# Transcript

[00:00] Thank you Radhika, and good afternoon everyone. I see very serious faces. You've been having very long sessions I believe.

[00:22] Before I start, I think if I'm not wrong, just looking at the schedule, you've had around three talks on AI. So let's start with that. If there was one big takeaway that you had from everything you've heard about AI, what was it? Anyone?

[00:45] **Question**: *One has to live with it.*

[00:48] **Answer**: Have to live with it, yeah.

[00:51] **Question**: *And try to study so that it will not overpower us. It should not overpower.*

[00:56] **Answer**: It should not overpower you. Okay, that's a good takeaway. Any other takeaways from the discussions?

[01:06] **Question**: *Job loss.*

[01:08] **Answer**: Job loss. *Jobs jayega* [Jobs will go]. *Hum sabke jobs jayenge* [All our jobs will go].

[01:15] **Question**: *Transforming sectors.*

[01:21] **Answer**: Transforming sectors, transforming the way we work, right? Yes. Sorry sir, yes.

[01:28] **Question**: *Compared to perhaps the US and China, we are too far behind. We haven't done enough. Just scratching the surface.*

[01:37] **Answer**: Yes, I agree. I agree and we'll talk about that.

[01:44] Now, these are all sort of good learnings to take from the sessions. Right now you're thinking that this AI animal, we have to figure out how to live with it, which is a very fair statement. I think what I'm going to talk about is going to scare you even more because what I'm going to talk about is what's going to be coming in the next ten years.

[02:18] And in your roles today as senior leaders in government, **if you don't understand this, then the game is already lost. Because it is the policies that you will frame today that will determine what happens to us ten years from now.** Ten years... what you said right now, US and China, we can't catch up. Imagine what happens when you end up in a situation with the future that I'm going to describe, where you haven't done enough to be ready for it.

[02:56] So no pressure, but the future literally sits on these shoulders, it depends on you all. And the kind of pivots that you all make right now. Because we have to make pivots. Going along the same path that we have been doing, we may like to tell ourselves all that is great, but that's not going to work. We have to make some changes. So that's sort of what I will be sharing today.

[03:29] The organization that I'm setting up, or I have set up in NITI Aayog, is called the NITI Frontier Tech Hub. And I'll talk a little bit to help you understand what we do. But one of the things we do is every six months we come out with a tech scan of what's coming in the next five to ten-year frame. And this goes to PMO, this goes to all the key guys and ladies, so that they can understand what's happening. And I'm going to share with you a shorter version of the latest tech scan in one hour, so that you can at least get an understanding of what's happening.

[04:18] So if you're ready, we will start. I'll start with what is the Frontier Tech Hub, so you understand the context behind this presentation. We were set up within NITI Aayog, and I came in from industry, I was asked to come and build this and get this set up, basically to understand what are the next big tech shifts that's going to happen.

[04:52] Now history shows that when the patterns start forming about a tech shift, it usually takes a decade for the impact to be felt. AI started in 1947, but the world woke up to AI when ChatGPT came out. Most of the world... a few of us who were studying it woke up to it much earlier, realized how crazy this technology was, but the rest of the world woke up to it years later.

[05:32] So any technology, and you can think about any technology, it takes anything from 10 to 20 years for technologies to mainstream. We believe that the number of years will now come down. It will be shorter. But we still believe that a decade is a fair timeframe to say that all the new technologies that's playing out today, they will be becoming mainstream by 2035.

[06:09] And that's why I again go back that it takes a decade, and if you think about AI—since that's been the topic of conversation—the two countries that you named, US and China, when did they start their work on AI? They started in the 1990s. The serious work. It was called the AI winter, and that's when most of the R&D happened. And then that's where the Transformer paper came out, etc. So these countries were able to spot the patterns. They were able to spot the trends, and they were able to start investing in the R&D and the capabilities to be ready for it. And that is why they are heading the game today, and everyone else is catching up.

[07:00] Our job, and I think this is the Prime Minister's vision, to say that we need to know what's going to happen in the next ten years. And we need to know today, and then we need to figure out how to be ready for it. So that's job one for us.

