Designing Indias Digital Future AI at the Core 6G at the Edge

opportunity, ensuring that India moves from being a consumer of global technology cycles to becoming a sharper of the world’s next intelligence and connectivity frontier. To kick off the discussion, I would like to invite Mr. Ashok Kumar, Deepthi Director General, Department of Daily Communication, Government of India, to deliver a keynote address. Thank you.

So my colleague panelist, the expert panelist here, the distinguished dignitaries in the hall, and other participants gathered here, Thank you,

Mr. Ashok. Thank you, Mr. Ashok. Thank you, Mr. Ashok. So it’s

my privilege to deliver the keynote address before such a gathering. So although the hall is like empty, but I suppose many of our participants are online. The theme of this session, AI at the Core, 6G at the Edge, captures the transformative journey which we have started now. So let me go back slightly back in the history. When we rolled out 2G, 3G and 4G, the vision was to connect V, human beings, and as technology progressed, we started connecting machines and objects through innovations like NB -IoT, as all of us know. Although they were not part of the original vision and we can say that those were evolutions, extensions and maybe we can also call that as afterthought.

When the work on 5G started at ITU way back in 2012, if you recall, after three years of deliberations with all the state 190 plus countries and also the sector members like industry, academia. So ITU released a 5G framework, they call it IMT 2020. And for the first time. The usage scenario, the three usage scenario envisioned by ITU included support for massive connectivity of objects and machines and also the applications which required very, very low latency. So, what we should say that for the first time technology was designed, it was not an afterthought even for machines, not only for we humans but also for the machines. As we know that the 5G journey started with 3G BP release 15 and that was also delivered in three parts, right?

Just to start early, so they had three part of releases of release 15 and then every one and half years or two years we have the next evolution of the 5G technology. And when we reached to release 18 and that is also called 5G advanced. So, basically AI, artificial intelligence began to be integrated at part of 3G. To solve the network functions or to solve some of the network functions requirement. So again, this was some sort of an afterthought, right? Because we started, our vision was not the native integration of AI into the 5G system, but as technology evolved, we started doing that, perhaps the precursor of the 6G. The shift now which we are seeing in 6G, the story is different.

If you look at the ITU framework for 6G, which was released two years back, so that has got six usage scenarios, they have envisioned six usage scenarios, and one of the usage scenarios is integrated artificial intelligence and communications. So now, the artificial intelligence is part of the initial thought itself, and more important, along with those six usage scenarios, what ITU conceived is the four overarching principles, and the fifth is the three main principles, and the sixth is the three main principles, and the sixth is the three main principles, The four, the key design principle we can say, and one of the design principle if you read is ubiquitous intelligence. So when we say ubiquitous intelligence, what we mean is that every element of our end -to -end 6G system, be it the user equipment or be it the radio or be it core or be it applications, everyone will use AI embedded natively into the system.

So the earlier generations, if you talk about connected humans and objects or machines, 6G will actually correct the intelligence as it is envisioned in the ITU document. And of course, 3GPP has started working on all those aspects. So this is a kind of… It’s a historic opportunity for me in India, particularly for our… ecosystem, that is our MSME, startup, academia and everyone. So it’s an opportunity not only to like participate in the standard so that our technology, our innovations becomes part of the standard, but also to build our own end -to -end 6G technology stack. So what are the different government efforts since I come from government, Department of Telecom, so I would also like to touch upon what are different efforts the government is trying to do to create a robust ecosystem of 6G research and innovations.

Of course, government alone cannot do everything, but whatever effort we are trying. So one of the important aspects is about whatever technology we are trying to develop, right, whatever IP we are trying to create, if that enters into the standard, 3GPP standard itself, it’s good for us that we are shaping the standards. The India is also. So, I mean, we started doing such activities from 5G onwards. Before that, we were not at all participating in the 6G, I mean, telecom technologies standard making. So, to support our startup, et cetera, onto this, so if a startup company want to, say, participate in 3GPP standards, that company has to be member of first our TSDSI and also individual members of 3GPP and that’s a cost, right?

