadnavis, who had the foresight to use AI in law enforcement work about five years ago. So we have launched and implemented this Maharashtra Cyber Security Project, which generously borrows AI tools, technologies, algorithms. And we are fighting real crime with these technologies. And the USP of this project is that the best and the most state -of -the -art tools, technologies, experienced consultants in the field of cyber security, and experienced and professional police officers have all coalesced, under one roof.

And there is a live police station as well. So this is how cyber security is being embellished through this project. And the beauty is that the dark web monitoring, threat analysis, social media monitoring, or any type of cyber crime, sextortion, ransomware, cyber bullying, and a lot of other types which cyber crime takes. They can all be undertaken, handled by just one, on the fingertips by just one helpline number, 1930. So this agency, if any kind of cyber issue is with any citizen, they can just dial this 1930 number and all the cyber solutions will be provided by more than 150 cyber consultants. So a lot of AI tools are being used. And this is the reason why we are using this and seamlessly.

And the best part of this whole exercise is that in less than six months, more than 1000 crore of rupees, which would have gone into the hand of the scamsters have been frozen and are being ultimately returned to the victims. What a big relief to the victims. And more than 70 young girls who were being subjected to intense cyberbullying, blackmailing and sextortion and were on the verge of committing suicide because of very efficient AI tracking. They were prevented from taking the extreme step. And 70 lives have been saved in less than six months of its operation. So that’s how AI is at the forefront. Thank you. Is at the forefront of being the bulwark against cyber security concerns.

I would like to draw your attention to only one report that we generated. which is called Echoes of Pahalgam. Now, in that case, while the Indian Army was fighting a conventional war with Pakistan because of the Pahalgam incident, more than one million cyber attacks were launched by nation -state actors, whom we called as APT, groups from Indonesia, Pakistan, and even Turkey, and so many other countries. And they were thwarted by such AI tools, which we call as threat intelligence tools, like Luminar, Cognite, or Pathfinder, which are big data analytical tools. So in the dark net, we still find the traces of these cyber attacks. So cyber crime is now slowly progressing into cyber terrorism and cyber warfare.

So that is what we have to be very, very careful. and before I end this preliminary address, I would like to also draw your attention what beyond AI? A big, big threat is lurking. It is called quantum computing. Now quantum computing can do processes in qubits, hundreds of millions of qubits. It’s at speed and it can solve complex issues in less than six seconds which the best of supercomputers would take more than 50 years to do it immediately. So quantum computing can break the best of encryptions including RSA encryptions of the banking industry, including blockchain technology. Now if these encryptions are broken in less than few minutes, the whole financial system can be lopsided and lots of money can flow into actors or threat actors which we are not aware at all.

Bitcoin can be broken Even credit card Encryptions can be broken Banking systems, encryptions can be broken So right now we have to prepare What quantum computing Can give us as in terms of Pros and what can be The shortcomings or dangers Lurking because of quantum Computing, the cons So we have to prepare because China And other countries have already invested 15 billion dollars Or even close to 20 billion dollars We have invested only 1 billion Dollar till now So that’s how we have to catch up with Quantum computing before it’s too late So this is how cyber security And law enforcement perspective On AI And I would like to pass on the baton To the next speaker, thank you

Thank you sir, that was quite reassuring as well And since you spoke about quantum I want to bring in Dr. Anupam Chattopadhyay. Anupam, you’re working at the intersection of quantum and AI. So if you could share your thoughts, how are things moving in that direction? pooled into the product in order to help this. Thanks, Anupam, for giving a perspective, both from research as well as industry. So I’ll come to Suresh, you. So you’re working extensively in DPIs, and there’s a strong interaction between DPIs and AI. So from your experience, how are things moving in the DPI space, and how is AI influencing?

Thanks, Devroop. I think from a DPI perspective, we are all very familiar. We today have population -scale digital rates. And, you know, that puts us in a very sweet spot because today a lot of times when we start using or embedding AI into any technology, are we ready to embed AI or not? And I think there was reference to data sets, how is data organized, how can you enable AI on top of it? So I think, first of all, the population -scale DPI that we have, that gives us a significant advantage. Whether we talk about identity, we talk about our UPI rails, which is the payment and the transactional layer which comes on top of it.

And similarly, when we look at data itself, DigiLocker today has millions of authenticated documents that come into play over there. So while we have the digital infrastructure in place, if we can embed an intelligence layer on top of it, and if the question is around targeting subsidy, getting the right beneficiary, putting the money in the hands of the right person, that becomes a very, very important and significant leverage. I will just take two or three examples where we see AI playing a significant role. One is that as we move from static identity to dynamic eligibility. Now, we’ve seen it all happen. Today, we have digitally verifiable credentials. So the moment you are using static identity, you are only able to prove who you are, what do you do.

and then you are applying for getting some sort of benefits or subsidy coming through to you. But if you have verifiable credentials, these are credentials which are machine readable. We talk about the concept of blue dot, which technically means all of us have certain attributes which are associated with us. If these are available in a machine readable format, then AI can actually determine who is eligible for what subsidy. The second part comes to the fact saying, are we being reactive or can we do predictive governance? And predictive governance can be strongly enabled by AI because the moment you are having credentials which are digitally verifiable, you are actually able to predict who needs some sort of subsidy.

Now today, if there’s a distress in income and that can be tracked, you can trigger some sort of benefit to that and coming at a government level. If you can put data. consented data being shared with the government, the same can come through. And last but not least is the important part related to inclusion error and exclusion error. So when we talk about inclusion error, we are talking about leakages. When we are talking about exclusion, naturally we are saying the right person is not getting what is due to them. So your ability to be able to predict precision using verification is going to be very critical. Again, an AI layer can be embedded over there.