[07:24] As we think about what's happening today, what are those big shifts in the next ten years, we apply four lenses. We look at what does this mean for citizens. What's happening on jobs, healthcare, education? How will these get transformed? Second, we look at industry. How will manufacturing get transformed? How will agriculture get transformed? What are the new industries that will get formed? Because there are multiple new industries that are coming up now. Third, we look at the exact view: government. National security is a key part of this work, where we look at what is happening to national security. That roadmap we've come out with, it's not public, but it has been shared with everyone including all the defense forces, and it's scary as to how rapidly our understanding of national security is going to change. And then last but not the least, we study the impact on global dynamics, geopolitics. Because technology and geopolitics is today intertwined. If you see all the geopolitical moves and what's happening, there is a tech play. Either they are fighting for energy, or they're fighting for GPU access, everything will trace back to technology.

[09:00] That's the work we do. I've put down the URL. Do visit when you have some time. And under that you will find the roadmaps, discussion papers, all the work that we put out there. Spend some time on it. It may seem futuristic because all our roadmaps are ten-year action plans. They tell you what to do in the ten years. So we work with the relevant ministries. But I request everyone to take a look at it because it tells you how fast things are changing.

[09:36] That's what we do in NITI, and what I'm talking about is a result of that work. So how do you spot shifts? This is the question I always get asked. We don't have a crystal ball, and no one can say this is how things are going to change or this is the next technology. But **technology works through patterns.** If you look at the evolution of technology, it leaves behind... like we human beings leave behind footprints, technology leaves behind thousands and thousands of patterns. And what we do, and this is something I've been doing in my entire career in the industry, is understanding these patterns. We model them out. You try to make sense of them. You figure out what these patterns are telling you. And these patterns actually tell very compelling stories. And when you start adding these patterns, you realize what's going to come next.

[10:48] I'll give you an example. Everyone was surprised when ChatGPT happened in November 2021, right? The whole world was surprised. "Oh, we never knew!" That's not true. If you had been following the patterns, you would not have been surprised. In fact, the only surprise was it didn't happen earlier.

[11:03] And the patterns have been telling you that. I won't go into machine learning and the old history, but one of the strongest patterns—I don't know if any of you remember or have seen this—you remember Alpha Blue [Deep Blue] which beat Kasparov in chess? First time a machine beat that. I will start there. That was a pattern that there's something happening here. This machine... it's not about intelligence, but this machine knows maths. Because chess is maths. The second pattern that got added onto that was when DeepMind's AlphaGo beat Lee Sedol, the world champion, in the game of Go. Which was a wake-up call for all of us in the AI world because chess you could still understand—it is a lot of maths into chess. Go is a hardcore strategy game. You can't play it just with numbers. And that was my "aha" moment, not ChatGPT, but my "aha" moment for AI, where I really woke up to it and said this is serious, was when that happened, when AlphaGo beat Lee Sedol. And then the third big pattern was Google's paper on Transformer, which actually showed how generative AI can work.

[12:47] For those of us who read it, it wasn't a surprise that... and ChatGPT didn't happen at one shot. OpenAI came out with various versions of GPT. Most of them didn't work. And then ultimately ChatGPT happened. So this is what I mean by patterns. Patterns happen over time and it's important to figure out how to recognize.

[13:10] We have been analyzing over maybe 10,000 patterns that have been taking place over the last few years, to understand the future, modeling them out. And this is what the patterns are telling us. First, the vision of the future: intelligence will flow through everything. **Intelligence is ubiquitous. You will have intelligence getting integrated into every single workflow.** I think that's going to be the job of AI. It's about diffusion of intelligence into everything. That's one big part of the tech shift that's taking place today.

[14:01] The second pattern is technologies now are not playing alone. Technologies are connecting. They're converging. So the biggest disruption will come when AI converges with biology to create what we are calling "programmable biology." Where just like you can write software codes, you can now code your DNA. And think about it. And nothing I am saying is science fiction. Nothing I am saying is something that is on paper not yet proven. Everything I am saying has been proven. Programmable biology basically means we now have the ability to hack life. Because we now have the ability to re-engineer the basic molecules that are the building blocks of life. Think about the problems that's going to come when this becomes mainstream.

[15:15] The third thing: technology is definitely moving towards physical. So earlier we used to say software is going to eat the world. The near future is going to be about physical AI. How software and hardware intersect to create embodied AI, robotics, physical agents.