So, at DOT, we are supporting TSDSI so that our startups can be member of TSDSI and 3GPP at a very, very low cost of 10 ,000, not 5 lakh, 6 lakh, and they can participate. So, that’s, it’s a continuous thing which we try, trying and doing. Interesting. In addition to that, as we know that unless we do our own kind of a research and technology development, even before the standard starts, building up and then take it to the standard. So to support that activity, we had come out with a scheme called 6G Accelerated Research Program. So that was floated, I think, two years back. And we have selected 100 plus 6G related projects in different area. That includes terahertz technology, artificial intelligence, machine learning, semantic communications.

And every aspect of sensing, every aspect of the vision of the 6G. And those projects are progressing. And we are trying to help them also participate into the standard. In addition to that, we have also supported some 6G related testbed like terahertz testbed and one AOC testbed, which is doing very good work as of now. In addition to that, there are many other. Programs which are sort of in progress. For example, recently we worked with. ANRF wherein we are trying to come out with a scheme wherein we are trying to build end -to -end system based on release 18 and evolve it to release 19, 20 and 21. As you know that release 21 would be the first release of 60.

So we are trying to do that and perhaps that will come very soon, maybe in the next two quarters that will be out. In addition to that, I would also like to take name of Bharat 6G Alliance here because we are also closely working with Bharat 6G Alliance as government. So Bharat 6G Alliance has created multiple working group on technology, on spectrum, on devices and some of the members of the alliance have been working on the technology and some of the members of the chair of those working groups are here as part of this session also. So, basically, Bharat 6G alliance is kind of suggesting government that what next to be done to be leader in 6G and based on that, we are trying to, I mean, shape the policies of the government.

In addition to Department of Telecom, our other ministries like Maiti is also supporting various 6G related projects. I would take name of the scheme of DST, which is RDI. So, once you have a technology, perhaps you want to scale RDI will come handy. And we have taken up with DST that telecom sector should be included as part of the sector which will be supported. In the RDI and Secretary DST had agreed to this particular aspect and whenever the schemes are getting floated, our companies, our startup in the field of telecom can actually apply. As part of DST, they also have, they have been running cyber physical programs. So, they are also, they are supporting some of the.

5G and 6G related projects. One most important thing which DOT did previous years, which was actually announced in the budget and inaugurated by our Honorable Prime Minister was 100 5G lab in 100 different institutes across the country. Those labs are actually operational. So those are some of the points where actually 6G research has also started because once you have good knowledge of 5G and if you are able to develop use cases or 5G network elements itself, perhaps you are ready to do something on 6G. And so my request to industry here, those who are online, that please adopt one or two 5G lab and try to work with them that what more can be done in the technology area.

With this, I want to conclude my address by inviting the esteemed panelists to deliberate and provide some answers to your questions. Thank you. Thank you. not only to the government, but also to the industry, MSME and startup and academia on the way forward on this

Thank you, sir. So now we are moving to our very next segment, the panel discussion. Our first speaker is Rajiv Seluja, Vice President, 5G Radio at Reliance Jio. Also joining us is Surojeet Roy, a Senior Telecommunications Leader, Head of Technology, Technology and Solutions, COE, at Nokia India. Sandeep Sharma, a technology leader and AI innovator, Vice President and Global Head of Emerging Technologies, Network Services at Tech Mahindra. The dialogue will be moderated by Radhakant Das, who heads the Technology Engineering and Innovation Function for Network Solutions and Services, NSS, at TCS. Before we start this panel discussion, I would like all the speakers to have group photographs, please. May I also request Ashok Kumar, sir, to be here?

Thank you. Thank you, sir.

Okay. Can we start? Great. So, good morning, our distinguished guests. My colleagues from the Government, Industry and Academy. All of you who are online, good morning to you all as well. So, this entire topic, which you can read out, A at the core and 6G at the edge, and Designing India’s Next Resilient, Innovative and Efficient Digital Frontier. We are at a historic inflection point where the intelligence is the basic infra. based on which the next evolution of this planet will actually continue. And we have seen until 5G, but in the 6G, a lot of hope. And we see that 6G not only emerges a faster network as an option, but as a distributed computer fabric.