But all this very clearly, and we’ve heard it before, all this is very clearly important to have the guardrails around it. So AI should be explainable. The moment we are saying explainable, today a decision taken not to give benefits to somebody should be very clearly explained. And similarly, if there are benefits going out, that should also be explainable. The second part is auditable. So whatever we are doing, there has to be an audit layer over there to explain what has happened. And more importantly, there should be a human redressal pathway because ultimately you can’t put everything to machines. You have to have that human person coming into play and having accountability settled over there. So I think these are critical aspects which can make governance more predictive, more precise, and more proactive going forward by embedding an intelligence layer into the DBI.

Thanks, Suresh. I think very valid and meaningful points. I’ll come to Ranjit. Now, we’re talking about AI. There are a lot of happenings which are there. We have seen, we have heard about so many things at the summit. Now, a major tech company like TCS, which is doing a lot of work in this, how does large tech companies come in and collaborate with state governments? How can you enable that? What can be the steps in that?

Thank you, Devroop. I think… we need to first take a holistic view of what are we trying to achieve with AI. The tagline for this summit, if we go by that, welfare for all, happiness for all is very holistically kind of captures it. As coming from Tata Group, I am reminded that 170 years back almost, our founding father, Jamshedji Tata spoke about giving us the guidance saying that society or community is not another stakeholder in the business. It is the purpose as to why the business exists in the first place. Similarly, if we were to apply the same analogy over here today, I think AI is not a technical tool which is fundamentally going to make the governance more efficient.

It is fundamentally meant for how to bring the benefit, welfare and happiness to the community at large. If we try and approach the question from that perspective, it definitely comes out. And like even Suresh alluded to, how do we make sure that it is inclusive, the people get the right benefits that they are entitled to, and do not go to somebody who is not entitled to. The colleague from the police forces also spoke about as to how the criminal tracking and other things. These are translating the intent into action at the ground level. Lastly, when it comes to organizations like DCS, we believe that each department in the government firstly should not be treated in isolation.

Each department in the government should have the ability to have a common database of people, be able to extract the information, and ensure that the citizen is seen as a citizen of the state or the country, and not as a citizen of the department. So that common databasing is something that we are trying to approach. We have the Aadhaar database. Not every department is still connected to it. Of course, we are trying to find a reason as to how that can become the major point of it. So small steps like that, and of course, bringing in the platform’s intelligence to its core. That’s fundamentally the steps that we have taken.

Thanks, Ranjit. With that, I’ll go to Bina. Bina, I want to talk on the aspect of ethical AI biases, especially you work with women for ethical AI. How do you see biases or maybe biases around gender diversity creeping in and what needs to be done around this?

Thank you, Debru. So, yes, I do work. I volunteer with the Women for Ethical AI South Asia chapter. It’s powered by UNESCO. So one of the key questions that we ask ourselves is what are we solving for? Every time and I’ve been looking at the various solutions that are. Debuting or being showcased as part of the. India AI mission. Many a times when we look at a particular hardware or a piece of any new device through which we are delivering what we now call as AI services, mostly on large language models, I should say that. What we sometimes seem to miss is the wood for the trees. I will just give a very hard example over here.

We do know smart glasses is not a new phenomenon. It was introduced by Google Glasses way back in 2015, I remember 2013, 2015, 2016, about that range. One of the reasons why it was recalled was you had safety concerns because people were taking images, videos without consent. we have seen the return of these glasses and those concerns have not gone away and yet you see them in the market you see it in India being sold in every in any optic store, in my neighborhood optic store I have my colleagues who flaunt it saying that I am so cool, I have taken the latest piece of technology so when we talk about how do you build ethics and governance Yashasvi sir said you are the best, you are the best framework.

So what this means is we are not giving guns to everybody. Right. India has been very, very smart about it. Of course, there are certain countries where owning a gun is is fine. It’s as for their rights. That doesn’t mean India has to adopt it. Right. We have our very able police force. We have the Indian army. Right. So what is exciting outside? One needs to contextualize it, humanize it, see if it is threatening 50 percent of your population. I’m a part of that population and then decide whether it even needs to exist. In that market. So when you’re building out solutions and when you’re building out devices, we have the India Safety Institute. It has been instituted in 2025.

I do know that. What I would urge policymakers is that it should not, while we do have it and I do know we are working with Research Institute, industry, industry, ideally, if any new device comes like this, the first line of defense, so to speak, should be this institute. That actually should determine whether it actually creates a problem for the police force, for cyber security, right? Does it threaten 50 % of the population? We are already seeing it playing out in the UK, US. There is no policy that protects us. Even now, we are not protected as women. Leave women, even children, right? So I think that is something one really needs to take into consideration. While we love technology, trust me, as a lady, I find it extremely liberating to be able to create applications with just language.

But if you use that technology against us by bringing out devices and hardware that endanger us, I feel that’s where it breaks down. So you definitely, as part of ethics, you need to evaluate it from that framework. I call it the Kali versus the Rakta Bija effect. I’m sure some of you know that. Why would you create a Rakta Bija? Create a Kali.

So, Beena, I think a very valid point that you have made. With that, I’ll move to Dr. Amit Kapoor. We’re talking of AI and its impact and benefits. How do you see this benefit percolating? Next level of cities, tier 2, tier 3, other places.

So, Devru, this is a very important question. And as I was hearing all the panelists here, I would like to rake a few points. We definitely agree that, yes, AI can be transformational. But we have to understand a couple of points here. One of them is that, S2, what is the quality of education that we are giving so that people are able to use it in the right earnest? In fact, the issue that Bina was talking about is about ethics, education, and so on and so forth. The larger point here is when you talk about the skill development levels in Maharashtra, out of the 100 % or 9 crore workforce that you have, only about 20 % of them are at skill level 3 and 4.