[15:30] Fourth: all of this is leading to a massive integration of your critical tech supply chains. We are seeing today the blurring of lines between hardware, software, material science, biology. And whoever controls this interconnected supply chain, that is what is going to determine geopolitical leadership. It is your control on the physical components of the tech supply chain that will determine whether you win or lose the race. One of the reasons China is doing so well is because of all the critical technologies, they have figured out how to build the complete supply chain dominance. Pretty much top to bottom. It's no more about saying, "I have great engineers who can design XYZ." That's not going to be enough anymore. It's going to be about your end-to-end stack play. How do you build systems?

[16:39] Technologies are going to become interdependent. You need AI for biotech. You need AI for semicon. You need energy for semicon. You need semicon for AI. You can't think of technologies in silos the way we used to. You have this mission here, you have this mission here, and sometimes the missions don't talk to each other, it's a big problem. Technologies are completely getting interconnected.

[17:13] The fifth pattern: competitive and geopolitical advantage will flow to those who command or secure trusted access to multiple layers of interdependent technology stacks. If you really want to have geopolitical leverage, you have to have a play across energy, compute, models, application. If you want to lead in AI, you need to lead in all of those.

[17:48] Sixth: it's much more than research now. The ability to combine infrastructure, industrial capacity, innovation. The game is changing. It's no more about how many research papers you come out with, but what's the outcome. What is the difference to your economy? What's the difference to social development? The rules are changing.

[18:20] Access to clean energy will be the greatest roadblock, not just for AI, but pretty much for all technology.

[18:28] And I think the biggest challenge we will have, and especially government will have in the next 10 years, is figuring out the role of human beings. What will human beings do? Sam Altman came out with a paper to say that no one should work. We should create a public fund where the profits flow and everyone gets a share of it. Is that feasible? Maybe not. I have lots of questions on that. How do you assume fair share? But the future is going to look very different from what it looks today.

[19:22] Why it matters: as I said, if you act on it today, it will decide whether we continue to stay consumers of technology, or we can actually start leading in technology. Your actions today are going to be very, very important.

[19:43] It's very clear the writing's on the wall. China has gone down that path. The US has gone down that path. Now EU is pivoting down that path. Success is not going to be about point solutions, or point systems. Or saying I am good in quantum, I am good in AI. Success is going to be about your ability to bring it all together to connected systems. Building connected systems, building end-to-end stack leadership.

[20:16] When you think about it from the perspective of a country like India, what will it take to be successful? One, there's just too much out there in terms of technology. With finite resources, finite budget, you have to pick. You have to pick what is relevant for you. When you're thinking sovereignty, you have to think trusted alliances. Sovereignty in most of the established technologies is near impossible. We still have an ability to build sovereignty in some of the newer areas. But in established technology, the choke points have been so hardwired, it's going to be impossible to break.

[21:13] India needs whole-of-government coordination. Our biggest enemy of technology diffusion in India are silos. Silos that exist between states, between departments, and even within a ministry. If you have to do systems space, stack space, you have to find a way to break those silos and start having an integrated, top-down strategy.

[21:29] We have to start linking innovation to productivity. All the R&D money that's going into India, what is the outcome? What does it change? It has to start changing things. It has to start solving problems.

[21:50] And last but not the least, countries have to invest in talent, mindset, and foresight. If you don't spot these trends early, you're going to be caught by surprise every single time. You're going to wake up after ChatGPT happens and say, "How do I catch up?" And it's going to get more and more difficult to catch up.

[22:21] Now this is a very simplified scan of the frontier tech radar. I have only put there eight of the most important technologies that we are studying deeply, but the real radar will have over 50 technologies. To prioritize, we look at three things. First, which of these is really going to give you accelerated economic growth? Second, we look at which of these are important for national resilience. Security of defense, food, health, water, communications. And third, what will give you geopolitical leverage? Technology control today is tied to trade. Which of these is going to give you higher negotiating powers when you sit at the table?

[23:38] These are the four broad areas we recommend. First, the intelligence revolution. Second is the biology revolution or biotech revolution. Third is the infrastructure revolution. Compute is today the leading act there, but you have space, you have a few other things coming up. And then the energy revolution. These are the four areas which are a must-do to achieve those three things. And of course, in order to do it well, you have to ensure that you have a proactive risk mitigation system in place. Technology has a good and bad side. As long as humans are involved in the tech play, we will misuse the technology. Let's be very clear. I refuse to say technology is bad. Because technology is neither good nor bad. It's us humans who decide whether we use it to cure cancer or we use it to kill thousands of people. That accountability stays with us.