It’s going to have a platform that enables the intelligence everywhere across radio, core, and age, including the satellite, which is non -terrestrial networks, and the sensor ecosystems. Devices in 6G and AI will take a major role. We’ll talk about how the 6G payloads or the designs will actually be AI -native, how it will drive the overall objective of bringing AI and 6G together. As a success. The professor has already pointed out the standards of already… will take in AI native to the 6G standards which is coming up in Magda’s next two quarters. It’s quite optimistic but yes, we are looking forward for the faster to come. And thanks for the government to give all the support to the industry, academia and the Vara 6G Alliance is also doing a great job and our Honourable Minister and Prime Minister are actually actively supporting and giving directions time to time to get this forward.

So we will focus on the edge interfaces at a scale. We’ll talk about semantic communications where like you would have seen India has really put a very strategic point of view that AI is, we will ensure that AI is kind of energy efficient. It will not be responsible for melting the data centres. It will be power efficient. And we will ensure that every compute capacity is being optimally utilized, not like we have enough compute and we will use it as much as possible. And data is a strategic fuel for this AI. And the networks, telecom networks, it’s not only 6G, but all kinds of connectivity networks, they will drive this data, this strategic fuel to the users, to the sensors, to the cloud, to the computing systems and deliver it.

So here we go. We start, I think, all our panelists are there. I think their names are already there in the backside of the screen. So I’ll just start with some of the questions. So maybe we’ll start with Surajit. Yeah. So Surajit, I have. I have the first question for you. The India in the context of India’s, in the context of 6G vision. where networks are expected to reason, self -organize and optimize across ecosystem, run, core, edge and of course when you say edge, it includes devices and the sensors. How do you see AI is transforming the RAN for the 6G in the day one? You may throw some light on that please.

Yeah, sure. So, I think we can talk about it in few steps. For example, first one is on the devices. So, we have many form factor of devices coming up. We already have smart glasses launched. We had this AR, VR glasses earlier where we could not see outside, but then now we have glasses which look more like the normal glasses we wear. But those are having this AI functionalities, right? you can do lot many you know work in the background and nobody would know that you are actually looking at something else while you are talking to a person so I think from the device perspective the intelligence is being built up in the devices handsets there was a talk that maybe all these smart glasses and wearables will take away the handset but then I guess these handsets are going to stay for a while I don’t know till when but at least for the next 4 -5 years those are going to stay and we will also have lots of wearables right and maybe some you know body patches as well which can sense your heart rate so as a person as a user I see that we will be having multiple devices going forward not only one device we will have handset, we will have smart glasses, we will have wearables right and this all will be having AI enabled devices capabilities, but because of the form factor, these devices might not be able to do all the inferencing tasks on their own, which means that there will be some inferencing help needed from the data centers, whether it is centralized or edge data centers, which means there will be lots of traffic requirements towards the network, especially in the uplink.

Okay. So, Rojit, if you would like to expand it a little further, the inferencing is now tiered or distributed, as what I am mentioning. So, what percent is, maybe you can take an average application, will be there residing in the devices or sensor side? What percent is in the RAM? What percent is in the core of the network premises? And what will go to the cloud?

Yeah, I think we do not have those exact numbers, but I think if I look at the data traffic as such, the WAN traffic. So, it is going to grow maybe six to nine times from now. till 2033 there is a position we have from Nokia Bell Labs and out of the total traffic in 2033 almost 30 % will be AI driven right. So 30 % traffic will be AI driven. It can be direct AI which can be slightly lesser but the indirect AI where you know once you use any application it drives you towards some other application and that increases your data. Maybe right now we have 5 % of the AI traffic it will go to 30 % in next 3 years. Not 3 years I think the projection is by 2033 around 2033.

So it might get to 30%. It is getting embedded to all our life faster than we have thought about.

So any of you would like to address this thing like how much of influencing would you like to see from the agency? Any of you would like to address this thing like how much of influencing would you like Like for example, of course, cloud has to do large part of it.

Yeah, on that, what I can comment, maybe Sandeep and Rajiv can also add. So first thing is, you know, it really depends on the use case. Physical AI use cases like autonomous vehicles, robots, I think autonomous vehicles are definitely picking up in US and China. But if I look at India, I think it’s going to take some time because we don’t follow rules. You know, we have a bad habit of driving. You know, I think the AI models have to tune to understand how the drivers drive in India. Right. So I think autonomous vehicles will take some time. But those are the use cases, autonomous vehicles, industrial robots, maybe robotic surgeries, where you need much lower latency.