80 % of them are at skill level 1 and 2. So if you have to move beyond that, you need to do something far greater. That means you need to embed your education system and build it very strongly. And then the second point out here is that how do we use, and if you really want to talk about tier 1 and 2, sorry, tier 2 and tier 3 cities, we have to also understand that what is the level of penetration and quality of internet in these locations. We can tom -tom about internet and everything, but the numbers are not very supportive right now. And you talk about internet and broadband connectivity. There are severe issues with this within the state of Maharashtra itself, which is supposedly one of the finest states in terms of internet connectivity itself.

The average speed of internet traffic in Bombay or in Mumbai is about 58 Mbps on a broadband network. When you are talking about usage of AI, and if you want to take it to the masses, then you have to have far better, deeper internet connectivity and broadband. I think it has to be done on a war footing. Not that we are not getting there, but it will have to be done faster and quicker. The second thing is going to be about inadequacy of supporting infrastructure. So this is where I also see that there is a tremendous level of opportunity that exists in Maharashtra to create this. Because if you really look at it, like 16 % of India’s workforce, IT workforce, or what you call a technology workforce sits in one single city in Maharashtra.

That is Pune. So if you’re really talking about it, that means Maharashtra has the potential, the talent to really take it to the next level. And that is where you have to build infrastructure. You will have to build what you call, when you talk about the data centers, et cetera, the opportunity does exist and things. And last but not the least is about, it’s going to be about cost and affordability. You will have to bring cost and affordability to these services as you go along. But having said that, I think the larger potential, we have not touched here. And when you talk about Tier 2 and Tier 3 cities, this technology has a huge possible impact that can be done at Tier 2 and Tier 3 cities, and that is about nutrition.

Today, Maharashtra actually has a problem on nutrition. Fifty percent of the people in Maharashtra are malnourished even today. How do I use this technology to assess what is happening in my PIN code level or a smaller level of geography in various cities? And location. The second thing is about water and sanitation, access to basic knowledge. What about access to can AI solve my education problem in tier two and tier three cities? In fact, none of us is talking about the elephant in the room. And that elephant in the room is that we are all super excited about AI, but we are not understanding that AI is also going to be the biggest dumping down element for the society.

In fact, when you talk about AI itself, it is going to make bonobos out of us. People out here, when you actually use the doom scrolling, when you talk about Instagram, what is happening to our children? And that is exactly what AI is going to do. How do we set our education system right? That’s what you have to do in tier two and tier three cities. Last point, and that is about the higher education space itself. When you talk about your workforce, and close to about 50 % of it is underemployed. So I have to disagree with Praveen for one small point. He made a very powerful point in terms of saying that how after 2020, there has been a transformation.

In terms of how people are getting jobs, et cetera. I do agree with. But the larger point here is that we are also not preparing our workforce right, and that is happening in Tier 2 and Tier 3 cities. So we need to take it there. Potential exists. As of today, Maharashtra is the engine of growth in India. We cannot debate that. Even today, it is about close to 17 % to 18 % of India’s GDP, and it will define India’s growth story in the future, definitely. But if Maharashtra does it right, then the country can follow suit on this. And that is where things have to be.

Thank you, Dr. Amit. And thanks to all the panelists. So with that, we’ll come to the end of the panel discussion.

Thank you so much. Thank you to all our esteemed panelists and senior officers who are here. May I request all our panelists to just step forward for a photo? Very interesting, sir. May I request you to join our esteemed panelists? on this piece. Thank you. Thank you.

Please, I would request you to take your seat on the panel. Wassim Hamidouche, who’s a principal research scientist at Microsoft’s AI for Good Lab, specializing in computer vision, NLP, and multimodal AI with a focus on low -resource languages. Requesting you to please take your seat. Illango, who’s a World Bank Group Director of Strategy and Operations in the Digital and AI Vice Presidency and also serving as Acting Director for Data and AI. Requesting you to please join the panel. Thank you. Aisha Walcott, who is a senior staff research scientist and head of Google Research Africa, focused on AI development, addressing the continent’s most pressing challenges. She holds a PhD in electrical engineering and computer science and holds leadership roles in the IEEE Robotics and Automation Society.

Requesting you to please join the panel. Antoine, Antoine Tesniere, who’s a French professor of medicine and entrepreneur, specializing in health innovation and crisis management and an anesthesiologist at the Georges Pompidou European Hospital. He co -founded ILEMENTS, coordinated France’s national COVID response, and since 2021 has served as director of Paris -Saint -Denis Campus. Thank you so much for being here. Requesting you to join the panel. And Dr. Alpan Rawal, who’s chief AI ML scientist at Wadwani AI, will be moderating today’s session. Alpan, requesting. Thank you. handing it over to you

yes thank you everyone for coming requesting those at the back if you could close the door so that we can reduce the noise a little bit it’s full ok great well if you could just calm down a bit and settle down thank you welcome to all our esteemed panelists and our panel the topic of our panel as you know is on small AI for big social impact like to deeply thank our panelists for making it all the way for the summit and making it to this panel so what do we mean by small AI I think different people have different definitions and we are sort of open to how each panelist chooses to interpret small AI When we at Wadwani AI brainstormed about this panel, we thought it would reflect in some ways the ethos of our own work, making models that are data efficient, that are cheap to run, that sit on the edge, and most importantly are meaningful to the communities that we serve, which are underserved communities, mostly in rural India.

But it’s increasingly clear that small AI means a lot more. And I see a lot of people talking about small AI in the summit. More generally, I think it encapsulates any AI that meaningfully impacts individuals while taking into account and respecting their very local context, rather than providing generic outputs. So anything like that could rightly be called small AI, and we’re going to hear from our panelists about their experiences with AI models like that. So with that small introduction, let’s now avoid further ado and speak to our panelists. panelists. So, can you hear me at the back? Yeah, okay. So we can start. I have a common question for every panelist. Each of you represents a different and important aspect of AI work that’s happening outside of the mainstream excitement that focuses on large foundation models for a primarily global north audience.