[25:13] As long as humans are in the loop, you will have great usage of technology, you will also have tremendously bad usage of technology. I don't believe in security anymore, because you know bad usage is going to happen. What I believe in is resilience. You know it's going to happen. How do you recover from it? Are you building resilient digital economies?

[25:37] So very quickly, I'm going to run through these four things. Why is the intelligence revolution important? This is the key to unlock exponential productivity. It's moving from a standalone technology. AI is now a critical control layer. It pretty much controls other sectors. Research, scientific discovery, healthcare, government, enterprise workflows.

[26:35] Over the next decade, if we can integrate AI into these workflows across sectors, it is believed that the world is going to get around 17 to 26 trillion added to the global economy by 2035. Any guesses who benefits the most from it? Which countries?

[27:01] **Question**: *US and China.*

[27:03] **Answer**: 80% of this goes to US and China. And rightly so, because they invested in it way ahead of the others. EU gets a large share. If you leave China out of Asia, then the Asia share is tiny. Same for Africa.

[27:28] But it shouldn't be that case for us. We have to change that. I firmly believe we have time to challenge this. **A good goal for us to set is not building artifacts, but saying that our AI strategy has to deliver 10% of this global value to India. What do we have to do to make that happen?**

[27:56] That's a great goal for India today. We worked with Global McKinsey to understand their arenas of growth, and we rehashed this for India. We basically found 18 areas where if you integrate intelligence into the workflows, the incremental unlock is going to be significantly higher. These 18 areas currently contribute to 1.7 or 10% of global GDP. The value is around 1.7 trillion to 1.9 trillion.

[29:23] But I'm going to be honest here. India is doing fantastic work to build the infrastructure, but when you look at adoption, India is the number one market for ChatGPTs. They are not the magic. The magic is how is AI transforming banking? How is AI transforming healthcare? How is AI transforming logistics and supply chain? The magic of AI is in workflow integration. And if you have seen IDC's latest report, India is lagging behind even countries like Mexico and others on enterprise adoption of AI. If there is one problem we have to fix to solve, it is that.

[30:54] Another area where technologies like artificial intelligence can have tremendous impact is reimagining or transforming sectors. Let's take the example of government. In government, while we use AI a lot to understand what has already happened, the real value is in predicting what *will* happen. A classic example is how various governments today are using AI to build digital twins to simulate policy. Before you come out with a policy, stress-testing that policy to see what's going to be the impact. This is widely in use today.

[31:47] China has built digital twins of most of their cities. So the entire city management happens through simulation. Singapore has built a digital twin of its entire power plant. So they can simulate, predict what will go wrong, and ensure that their maintenance is predictive rather than reactive, bringing downtime down to pretty much nil.

[32:22] We have to start thinking about technologies beyond conversational AI. I think that's the first challenge for enterprise and government in India.

[32:43] The second trend we're picking up is we are moving towards physical AI. It is the rise of digital labor. Today we are in the world of agentic AI. Which is different from AI agents. AI agents will do one thing that you train it to do. Agentic AI can do multiple things. It can drive the entire workflow. A simple example: before I leave the house, I will tell my agentic AI, "I'm going to make Palak Paneer. Make sure everything is organized." What will it do? It'll figure out the recipe. It'll figure out the ingredients. It knows that I come home by 7:30. So it will choose which vendor it's going to book from based on delivery times. It'll add everything to my cart. What I haven't yet given it approval for is to pay. I'm very scared of that. So it will just ping me saying everything's in your cart, can I pay for it?

[35:28] That's the power of agentic AI. Now imagine it being used in banks. Imagine it being used in each of your departments. What is happening is, we need to now put that intelligence into a physical form. And that's where physical AI is becoming so important.

[35:50] I'm going to show you a ten-second video. Just look at the left and right. Same company building robots. The left is where they were in 2023, and see how clumsy the movement was. The right is where they got in just two years, 2025.

[36:20] [Video playing sounds].

[36:30] You see the difference? Two years. Now you know the meaning of exponential, right? And not a surprise that they are already landed up in homes. This is not AI, this is the real video.