Those are the ones where the inferencing might be needed at the edge. But I think for normal consumers and normal use cases, we can still manage with the inferencing at the central location. Right. But the main problem is. having a centralized data center establishing that is a problem because I think the power consumption and the power requirements site infrastructure, those are a major challenge and that’s why we see a trend that the data centers are gradually moving towards the edge. Maybe not driven by only the use cases but maybe driven by the infrastructure.

Yeah, a lot of, a heavy dose of this data center related concerns were there for last four days in the summit. So Rajiv if I can come to you question for you is are you witnessing a shift from telcos as a connectivity providers to intelligence utilities and how does your organizations plan to deliver intelligence at the lowest cost?

Right, you know, so in the past decade like Ashok sir also mentioned was about democratizing the connectivity, right? Today more than 99 % of India’s population is connected by high speed broadband, right? The next decade is going to be about how can we democratize intelligence. So how can the last citizen of India have the strongest intelligence ecosystem built? That is the whole objective towards which we are working. And like our chairman said yesterday, you cannot rent intelligence. We cannot, as India, we cannot afford to rent intelligence. We need to build it. We need to scale it. And the complete infrastructure that we are building, we are building up from connectivity to the cloud, to the edge, and then the intelligence ecosystem on top.

So just to add to your previous question, we believe that most of the simple agentic and inferencing workloads will get handled at the edge. And only the multi -step, multi -agent, complex workflows. those are the ones which will get handled at the central location but our whole focus is how can we create an ecosystem an end -to -end ecosystem which can ensure all pervasive and an affordable intelligence to every citizen of this country that’s the whole focus

Thank you So Rajiv, what you are actually referring to is if we distribute the inferences and the processing across so the power requirement will get distributed and also we will not have a lot of concentration of power consumption and the data centers itself it’s a good thought so maybe Sandeep, we’ll just come to you how do you see AI and 6G anchor use cases can deliver the ROI within next, let’s say one and a half year from the India’s priority sectors such as BFSI, manufacturing, healthcare, mobility and how do you see that and how do you put the metrics as a success?

I think fairly good question honestly speaking and if you look at AI and 6G are two parallel things. They are going to merge but as on today we see lot of AI traffic is getting generated. Maybe it’s 6G or 5G or maybe on wireline. And the pattern is also evolving drastically the type of AI traffic that is running. So till now, till 2G, 3G, 4G we thought of only voice and data is the actual traffic for which network should be defined. But going further, depending on different type of use cases, different latencies use case need, network has to be defined for three parallel dimensions which is latency. Latency is there today in the network but we don’t take much of attention because most of the use cases are not latency sensitive.

The other thing is coverage. Coverage is equally important. reason being the uplink sensitivity of the traffic is getting more and more relevant in the AI type of traffic. And finally, one thing that we all should be aware of, the token economy is something which drives all the use cases. How much token you are going to consume, at what pace, at what latency, drives many of the use cases, efficiency or not. So if we bifurcate it from the industry to industry perspective, if you look at the industries which are more sensitive to delay, or maybe the robotic surgery, the hospital industry, and maybe the floor machines where robots are taking all the production control, their latency plays an important role.

So we should be using 6G -centric or the 5G -centric networks to realize as good as low latency, so that the tokens which are exchanged should be acknowledged well in time, and we have a faster time to resolve. And even we have observed that even if you reduce 10 to 20 % of latency, the efficiency improves drastically so it’s no more a network KPI it’s a productivity KPI for those use cases if we talk about the coverage perspective AI is going to be more uplink heavy more bursty and we need persistent traffic around it requests will keep on coming and that persistence in uplink will only be achieved if you have a good reliable coverage and the scenarios like when you have to do a lot of tracking of the assets lot of monitoring of the assets you need to have certain use cases realized on that those are the immediate use cases that industry will look at and finally all these AI specific things will only scale when you have some national framework around it you have certain national sandboxes around it so whatever is coming into the ecosystem it’s well tested across a diverse set of vendors diverse set of customers diverse set of ecosystem players because use cases for AI may not be related to the one use cases which we have seen so far so these three dimensions we should look at and once we look at the economics of the token then coming back to the question that you asked Rishabh where the influence should happen I think it’s not about only the where but at what cost so that defines how the influencing traffic will shape up

has also urged some of the industries to take over or adopt a couple of these labs. I think what you are suggesting as a part of sandbox on the applications, they are already happening. I think more the Department of Telecom and GOI should be working on that part. Okay, Surajit, we’ll come back to you again. Again, let’s say for the next four years, until 2030, how do you see the evolution, Surajit? And starting from the devices, use cases, traffic growth, and how do you see the impact of AI derived from the networks? One is tokens, the number of tokens, we’ll start using the KPI, which Sandeep has already mentioned about. So, what’s your opinion on that?