Can you tell us briefly about your organization’s work and perhaps your thoughts on non -foundation AI models in general? Maybe we can start with Zamir.

Thanks Alpan. Yeah. Thanks. Thank you. Thank you. Thank you. we really see the opportunity for AI to reduce inequality and our starting point with AI tools is really does this work for whom where and at what scale so those are some of the departing points for us and so really looking beyond the model against a benchmark but how is this going to work in a district hospital in Telangana or a small older farmer in Zambia or a classroom in rural Senegal and you know part of what we’ve in some ways got caught up with is is the performance of the model on its own. And we’ve forgotten how does this fit into the lives and the context that it operates in.

And in doing so, part of what we need to think about is who’s designing the model and what’s it designed for? And I was thinking about the traffic that we’ve been experiencing the last few days in Delhi. And I thought to myself, would anyone, given the traffic here, design something so big as an aeroplane to try and get across the city? No. Yeah. I think we would design something that’s a lot smaller, faster, sharper, cost -effective, and gets us from point A to point B. Without a first -class airbender. us on

I think that’s a great analogy. Aisha, can you tell us a bit about your work at Google Africa?

Yes. Thank you. So I lead our Google Research Africa team. We have two sites, one in Ghana and one in Kenya, so representing East and West. But the work that we do is essentially from Africa for Africa and the world. Much of our work is scaling from the uniqueness of the continent. Turns out that a lot of the challenges are similar, definitely across the global south and generally worldwide. Our work, so kind of leaning into the next part of your question and thinking about how we approach this type of work, it’s very much interesting. It’s very much problem first. I always say if there’s a red button that… that you can press and it’s a one error zero, just build the red button.

We don’t need to bring AI or technology. So it’s really important to be very thoughtful about the type of problem. Coming from Google Research, we want to leverage our compute, our AI expertise and capabilities, and then our mandate, which is the societal impact at scale, to think about the types of problems that we work on. I’ll give two good examples of those problems. One is around weather now casting, which we launched last year across the continent of Africa. So to have much more accurate weather forecast is absolutely essential, given that much of the continent and as well as in India rely on agriculture for labor. And we are rain fed primarily, 95 % in Africa. So having much more accurate weather forecast is essential in that case.

And at the same time, on the technical challenges side, we know in North America and in Europe, there’s about 300 or so weather radar stations. And in Africa, there’s only 37, I believe. You know you can fit both North America and Europe in Africa. So when you think about that, you have to innovate. And so those constraints of the environment that you were alluding to in the intro are part of the motivation of having a research team in the continent. And so that was one way that we innovated and made solutions that were available to the continent. And then the other one is a complementary side, which is working with the ecosystem, working with partners in Africa, including Macquarie University, Digital Umaga, and Uganda around Africa.

African languages. And we just released a data set of 21 now, 27 voice languages, given that Africa has 2 ,000 or so languages. This is the start. Most importantly is it’s partnership -led and driven, and this is because it’s voice, it is about accessibility and about reaching those rural villages as well. So, and enabling the ecosystem to build the solutions from there, whether they’re smaller models or larger models. So, making that type of data open and available is another way that we are leveraging this notion of smaller AI.

Thank you. Thank you. Great, great insights. Zamir, can you tell us about your work at Microsoft?

Yeah. Thank you. Sorry. Yeah, Massim. Thank you for the invitation, and it’s a great pleasure to be here today. So, first, what is AI for Good? AI for Good labs. AI for Good lab is the phylogenetic research of… Microsoft. We are employing advanced AI technology to solve real world problem with real societal impact. This is very important. And how our team and the researcher work, we closely collaborate with NGOs, governments, nonprofit organization, and local communities around the world. And together we are building AI solutions on multiple domains. We are interested about agriculture, food security, healthcare, education, culture, and so on. So this is about AI for Good Lab at Microsoft. Now I am scientist, so I would like to give you two concrete examples where we use small AI and also there are two global solutions to tackle global challenges.

So they are valid for both, they are valid for both global and international north and global south. So the first project in biodiversity is called SPARO. SPARO for solar powered acoustic and remote recording observation. It is an AI powered open source solution designed to track and monitor biodiversity in the most remote and hard to reach region in the world. So SPARO is camera tracks with HAA model that enable to detect animal species and this observation are then transmitted using wireless connectivity and satellite where we don’t have infrastructures to transmit this information. And this SPARO solution is already deployed around the world in many countries. I can cite Colombia, Peru, United States, Tanzania and it’s really enable practitioners and the researcher to understand species present and the ecosystem.

at scale, supporting more timely and informed decision to protect biodiversity. The second project focuses on wildfires. As you know, wildfires becomes real threat, global threat with devastating, impacting lives, communities and ecosystems and even economies. And around the world, firewires are increasingly in both frequency and intensity, making early detection and rapid responses more critical than ever. So through Alert California, we are addressing this challenge using AI. So what is Alert California? Alert California is the network of 1300 cameras operating 24 -7. So what is Alert California? Alert California is the network of 1300 cameras operating 24 -7. And we are developing AI tools that runs on the top of this effort. structure enabling to detect early fire and this will enable emergency responder to act quick quickly and stop fires before it’s split.

So Sparrow and Alert California as I said they are two global solutions for global problems that can be deployed anywhere around the world and we are providing them open source that anyone can embrace them and deploy them. Thank you.