[36:47] [Video playing sounds].

[37:20] China has the largest deployments of humanoids today, mostly in hospitals, senior care, and in homes. And I'm not sure if this is true—it's a rumor—but the rumor is a lot of the training data was created watching Indian household maids at work. We keep talking about "our data," our data is getting used in a lot of interesting ways.

[37:50] What is the ultimate manifestation of this? Recently a group of Indian leaders went to China. One of the biggest "ahas" was China's dark factories. With lights dimmed and no workers in sight, this car factory in China uses hundreds of robots to turn out dozens of electric vehicles an hour, 24/7. This is a dark factory, an area of a plant so automated that in theory the lights could be completely shut off.

[38:33] These dark factories are spreading. Productivity is much higher because you're churning out many more cars. Assembly lines are moving much faster. I was asking some of my friends who visited... How many human beings manage the car factory? Or manage the phone factory? Any guess? 10. 10. They said the car factory had at best around 50 people. And the phone factory had around 100 people. That's the future.

[39:15] Productivity has gone through the roof. Now yes, we can say we don't want that future. But India can't be the only one to say we don't want that future. You have to get the whole world to come and say we don't want that future. If we decide to do that, this is the time to start those discussions and negotiations.

[40:07] In 2026, the total industrial and government spending in AI was 1.5 trillion. That's how much the world is spending. Progress is not slowing down, and this is becoming a reality.

[40:57] Just a few numbers for you. This is the entire physical AI opportunity. 1.5 billion already in 2026, and will become 15.32 billion by 2032. Last year in 2024, there were literally around half a million robots that were installed. Asia accounts for 74% of that installation.

[41:42] What is interesting is humanoids. Look at the humanoid sales. The top four companies are Chinese. Agibot, 31.9% share. Between Agibot and Unitree, the top two Chinese companies, that's around 50% share. Let's analyze the supply chain of a robot. From head to toe, what are the different parts that's required? Can you count the red flags on this? Can you see the dominance of the red flags in this?

[42:52] This is what I meant, that it's not enough to just control one part. You need end-to-end stack leadership. No one can come in and disrupt this. It's going to take tremendous amount of investment and effort to displace someone who literally has leadership in every part of that supply chain.

[44:02] Job displacement is going to happen. See it has always happened with every tech shift. What is new is this is happening very fast, and there's going to be a huge lag between jobs going away and us figuring out how to reskill people for new jobs. The new jobs that are getting created unfortunately are not like the old jobs. Data entry operators employed millions. The new jobs don't do that. The new jobs are all hyper-specialized. And they employ at best 100,000 people. One of these new jobs which is right now very much in demand is haptic engineering. I looked at every institute in India to figure out who's teaching haptic engineering. Finally I was able to find one. But they're not teaching it, but they do have a lab. Are we even skilling people for the right jobs?

[45:18] I'm going to move on to the next trend. What happens when you can program molecules and DNA? You basically can create anything new. Or you can re-engineer anything. From your DNA to molecules that make up medicines, or make up proteins or molecules that go into materials. **Biotechnology is the next mega general purpose technology.**

[45:56] I do hope India will really focus on this as a priority, because **this is going to be way bigger than AI**, and this is one place where you don't have hardwired supply chain choke points. And this is one place where India actually has a strong ecosystem already, thanks to the vaccine work that we had done. Genome sequencing cost in 2003 was $3 billion. That has come down to $600 today. The first customized CRISPR therapy was done for Baby Tayt who was born with a liver genetic disorder. And I check every week, and touch wood, seems to be doing very well. FDA approvals are really growing, which means it's moving out of labs and into markets.

[47:28] This is becoming programmable and manufacturable at scale. Everyone used to think of biotech as health. Biotech is way bigger than health. It is going to flow into materials. One of the companies that make aircrafts has announced a project where they are using the technology to create self-healing material for aeroplanes. There's already pilots underway to build roads with self-healing cement.

[48:13] If you can re-engineer molecules, you can re-engineer anything. This is going to be way bigger than AI. The race is on between China and US in terms of just how much they're spending and the size of the bioeconomy. The research that is coming out of China is mind-blowing on this.