Yes.

Yeah, I think we touched upon it, I think, but just to be more specific about it, So the uplink traffic is going to see a significant increase. So currently we see a downlink to uplink ratio of maybe 10s to 1 or 12s to 1. I don’t know the exact number, but that’s the range we’re talking about. But with this AI applications, we are predicting that this pattern will change to maybe 4s to 1, 4 in the downlink and 1 in the uplink. So what it means is that you need much higher data rates in the uplink. Today the networks are sort of not built for that, which means there will be lots of enhancements required in the network. This can come a bit from the 5G advanced, and then more enhancement will come when we go to 6G.

There will be lots of improvement on the spectral efficiency in the uplink, and then using AI in the RAN. We can improve the coverage. I’ll give you some example how that can be done. So, for example, you know, the communication between the transmitter and receiver, it involves the signal received, the interference, the noise floor, right, and the scheduling. And there is lots of data, huge amount of data which is involved there. So, I think with AI using the deep learning algorithms, we can create, you know, some logics which can help optimize this entire communication. So, and then with AI, this communication can be adaptive as well. So, we are talking about something called DeepRx, DeepTx, where Nokia is very much, you know, engaged.

And we have done some initial proof of concepts. And using that, what we have seen is that even in an environment where you have the signal to noise ratio, which is much worse than what, for example, 5G can decipher. using AI you can actually decipher those signals and that can give a capacity increase maybe 25 -30 % and what you can also do is you can have higher order modulation supported. So this is going to increase the capacity of the network and then as I mentioned the multitude of devices which will require lots of low latency use cases, much higher capacity, we are talking about minimum 400 MHz of bandwidth when we are talking about 6G. So today 5G networks are primarily running with say 100 MHz typical bandwidth.

We are talking about 400 MHz of bandwidth which might be required and we are talking about 5 times spectral efficiency. So which means 5 into 4, you are talking about 20 times more capacity coming out from 6G networks. But I think this is an evolution, right? So we are doing the standalone networks right now and you know this voice over NR, slicing, this will… you know, get, I would say, advanced and will have the entire network having slicing capabilities, voice will transform to voice for NR and then gradually you go towards 6G where you will be building the networks which are more AI native.

Very interesting point you brought in, Surajit. So what you mentioned is, it’s very interesting, I didn’t think about it earlier. You were saying even a single digit designer, I can extract more information. Actually, we are going to improve Sagan’s principle. That’s a good aspect, right? Exactly. And also, one thing if you can just throw on, the tokens are smaller packets. You just have instructions and some questions. Why it should increase the opening bust? Ideally, it should not. There are a lot of popular talks like it is going to the 6Gs or the AI is going to reverse the traffic pattern. But why? Just tokens.

So I think it depends on the, because you have to send the contextual information, right? For example, you are standing somewhere and you want to send a 360 degree view of where you are and you want to send it to the inferencing application so that it can help you, you know, understand whatever question you have. So I think that contextual information, sending it upwards, right, it will take lots of data requirement and primarily we are not doing it today. That’s the main reason because and this type of tasks will increase and that’s going to increase the uplink requirement.

Rajiv, do you want to add something to this?

No, the only thing which I wanted to add on the uplink side was that, you know, there are going to be multi -modal agents. So right now the traffic that we see is which consumer initiates, right, but when the agents and then on top multi -modal agents who are orchestrating end -to -end workflows, then they start initiating the traffic, that’s when the uplink also starts. So you will have multiple agents.

A2A traffic.

Yep.