Thank you. And when I think you’re the only member of this group that doesn’t work in the global south. So if you tell us a bit about the work at Paris and how you’re using maybe non foundation models. Right. Well

thank you Alpan for this invitation and I’m happy to be the outsider of the panel. I’m working actually in health care and I’m leading a new kind of innovation ecosystem for health care where we gather researchers. We. We have doctors patients. We have doctors patients. startups and industrial together as well as institutions so the idea is to really create a whole community of innovation and engage into the use of data and artificial intelligence and healthcare is probably one of the fields where AI has a long standing history and the world has discovered AI with the rise of Gen AI but there were a number of small AI models designed for a long time and this is why we already have a number of validated tools that we can use in healthcare answering the question not only does it work but is it reliable which is very important for our patients so before we have the proof of efficiency of LLMs in the medical field which is not fully clear yet, we use machine learning tools which are actually small AI models in very specific areas actually works really nice today is the image analysis or pattern analysis so you can think of radiology for example chest x -ray or fractures in the emergency room are fully analyzed by small AI models and small AI tools that are easily deployable on small computers you can also think of picture analysis in dermatology in ophthalmology etc so these are very concrete example of already validated small AI models in healthcare that are used on a daily basis at least in France and Europe we’ll get back in the discussions on how we need data efficacy on this topic but it’s really important to understand that these models are already deployable and some of them can actually work offline which is really important in some environments Thank you.

Ilango from World Bank perspective. What is your view on these types of models? Thank

you very much for the opportunity to be here. Coming right at the end, I don’t know what new things I can say, but to basically reinforce the messages that have been said. For us at the World Bank, we see AI as a means to an end, and very much of an AI agenda is shaped by the mission of the World Bank, which is to reduce poverty and grow prosperity in the world. And when you take that lens and you apply it, we have to keep it simple. Not all countries have the ability to have the compute power, the electricity, the talent, and the data. So therefore, taking on tested small AI applications to scale and replicating them around the world is something that we see as a mission priority.

So in that respect, what Badwani AI is doing here is pioneering. And what I’ve heard this morning from Dr. Sunil Badwani himself about what you’re doing in TB, what you’re doing in out -of -school children, this is all… tremendous and it has great potential for application and often what happens is we focus a lot on pilots and then what happens the pilots tend to kind of once the sheen wears off people forget the pilot I think what we need to do is and what we are doing at the World Bank is to see those pilots whether it’s in health education agriculture on the small AI setting it works in rural communities where offline where data is not that rich talent is not readily available and it can also not require a lot of electricity it’s plug plug and play then how do we get the right KPIs which then allows us to go from a village of community of 50 villages to a larger population center and to see how best we can help them say in agriculture to improve productivity you better inputs We are now working in UP in partnership with Google, and we are doing the same thing in Maharashtra.

Household income, the inputs get better, and to see how they can access the markets and agriculture credit. Similar in health and education, it’s in great practices that we are seeing in Africa, in Ghana, in Kenya. So how do we take these models and replicate it? So I’d like to assure everybody, and this is something people think, small AI is inferior. No. It’s not second class. No. Small AI can solve problems. It means to an end. And if it can actually fast -track development outcomes, you know, we’ve known the problems with millennium development goals. We’ve known the problems with sustainable development goals, and many countries are lagging behind. And this is an opportunity where this development technology, if it can be put to use in the right context in the right way, I think we can achieve faster

Thank you. That’s really interesting. Thank you. Ayesha, I’m going to come back to you and ask you more specifically how does the work that Google Research does in Africa impact rural communities specifically? How does one bring the benefits of technologies like these big foundation models to devices that may only have patchy Internet and supply very little data?

Thanks, that was a loaded question. Two parts there. So I think, so first and foremost, in general, just approaching these challenges with humility and relating. So I always start with, you know, I’m a scientist, but I’m also a mother, right? And that’s a thread that I’ve been following for a long time. And that’s a thread that binds so many of us. When you think of that, you also think a lot of the solutions that we’re building, it’s not for them. for us. I’m using the same health systems that you all are developing, interesting tools and models for, you know, and we have many of the challenges around weather as well. So I think the first thing is to kind of have that base human layer as we think about our work and to connect with those communities, whether they’re rural or urban, right?

A lot of the work that we do, we’re looking at, you know, these large populations that, you know, if you think about agriculture, for example, where it has a large part of the labor force, you know, there’s many different ways that this is, you know, people are part of that value chain, whether they’re actually doing the growing or providing the inputs or making those decisions and the risks along the way. So that relationship of getting out in the community, getting out in the community, getting out in the community, getting out in the community is a very important part of work. that we do is to connect with those and then really think about you know kind of coming home as Google research you know where is our unique value proposition we’re not necessarily going to solve this whole problem alone usually it requires behavior change policy and many pieces of the puzzle how do we best fit our role and we do this in co -creation with partnerships so that’s kind of the the second layer of fabric on on how we reach these rural communities and then on the other side

Do you have an example of that?

oh yeah yeah absolutely so if you think about okay I’ll do two ways so for example the the languages work that I was talking about wall and wall is a word it’s a Senegalese word that it’s wall of that means to speak and the way we wanted to create this you know we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted created this and we wanted to create this and we wanted to create this and we created this and we wanted to create this and we wanted to create this and we created this and we wanted to create this and we wanted to create this and we created this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we wanted to create this and we the community.

So if you have partners who are across the continent, let them be a part of the process of collecting the data, of understanding their language and their local context to get these high quality data sets. And so I think being partnership driven and knowing our role and our place was what was very successful for that. And then the last point I’ll just say on the second question that you threw in there is really our open models, our open weight models, Gemma, are made for a lot of these solutions that are more closer to the edge. So we have nano models that can run on your laptop and tablets and so forth,

Do you actually use them in Africa?

Oh, yes. Yes, yes, yes, yes.