[49:12] India needs to significantly raise its ambition. Our goal right now, we believe we will be 550 billion by 2035, and 1.2 trillion by 2037. This is going to be around 4-5% of our GDP. Like we are saying a 10% target in AI, we have to have a 10% target here too.

[49:46] What will determine winners? Thankfully, there is no hard-wired choke points. But yes, you have to play across the system. You can't just say I am good in biosimilars. That's not going to be enough. You have to ensure you invest in the hardware, the biofoundries. You have to ensure you have the data in place.

[50:18] One of the things we have to be very careful about is the risks of biotechnology. It is going to be way more destructive than AI. Wars are already being fought on these lines. The future of warfare is biowarfare. It's about DNA mutation. If you start reading the scientific papers you will realize this is already happening. This is not science fiction anymore. How do you govern this? We need to think from now.

[51:27] The third bet, as I said, has to be compute. The good thing about compute is that the landscape is completely changing. Compute is moving from centralized data centers to the edge. In order for it to move to the edge, you need energy efficiency. You need new materials. India is getting into the 2D materials game too. And then the newer areas, quantum, neuromorphic chips, which is what goes inside your brain. They are already in markets.

[52:49] While traditional compute pathways are very hard-wired, the new opportunities do create opportunities for countries to leapfrog. Are you investing in those areas?

[53:09] Compute is moving to the edge. Most of us wear a smartwatch. It's collecting our data, it's feeding the data into our phones. In the near future, you will have intelligence loaded on this watch. It will process the data on this watch. It won't have to send it to my phone. And it'll actually take actions based on the data. Once we crack the energy code, you basically chuck your phone, and it creates your digital twin, and it tells you what is going on inside your body. This is an engineering problem. You just need to throw enough mindset at it, and enough dollars at it.

[54:00] Compute creates the need for intelligent sensors. I do believe this can be India's leapfrog moment. Because we have the know-how to design the architecture.

[54:37] Traditional compute is hard-wired. Look at the supply chain. EDA software, US dominance. Lithography, one company in the Netherlands, ASML. Wafers fab, US, Japan. Semiconductor materials, Japan, South Korea. Critical minerals, China, complete dominance. India does have critical minerals, but we don't have the ability to convert it. Foundries is Taiwan, and packaging is Asia, US and Europe.

[55:30] This is a very difficult supply chain to crack. Sovereignty has to come with trusted alliances.

[55:54] The next one is the energy revolution. You've all heard about the great news on nuclear from India. Nuclear has to be the way forward. When it comes to clean, it's the cleanest. Most reliable energy source in terms of producing power, 93% of the time. Cheapest forms of energy, because though initial costs are very high, it has a long life. The India news was so positive, because I think if we can scale it, we can rewrite the rules of the game. If we are dependent on uranium, we again have a very hard-wired supply chain. It's five countries controlling 71% of generation capacity.

[57:33] The last but not the least is, how do we ensure we are building resilience into the system? So when we are down, we can get up very quickly.

[58:00] I stopped using the word cybersecurity. If you look at the risk channel, there are four broad areas: cyber, social, emerging threats, and operational. In cyber, the risk is moving from devices to your national infrastructure, power grids, supply chains. Social impact, we used to worry about AI ethics, bias. Bigger worries: what I call hyper-personalized disinformation at scale, cognitive warfare. Emerging threats: what's going to happen with quantum as it breaks encryption? Bioethics, biosecurity.

[59:28] The biggest risk for India is the operational risks. The capability gaps, operational silos, and the lack of standards. How many of you use your own processes and standards to procure technology? Everyone. When we do that, we are killing interoperability. And when you kill interoperability, you increase vulnerabilities 100x.

[1:00:25] Key takeaways for India: In technology, it's all about timing. **You get the timing right, you win. You get the timing wrong, you don't win.** We have to build capacity and alliances across the full stack. We have to stop thinking point solutions. We *have* to think stack.

[1:01:14] Your action is absolutely a must-do. Shared infrastructure, institutional reforms, talent development will not happen without government. If we don't do this, then the aspiration of Viksit Bharat is going to be challenging. Prime Minister describes Viksit Bharat not about incremental growth, but transformation. And that transformation will only come from technology.

[1:01:53] This is not optional. This is a must-do for us. We have four clear outcomes in front of us: higher productivity, globally competitive industries, strategic autonomy and national resilience, and world-class talent and innovation.