Good. So Sandeep, the next question for you, there are a lot of AI6Z pilots are happening. I think whatever the organizations we have seen in last four days of AI Impact Summit, a lot of them are there. And what specific coordination mechanisms or co -creation models do you think we all should work together as industry, academia, government to ensure that these pilots, they align to the standards, they just don’t build on the silos while 6Z standard is maybe two quarters or maybe couple of more quarters away. So how we put the standardization as a perspective which can be adopted later stage. Also safety guidelines. A lot of safety issues will come. The more we are excited about how great things can happen, also the more of the things are exposed.

And we have seen the goals. We have seen the work, what it’s doing and how things are really getting into out of control. and so any AI maybe outcome based AI native deployment if you just can throw some light on that

Frankly speaking your question is so long I am not sure how long should be the answer I got it I got it just kidding so I think if you look at the perspective lot many good things are being done in the country there are lot many good organizations as we heard in the keynote that there is a Biosense there is a DSDA as well so lot of coordination is already in place the problem is with the pilot and the scale gap is not a technology gap is basically a gap of how we put things together in the frameworks which are scalable and referenceable as well so as I mentioned in my previous response that we need to have some national frameworks around AI native architectures once that is in place I think the quicker thing can be done is that let’s align the fundamental what type of use case are being driven and how the data needs to flow around it.

Other part is that India, as a consumer, we have a huge amount of data across the industries. And data is like bread and butter for AI. There’s no AI if there’s no data. But the data today is either siloed within the industries. If you look at the sector, they don’t combine the data together. So a national framework of putting the data together creating national exchanges where data can come in and people or organizations are allowed to train the data, train the models with the data. And we can have certain models which are more industry specific. And that plethora of variation of data, putting it together gives a very useful reference of creating frameworks which can be referenced or replicated, not only in India, globally as well.

And certain organizations are already in place to take care of that. I think more and more efforts, more and more programs are needed. thirdly if you look at more and more the safety guardrails I think we need to have certain framework in place how the AI is audited monitored within the telecom network as well we can’t allow some model to take a change of any parameter in live traffic if we can’t audit it maybe policy frameworks for intervention if certain models are changing in network parameter how and why they are changing certain explanation needs to be brought in and it can’t be done in isolation reason being if you do it in isolation again there is no clarity will come in to have a national policy around it will improve the reliability or explainability of the models hence people will come together rather than creating a differentiation of another layer of security that may encompass certain things that should be known to the larger audience and certain things that we all should do as an industry that let’s contribute more and more in these forums which government has started like Bharat 6G Alliance.

I am part of the 6G use case group, work very closely with Shokji and I think many things are already in place. We drafted certain white papers which could be referenced around AI, what type of implications it will bring into the network and we collaborate well with the 5G 100 labs as well. The things that we have done already, we should encourage them to take these things to the next level. Certain things could be referenced, certain things could be evolved. Not everything may be done right, but there is an opportunity to do everything right in 6G in terms of coordination, in terms of national referenceable frameworks. Good. Have I missed anything? It was a long question.

No, no, no. Thanks for that answer. So what actually we are seeing is in the security perspective, we are seeing we have… already a work is happening on interest in terms of policy perspective. These are in place and typically what if I have to bring in the DPI stuff or public and digital public infrastructure. So you have a lot of learning from all these sectors and how to deal with it. Even in the telcos, even in all the industry sectors, how to keep that as a creative, how that particular data which is not been seen in the telco ecosystems but still we are all responsible to deal with that. I think with the AI coming in, maybe some way we need to understand the DPI of DPI.

And just to add here, you brought a very good point. I think whenever I give a reference, the importance of open ecosystem, interoperable ecosystem, I always give an example of UPI. UPI wouldn’t have been a success if we had not promoted the open ecosystem around it. I think same mindset is needed in the AI era.

Good, good. Thanks. So Rajiv, I have the next question for you. how does the AI native telco change the way enterprises consume technology and what are the new value pools that will emerge out of it?

I think it’s a very good question see Sandeep brought out a very important aspect of latency and the second was on the uplink so first of all 6G is the solution to both the problems right enterprises in particular will benefit a lot from the advent of and the confluence of 6G and AI so there are three major drivers of value pools which enterprises can derive from 6G and AI the first would be demand analysis so you know they will be able to analyze what kind of demand is coming today their entire data is limited to the research that they do right But with the new data streams flowing in, they will be able to understand what new services can they provide to their customers and how can they embed intelligence into those new services that they are able to deliver.