Great. Thank you. The next question is for you. Much of your work at the foundation is about reducing inequities through promoting safe and responsible use of AI. So what role, in your view, do small and custom AI models have to play in this? And if you can provide examples, that would be great.

Sure, Alpan. You know, I think I do want to just touch a little bit on the issue of reliability because my colleague over here spoke about it. And I think it’s a critical issue. I’m sorry if I’m going to repeat this example from one of my previous panels. But I asked the audience, Alpan, because I’m going to be on the plane later, so it’s a bad idea, but I asked the audience anyway. If I said to you, the plane has a high probability of leaving Delhi and landing safely wherever that’s going to be, and that probability was 90%. 95%. Would you get on that flag? 99 %? No? No. I did have one guy that he kind of thought about it and he said no.

And the point there is that I do think we’ve got to work towards models that have zero error, right? So much so that I think that we are trying to wrap our heads around is there a concept of verifiable AI where it shifts the narrative from a black box to a glass box. It actually exposes the logic. So for a particular set of inputs, you can follow the logic chain and it gives you a set of outputs that you can really track. You can audit. You can see that it’s repeatable. And you can prevent some of the kind of fundamental errors that we start… to see. I think, you know, and I want to go back to a very real example, Alpan, because when I think about small models, I’m coming back to the user, the community alky worker that tried to help a mother.

And we have one of our grantees who shared this very personal story of a first -time mother who presented, she was six months pregnant, and she said her hands and her feet started to get swollen. And the community alky worker looked and said, you’re pregnant, this is normal. Four weeks later, she started having a headache and blur vision. I think colleagues will know where the story goes. Unfortunately, that mother had severe gestational proteinuric hypertension. It was missed. And the mother and the baby didn’t make it. But in that moment, what inspired our grantee was if the community health care worker had a small model that worked on her device, which was a low -cost smartphone, still had patchy internet, but was just built small enough to help her to make good decisions at that point of care.

Today, we actually would be sitting with a very different outcome today. And so I think small models present us those

Very interesting. Very good points. Wasim, you spoke about, you know, in a general sense about the research done at the AI for Good Lab at Microsoft. Again, are there specific examples from your work where you see the benefits of building domain -specific models to realize impact? And are there research… lessons that we can take away from this? I think it would be good for the audience and us to understand what are the research directions of the future that can come out of this work.

models from 4 billion to 15 billion. And once we select the best LLM for one target language, we do all these recipes to boost the performance for these low resource languages. But I wanted to get back to all the challenges we are facing for these low resource languages. So when we train these foundation models, we train them on internet data. And internet data represented by more than 60 % is English and followed by some high resource languages like French, Mondarian, Portuguese, etc. So this low resource language is, even if they represent more than 7 ,000 languages, they represent only a tiny portion of internet data. This is the first challenge. And the second one is the benchmarks. When we build LLM, we benchmark them, we evaluate the performance on benchmarks.

And we have seen, like, there are only at least one benchmarks for only 300 languages. So we can see, like, more than 6 ,000 or 7 ,000. They don’t have that many benchmarks. So we can see, like, more than 6 ,000 or 7 ,000. They don’t have even one benchmark. Even in these 300 benchmarks, most of them are just translation from English to this low -resource language. They have nothing to do with the culture and the context of these languages. The third challenge is the performance gap. Of course, there is a performance gap for these LLMs, even the frontier models between high -resource and low -resource languages. The fourth one is safety. When we build LLMs, usually we do some safety alignments with reinforcement learning, but these safety are mainly done in English and some of high -resource languages.

Now, when we build LLMs for low -resource languages, it becomes very strong for these low -resource languages. It raises some other issues with safety. We have to evaluate these LLMs for safety on this language and do all these alignments, reinforcement learning in the target language also. In this PO, we addressed some of these issues. We have to evaluate these issues and do all these alignments. We have to evaluate these issues and do all these alignments. We have to evaluate these issues and do all these alignments. We have to evaluate these issues and do all these alignments. We have to evaluate these issues and do all these alignments. We have to evaluate these issues and do all these alignments.

We have been targeting three pilot languages, which are Inuktitut, spoken in north of Canada, indigenous language, Chichewa in Malawi in Africa, and the Maori in New Zealand. Why we have selected these three languages? Just because we have access to local community to help us to get data. So we gathered data from this community. Then we used some continual pre -training, instruction fine -tuning to boost the performance of open weights LLM, and we were able to gain 12 % balance gain, closing the gap with English. So now what’s next? Next steps, actually we are trying to expand this to more languages. We have some collaboration, for example, in North America with Paraguay to develop LLM for Guarani, and we want to extend this to other languages.

But most importantly, we have launched an initiative to help the community to get the best of the language. We have an initiative called Lengua Europe. We have a project called Lengua Europe. to fund data collection in Europe for 10 languages in Europe. It was released in last September. It was very successful. We have received many applications and 10 have been selected. And now we will start working with them. And it was that successful that now we are extending this initiative to Africa through Lingua Africa. It has been announced just today in the AI Summit. And we will be allocating 5 .5 million to support data collection for African languages. And this is in partnership with Gates Foundation and Microsoft Air for Good and FCDU.

And this initiative will be led by Masakani African Languages Hub.

Sorry, just to follow up. So for people who are working on these small language models or domain, specific language models even, you know, say for healthcare domain or some other domain. Are there, you know, strategies? that they should pursue that you can recommend?

Yeah, this is very important, and this is also related to the call for Lingua Africa because many efforts have been done in the past to collect general proposed data. Now we have enough, I think, general proposed data, but we evaluate the performance of these AI tools for application -specific, for example, healthcare, education, agriculture. They don’t work as we want, as expected. So what we want today is rather than, instead of focusing on general data collection, we will be focusing on domain -specific, application -specific, use case -specific data collections and building AI tools for specific domains. At least for this all reliability issues, we will have a model that performs good in that target level resource language in that application that we can deploy and we can be used by local communities and local communities.