[1:02:40] That's the work we are trying to enable in the Frontier Tech Hub. The ones in red are going to be released in a month or two. The ones in black are already available on the website.

[1:03:13] I think we have exactly ten minutes. I'm happy to open it up to questions.

[1:03:20] **Question**: *Ma'am, a very good afternoon to you. And very happy for the brilliant exposure. Ma'am, we all know that the technology stack was earlier a tool, but now it has become a collaborator. You mentioned something very interesting about patterns. Pardon my also studying the pattern on governance because technologies we know keep coming back in different forms. 2008 we had this huge GMO debate on brinjal. We were not clear about the policy, nothing happened. CRISPR is here already as you mentioned, and we are still not very clear whether it will be applicable at least on some congenital diseases or not. Say for example, in the tribal areas you have hemophilia as a major case, will we be able to utilize CRISPR for that population, we are not clear. 3D printing came, IoT commercialization has already happened in the private sector. We are not using it in libraries, we are not using it in multiple forms. 3D printing could have resolved our Pradhan Mantri Awas Yojana, our ODF issues. China used it brilliantly to build hostels to curb their urbanization, migration issue. Of course, crypto, we have a very clear stance that we don't believe in crypto. Digitalization of our artifacts, I was in the museum sector before coming to tourism, we did not have a clear policy. I steered the NFT for India, but the finance guy really told me that you can't do it, we went ahead and started capacity building because apart from the physical and digital stack, the cognitive stack also needs to be built which is a lot of commercial drafts, lawyers where lawyers are required to do that. So I am not very sure because, ma'am, the devil lies in the details, so I am not very sure, given the patterns of the earlier technologies, blockchains, drones, how do you think we will be able to... and that is why I am not very fearful because I see that we were not able to jump onto the bandwagon of the earlier technologies and so this may also just come and go. So what is your comment on how it can be embedded? And if we are not able to... it cannot just be the technology as I read which you very clearly mentioned, unless we learn to build the cognitive stack with the physical and the digital, I don't think we will be here. And that requires not just us, but it requires working with commercial people, it requires working with lawyers, it requires working with scientists, technicians, and I think transnational policy building is what is the future, ma'am. Your comments on that. Thank you.*

[1:07:05] **Answer**: I think you've said it all! You've sort of said it all. The problem and the solution. See, when you look at... I don't spend a lot of time on historical patterns because unfortunately we can't learn much from history in today's world. We can learn what mistakes not to make. And I think when we look at historical patterns, the only thing we should take away is what should we not be doing. I don't think we can use historical patterns as an excuse to do or not do something. Because this is where the change mindset has to come in. There's a lot of things we can say we didn't do in the past, but we *did* change. India was a country that built a national ID for a billion people. India was a country that then took the national ID and we built the India stack way back. So it's not like India hasn't done this, right? It's our willingness to replicate what Nandan did in DPI across the other sectors. Are we willing to apply that thinking here? So yes, to answer your... agree completely to your solution. If we don't do this, if we don't build cross-functional teams, breaking down the silos across ministries, we won't win this.

[1:08:24] **Question**: *Ma'am, you said that AI will change multiple sectors. Now what is the right approach for any government? Is it wait and watch, let AI reach its plateau, then see its application in respective sectors? Or should an early adoption be made because when this is evolving stage, its use cases also will evolve, keep modifying them. For a country with limited resources, financial especially, what would be the ideal approach for a developing country like India to integrate AI into government systems?*

[1:08:58] **Answer**: Great question. Can I combine the two questions and answer it because we have just 2 minutes...

[1:09:07] **Question**: *Hi. From the Ministry of Science and Technology. So what is our plan as a country? Have we targeted something or we are having some very concrete thing in mind? And related thing, in the article you have mentioned in the presentation also you have shared about that integrated approach. At least one sector where we can see a lot of happening around the world is the drone area. So where do we stand as a country?*

[1:09:30] **Answer**: So, I don't know where to start. There's so many great questions. Let's tackle the wait and watch part. If you wait and watch today, you can pretty much guarantee we will stay consumers. Because this is going too fast, this decade is moving exponentially. By the time we wake up and say, "Okay, let's look at this," the game is long lost. So wait and watch should not be an approach for a country like India. But what do you invest in? I understand finite resources, finite budgets. And that is why prioritizing is critical. India may not be the number one to come up with the best foundational models, or to say let's figure out an LLM that beats GPT. And frankly it doesn't matter. From a country's benefit perspective, the real game of AI is in workflow integration.