The second important value pool which enterprise would be able to deliver is the workflow automation. So today, a lot of work which is manual will get automated. They will be able to orchestrate end -to -end workflows and humans will go up the value chain. That is the second important value which enterprises can derive from the confluence of 6G and AI. The third most important part which enterprises can derive out of 6G and AI is how can they make their end -to -end processes, how can they make their end -to -end security framework more robust. And see, till now, whenever we used to speak about digitization of enterprises, it used to stop at ERP implementation. Or basic business process automation.

Now, AI and 6G are going to take it to a completely different level. India is a wireless economy we don’t have fiber penetration in this country so the only way enterprises can go and reach to the last mile in this country is through the confluence of 6G and AI

so another extension to this question so how do you see sovereignty of the entire ecosystem we should deploy when you say sovereignty it is a very complex question one side that we think that ok if we make it open only it will grow at the same time sovereignty is also asked for every country maybe you can actually make entire European continent will only have one sovereign stuff what do you see on that and how much we should make it as a sovereign how much we should make it open what is your viewpoint on this

see this token economy in which we are going to go in the next 5 to 7 years so sovereignty is going to be a token sovereignty right in my point of view it will be very important for us to build our own intelligence and then deploy and scale it. We cannot be dependent on the world to deliver the intelligence to us because that will simply be too expensive for us to handle. So in order to make sure that intelligence reaches the last person in the most remote area and the last remote enterprise in India, the most important thing we need to have is have a token sovereign. We need to have a sovereign AI ecosystem, an end -to -end ecosystem starting from device to the cloud to the edge to the intelligence layers on top.

This end -to -end ecosystem has to be sovereign and we don’t have an option in this.

So you are saying end -to -end ecosystem platforms or stacks are sovereign?

Yes.

Token may or may not be sovereign or you can classify as a sovereign or as a general public one?

Correct, but we are basically calling it as a sovereign token. What I basically mean is that right from the time the request gets initiated by an agent or by a human. to the time an inference happens and the value gets delivered to the human or to the enterprise, this entire value chain has to be made in India, has to be sovereign.

So you have some view on it?

I think I was just supporting him with the gesture, but honestly speaking, the level of intelligence that country needs may not be a priority for the intelligence for other regions who are creating their own intelligence. So having a sovereign AI has an economy sense as well and has an importance for our own social values that we build in the system. AI is not only about telecom, AI is a bigger base. What we are getting as a query output as a new generation, they should be very well aware of what is right and that can be ascertained only if we have certain sovereignty developed in the ecosystem for our own nation.

So while we are talking about sovereignty, we should be very specific about it. so there is something which we need to keep as a there are certain things which you need to keep it as open stuff for learning from each other community learning across the country across the planet and all so that’s something would you like to comment and maybe I would request Surajit to also comment on that as well

I think just to start and Surajit will elaborate more the era of going further will not be a abstract one or abstract zero we need to look over the hybrid ecosystem which works best as a mix of which type of AIs as a mix of which type of compute and mix of which type of influencing and industry or the economy is going to be more use case and I would say efficiency driven so we should be leveraging which is best for to satisfy that particular use case

Surajit thank you

Yeah, I think just to add, I was reading one Niti Aayog report, you know, where we are aiming for a 30 trillion economy by 2047 as part of the Vixit Bharat initiative. So out there it was very much mentioned that there are approximately 490 million informal users, you know, workers. So for example, all these carpenters, drivers, so they are the informal users and they are not yet equipped with all the applications which might enhance their productivity. So I think from that perspective, AI use cases can be significantly helpful out here. We can have, you know, smart robots working in the, you know, fields for helping on the agriculture. then there may be use cases where maybe an electrician or carpenter, you send a video of your work and they can, that AI can generate a list of the tools they need, what all steps they have to come prepared with, right?

But for all this, I think the most important part is the model needs to be trained based on data which is coming from India. Because if you train the models based on data which is coming outside India, then maybe it is not tuned for, you know, the India specific use cases. There will be a bias there. So I think that’s why it is important.