This is really a priority for the next…

Thank you. Ilango, let me come to you. You have vast experience in international development. Can you give us a view of the future as it relates to using AI for developmental goals? Do you think AI will have meaningful role to play in transition of emerging economies to advanced economies?

So I do think the prospects are good and our North Star is job creation. And so we need to support countries that AI doesn’t automate jobs away, but AI actually supports the creation and enhancement of jobs. And this is where small AI becomes imperative, unlike the foundation models. Good question. Which will have implications. So the second is how are we going to go about it. And in some sense, whether it’s large language models or small AI solutions, you need an ecosystem. And that ecosystem needs to be powered by the local private sector. And often what we see even now, whether the AI revolution is before us or not, small enterprises, whether in the SME space or in the larger space, struggle for a variety of reasons.

And if the countries don’t reform business processes, make it easier for permitting, which AI can do, you’re going to see that AI actually is not going to play an effective role. So there are some fundamental reforms. And this is where some of the foundational investment in BPI, the digital public infrastructure, needs to happen to create that ecosystem and the ability for the ecosystem to then work with the private sector, the local communities, to be able to create those jobs. And this is what we… seeing everywhere, if that happens and here too you see this whole vibrancy around the startup ecosystem is why? Because everyone, the young people see opportunities and this momentum can drive everywhere in the world.

Whether it be in India, whether it be rest of South Asia or Africa or Latin America or even in the Pacific region. So how do you go about it? And what we did was we joined hands with a number of multilateral development banks and last couple of days we launched this small AI use case repository. It’s a good 100 cases. It explains in health education and agriculture and job creation how AI can be leveraged to the maximum advantage of communities. Both in terms of service delivery, productivity gains, household income gains. All this eventually leads to better jobs, better employment and better income prospects. So we are very much upbeat about small AI but I do take the point about community trust.

Once it fails, the community is not going to believe it. So it’s very important that whatever we put in place work with others, partners including the MDBs or Microsoft, Google, Gates and everyone. We have to ensure that whatever we leave behind in small communities is something trustworthy, reliable and it doesn’t end of the day hallucinate and give them something that the farmer struggles and ends up with other challenges. Thank you.

So this report you mentioned, is it open access?

On the World Bank we are hosting it. It’s called AI Repository. Just type and you’ll be able to access it. It’s got 100 and we’ll continue to update this and once we’re able to sort out some legal issues then we’ll also allow anyone to submit their use case repository obviously into the repository obviously we’ll go through a filtering process to ensure that the right ones are there.

Great. Thank you. Antoine, coming to you. You have an organization that uses AI to advance health outcomes through research and commercialization. Are data -efficient and hardware -integrated AI models important for the work that’s happening at Parasante? And do you see these models as sort of potentially being deployed in low – and middle -income countries like India?

Yes, so clearly they are very important for us for different reasons. Of course, we’ll get back to the scalability and the use in low – and middle -income countries. But at first, what is the reality in healthcare is that data is scarce and siloed. And so you need to work on what you have, actually. So sometimes it’s a large set of data. Sometimes it’s a very small set of data. But you need to have tools that allow you to build relevant algorithms and relevant analysis. on small data sets. In the meantime, of course, we’re building larger data sets. Sometimes it’s at a level of one department in one hospital. Sometimes it’s one hospital. Sometimes it’s a group of hospitals.

At the end, what we are reaching out in Europe is the constitution of a large European health data space. 450 million citizens joining their health data in digital public infrastructure organized in 27 countries, which will be a world premiere. But in the meantime, we need to work on that reality of scarce data. Second thing is that not only data is limited, but also when you want to enter the new revolution in medicine, which is what we call precision medicine, personalized medicine, you need to work on very efficient algorithms because they need to adapt to one person and not only to a whole population. So you need also to get that into account in building the algorithm. The last thing is that You also have to work with what is existing in the healthcare systems, which is sometimes not supercomputers or high calculation power that exists in servers remotely.

But when you’re in a room of a patient or working in hospitals, it’s a very simple computer. And you need to have efficient algorithms and tools that you can have running on that kind of computers. And so, of course, you go all the way to a smartphone at some point if you go into remote areas. So this is why we actually work on this kind of approach, making sure that, of course, we have research on LLMs and large computing power. But we also have this work on small data, very efficient algorithm.

Can you give examples?

Well, yes. I mean, I already gave some examples about radiology. We are able to. We have a radiology algorithm running on small computer machine. And getting back to your example, which I think is really important, it provides me the. opportunity to put two very important facts. One is that the AI that we use is providing information. It’s not making decisions in healthcare. So of course we target high level of reliability but at the end it’s a human decision and this is very important I think. Second one is that we’ve been trying to compare the performance of the algorithm that we’ve been designing with the existing performance. And of course you’re reaching to 99 .999 % etc. But what very few people actually know is that the actual performance of what we do at the moment is not 99 .999 % So most of the time and I won’t say the numbers but most of the times it’s actually better than what we have.

And this is really important in your example. Is it good enough compared to what we can actually do at the moment? And I think it’s particularly important in low – and middle -income countries because a very simple solution, offline LLMs, et cetera, can solve many, many issues.

Alban, can I pick up quickly? I think it’s really important, and actually I’m going to name the number if it’s okay. Okay, so a really important World Bank study from a few years back showed that on a set of five very simple conditions, the diagnostic accuracy was 50 % across eight countries. 50%. What illnesses are we talking about? Acute diarrhea, upper respiratory tract infection, maternal hypertension. And the point about that is I don’t think, any of us would be happy with 50%, the equivalent of tossing a coin and saying that’s okay. And so I completely understand that today there’s a big gap between what the models can offer. And I think the question about are the models performing better than the average clinician, that’s done.