[1:11:15] Our investments need to go into driving integration, into driving usage. And not trying to be everything to everybody. Because if you do that with finite resources, it gets diluted too thin to matter. My recommendation would be pick five things, but go all in. Pick those five usage cases that's going to transform value. It doesn't matter which model it uses. Why do we want a sovereign model? Sovereignty should be at the data layer.

[1:12:00] Someone asked me that "do we have an R&D target?" The broader target that is generally discussed is we have to get to 2% to 3% of our GDP on R&D. But if you double click on that 2% or 3%, where is it going, what is the output? Today, US and China get a massive output from every dollar they invest in R&D. We are lagging behind significantly. The ROI on our R&D dollar is very poor. Because we don't connect it back to outcome.

[1:12:44] We have to fix rather than chasing another new target. We have to sort out the plumbing to use your words. Without that, money will just be sprayed and spread.

[1:12:59] What's the goal for AI? My goal for India would have simply been, if there is between 17 to 26 trillion getting unlocked by 2035, what does India have to do to make sure we get 10% of that value? We are targeting to be a $10 trillion economy. $2 trillion, which is an additional $2 trillion unlocked... think of what that would mean. It's 20% of your GDP. That comes from technology. Just AI alone. Let's align backwards from there. What do we have to do?

[1:13:41] Thinking about what to focus on and where to pick your bets, is an important thing given the limitations of our finite resources.

[1:14:02] **Question**: *In the article which you shared, you have mentioned that China in their five-year plan are planning to invest certain amount in R&D, some percent per annum increase. So what is our plan as a country? Have we targeted something?*

[1:14:15] **Answer**: I don't think there is a... maybe some of the others can answer this better than me. Broadly what gets discussed is India should reach that 2 to 3% of GDP in R&D investments in the next 5 years. But I don't think that's an official target. I am not aware of it. DST might have one, please talk to them. But we definitely have to move up on our R&D investments, it's tremendously small.

[1:14:56] Drone is a manifestation of physical AI. Its importance has come up because of the wars. See, my thinking is a little different. I don't think about the artifacts. Like a drone is an artifact. But if you look at what's the technology that's going into it, and is that something we should be betting on? Which is you're making autonomous edge intelligent, and you're giving it mobility. Which is the whole concept of physical AI. Drone is a very useful form because it doesn't use land, and therefore you can use it in warfare, transportation. What I do know is the drone ecosystem in India has grown nearly 10x in terms of multiple players coming in over the last few years. And there's some fantastic work happening. They're finding markets, which tells me that it's a positive sign. Is it enough? Technology is never enough.

[1:16:16] **Question**: *Ma'am, last question. Just simple. Please.*

[1:16:18] **Answer**: We are standing between lunch and...

[1:16:21] **Question**: *Very simple one. It may be negative also in a sense, but since you mentioned that, you know, the AI $10 trillion dollar economy, I mean, we are hoping to pitch for a 10% and you said that we have not invested so much. So I would just like to understand, what is your take on why we did not invest? Let's say... any thoughts going on that? Why we did not invest enough as compared to other countries? Is there a reason why we could not do that? Have you done some study?*

[1:16:47] **Answer**: I don't want to get into studying history, sorry. Honestly I didn't spend time thinking about what we didn't do and why we... I'm not a student of history. In school I literally scraped through history classes because I just didn't believe in history. I am a student of the future. The only thing I'll look at history for is are there any learnings on what *not* to do. That's all I'll look at. And then, start defining the future and saying how do we do this. And you, as government, have to have that mindset. Otherwise, how do we move the country forward? I'm sorry to sound like this, but as a citizen, it's my appeal to all of you. Stop using history as an excuse! We have to start looking... there are enough leapfrog opportunities coming, and we have to start looking at the future and say, what do we do newly? The playbooks are changing so drastically, 90% of that will not matter. Because we can't repeat the same thing anyway. The changing world... we have to realize that.

[1:17:58] So on that note, thank you very much everyone, and have a good lunch. Thank you very much!