So the cultural perspective you are hitting upon, the cultural perspective has to be understood by us. So there is a bell. So are we going to have some questionnaire sessions, Q &A sessions? Any questions? Maybe we have just two minutes left.

Can you hear? Yeah, we can. In fact, I had a lot of questions. Okay, so let me first, my question would be around interoperability. So in mobile world, we see that whatever user equipment you buy from the market, it works on all the operators, right? When we are moving towards having AI -related applications, we see there is some problem. So I was looking at meta glasses they were exhibiting. So basically, the meta glass being, say, coming out in the market will only work with the meta. So should we not think of creating some AI? API sort of architecture wherein a product created by one. user side should work in different applications. It should work with Google.

It should work with geo -applications sort of. That’s the first question which I have. The second question is about the model training which Surjeet was trying to address. So I mean the advantage of India is like any applications we can scale to a billion users. That is one. And the second advantage is that we have a huge data set on many aspects. So how to leverage these two for AI because although we may not be good at LLMs, various LLMs which we have today. Of course new companies have started. Servum and all have started working on. But we are good at having a data and the market. So how to leverage that so that. I mean models are trained here models are utilized here so these are the two questions which I have in mind thank you.

Sir I will try to attempt to answer your question the first part I think Sandeep also mentioned see this entire ecosystem end -to -end ecosystem has to be open has to be API driven and loosely coupled so that you know there is no proprietary interface from one point to another so the whole work which is going on right now at least in our organization is to make sure that how this end -to -end ecosystem can become open can become efficient and can scale you brought out a second very important point about India’s scale right and this scale is going to reduce the cost of intelligence and that affordability is a also a very important factor for us to deliver value to our you know 140 million sorry 140 crore people that reduction in cost is a very very important factor The third important point which I want to make here is that when you talk of LLMs, they are important.

But delivering intelligence is not about LLMs or training LLMs. It is about delivering this entire ecosystem to the last mile. When it comes to LLMs, the way we are building intelligence in India is in every language. These models have to be trained. So every person, whether it is from the south, from Kerala, or from Assam, or from any state in the northeast, they should be able to get this intelligence in their local language made. That is the whole work we are doing right now as part of Jio.

Thank you, Rajiv. I think just to answer the second part of the question, there is a lot of data, how we can ensure. I think a framework of having is centralized data exchanges and centralized processes. Training exchanges. where enterprise can port their data with a certain anonymization so that no confidential data passed out but industries can come and train the models specific with the data which is available from the enterprises or from the end users within India. But I think central exchange mechanism is need to be placed.

So just to add, I think democratizing the AI is also very important. It should be accessible to everybody at much lower cost and I think in that direction putting GPUs at cell towers can be one way of doing it because what you can do is when the network is not very busy and the resources are free, those resources can be given to the users to train their models or do some inferencing functions because those resources are there at every site. So that can be one way of helping on this direction.

Thank you. So you have another question? All right.

Morning. My name is Sidhu. I’m from AT &T. One quick question, now that Rajiv is also here. See, across the world, telecom companies are realizing that not having a network API exchange and then monetizing that is becoming a problem for many of the large enterprise customer use cases. For example, if a bank wants to understand their customer behavior, customers have got multiple networks, so they don’t get the visibility, right? So with OneEdge, I think Jio and Airtel have also joined hands last year. On the U .S. side, some work is happening, but I wanted to understand how much of this monetization of the network API -centric economy is materializing from India’s standpoint. I know Jio covers almost, I don’t know, 40%, 50 % of the overall population in India, so you might throw some.

I don’t know if you’re going to throw some. I don’t know if you’re going to throw some. I don’t know if you’re going to throw some. I don’t know if you’re going to throw some.

I will try to answer this quickly because the time is up and we can take this discussion offline. But see, we are committed to an open AI ecosystem to drive value. And like I said, enterprise value cannot be delivered unless the end -to -end ecosystem is open and connected. I think we are ringing the bell, but I will take this discussion offline. Thank you, sir. Thank you.

So we

have time to stop. Any other questions that we can remind of now? Thank you. Thank you, everyone. May I request Radhika and Das to hand over the memento to all our speakers? May I also request Ashok, sir, to please come on stage and kindly collect your memento? Thank you. Thank you so much. Thank you. Thank you. Thank you. Thank you. Thank you.