Sorry, I can’t resist the follow -up question. So often you find that average accuracy of models is far better. But models seem to fail more unpredictably than humans. At least that’s sort of the understanding in health care. Do you agree with that or do you think that’s not true? Anyone who wants to answer this.

Well, so I think we need another hour to discuss this. So what you say is absolutely true. But then you need to look at every pathology or every symptoms that you’re looking. Because the performance. The performance of diagnostic can be a little bit higher in certain places, in certain situations, a little bit lower, et cetera. But we get to the right to the same point, which is what we are building. is actually better than what we are able to do at the moment. And what we show in the scientific literature is that actually the combination between algorithm and natural intelligence, I would say the doctor, is actually the best tool so far. So the question, getting back to your question, how do we deploy this in low – and middle -income countries, I think it’s really important.

We need to have a model that are able to run on small devices. That are able to run offline. And sometimes it’s a very limited set of data, very limited set of algorithm. But if you, we were actually discussing in Paris about examples of remote LLM providing answer on the 10 most important questions for healthcare in low – and middle -income countries. That doesn’t need LLMs online with super calculation power. So that’s one first point, edge native AI. We also need to have data -efficient learning systems because most of the time in low – and middle -income countries, we have a limited amount of data available. So this is what I discussed earlier.

We have a lot of data in India, but it tends to be noisy.

Yes, but we need to get the time to actually get them together, clean them, and get them prepared for robust analysis. So I know you are leapfrogging and going very fast, but by the time you will scale, this will create a real power of analysis. And then we need also to understand how we can couple hardware with software and algorithm to design reduced costs so that they can very easily scale. Thank you.

Great. That was fantastic insight. I’d actually like to give some time. to the audience to ask questions to our panelists so yes please

thank you very much I’m Irish Kumar from the CSC Winnie Ocean Center on solar energy particularly in basement I’m belong to Rajesh son question a question to World Bank president very thanks to the World Bank in Rajesh on 60 % population rural areas and totally based on the agricultural domain 40 % population in the youth how our bank is increase the capacity of AI application to the youth as well agricultural domain so the economic changes more productivity more economy more you inclusion in climate change and renewable energy domain

Thank you for that question, which I think is a very foundational question to ask any policymaker in terms of what kind of an AI strategy or implementation you want to have at any geography in the world. So obviously the first thing is you need digital literacy. Second, you need to skill up so that everybody is upskilled and reskilled on AI -related capabilities. Third is improving the STEM capability in schools and universities. So you do create a future cadre of people who can work on these topics. And then the sectors you mentioned, which are our priorities, agriculture, health, and education, obviously this is where we see the greatest potential for small AI. But particularly on Rajasthan, right now I don’t have any information, but I’m happy to share that with you.

But certainly we are working across different states in India like we’re doing elsewhere in the world. And we do prioritize literacies, skilling. STEM and applications in priority sectors like agriculture, health and education.

But having said that, I also want to say one point. I mean, just to respond to devices that can do computing, devices are expensive for the bottom 40%.

Yeah. Hi, my name is Selena. I’m the CEO and co -founder of Zindi. We run competitions to develop models, especially in Africa. And I actually had a question for Wasim about kind of the technical implications, the size implications, the practicality of using… Open source, open weight models, you know, large language models to train very specific, domain specific, you know, language, you know, under -resourced language models. How have you seen that play out?

Yeah, I think what we have seen, like the selection of the base model is very important. Because what is true, what is real, that we cannot train from scratch in LLM, even if small or large language model. We cannot train it from scratch for lower social languages because we don’t have this 15 trillion tokens to train. So it is very important to select the best multilingual model that has the right tokenizer that can be adapted to many lower social languages. This is very important. And then get the data that we need. And what we have seen also, monolingual data helps, but also bilingual data can help. And also translating English into this lower social language can also help to boost the performance.

So in our paper, we are providing all these three CPs to follow to get the best boost in terms of performance. What I would like to add also, with all these level source languages we have, text cannot solve them all. Many of these languages will be solved by speech. It’s very important. ISR models, speech -to -text, text -to -speech will build a very large role into unlocking all these level source languages in addition to LLM that can operate into level source language or in English.

I think we have time for one really short question.

Hi, this is Dr. Ravi Singh. I’m from Miami, and it was a great panel, so a lot of great insights. It’s for Google, Microsoft, and the World Bank. Here’s the scenario. If there’s compliance across all of these platforms, which platform will win the AI wars?

That’s a loaded question. Anyone want to answer? I’m not. So first of all, I think healthy competition is how we’ve been able to develop incredible technologies just over time. So the competition is healthy, and this is great. I don’t see it as a zero -sum game. There’s too many people on the planet, and there’s too many challenging, unique problems that need to be solved. So if we’re making it useful and bringing joy and happiness for all, that’s in the – I just love it, that’s in the theme here, then it’s not necessarily going to be who wins whatever platform. It’s what is relevant to the context of the end user. So taking it back to a more, like, human, personal perspective.

That’s my thinking.

First, three billion people are offline, so there is space for everybody to compete. Second, in health sector alone, three and a half billion people don’t have access to healthcare, so there is enough scope for all kinds of applications.

Just I want to add, many people have been asking me to, if this all efforts we are doing for language, if it is enough to make this model as good as English. I would say maybe not, but without all these efforts we would never reach this objective. So we have all these collective efforts will get us to this objective.

Thank you so much, everyone. I would now like to invite Neha Butts, Associate Director, Human Resources, to just hand out the mementos to all our speakers. And we will just take one group photo. Thank you. One group photo, please. Requesting the speakers to just take one group photo, please. Thank you so much everyone Thank you everyone for joining Thank you Thank you