YT YouTube2Text

YouTube2Text → Transcripts → India AI Impact Summit 2026: Session on Generative AI and Future Networks Session

India AI Impact Summit 2026: Session on Generative AI and Future Networks Session — Transcript

by Software Technology Parks of India · 8,409 words · 1,368 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00all our esteemed guests today. Good
0:02morning.
0:03Thank you for joining us on this very
0:06important session.
0:11As we all know that's India AI impact
0:14summit 226.
0:16It's a global inflection point for all
0:18of us and we all would agree that these
0:22five days will be history history in the
0:25making for all of us and it's so proud
0:27feeling to be part of this history.
0:30uh without
0:31wasting any more time I'll request the
0:35speakers for today's first session which
0:36is generative AI and future networks to
0:40come on the stage. I'll request Sri S
0:43Abbas to please come on the stage. Sir
0:49put your hands together.
0:53Next I'll invite Shri Sham Mikar to be
0:57on the stage please.
1:01Thank you sir.
1:04We'll ask Mr. Gurinda Singh Alwalia.
1:07He is CEO Digital Twin Labs USA. So
1:11please be on the stage.
1:18Next I'll invite Mr. Manoj Gurani
1:22from Nokia. Yeah. Thank you sir. Please
1:24be seated.
1:27I'll ask Mr. Colonel PK Chri to please
1:31be on the stage
1:37and yes ma'am Janet White please be on
1:40stage. Thank you very much ma'am.
1:47So as you can see the kind of speakers
1:49you will see today. This is the power
1:51pack. It's a global speakers we have the
1:54privilege to have today.
1:57Uh I have a privilege to call Sri Anil
2:00Kumar Badwa GI for giving up the welcome
2:03address. As you all know Badwaji is a
2:06regulatory specialist over 30 years of
2:09experience in leadership roles in policy
2:12regulation economic policy competitive
2:15regulation and change management and we
2:17all know that he has been adviser to TRI
2:20also. So put your hands together Shadwa
2:23G for your welcome address. Thank you
2:25sir.
2:33>> Thank you Dhanj.
2:37Dignitaries on the dis Shriast Abad Sham
2:41Madriikar Gurinda
2:43Singh Aluya G Mano Gurani G Mr. PK Ching
2:48and of course our friend from GSMA Janet
2:52and uh ladies and gentlemen.
2:55This session has been cured on behalf of
2:59D actually I I first of all want to
3:01welcome you all and thank you for
3:04sparing your valuable time today
3:06morning.
3:08Whatever AI and AI is changing the
3:11world. We all are hearing we are are
3:14listening to things and There is a tweet
3:17of Matt Schumer which became viral three
3:19days back. We are actually living in
3:21such times. But what is making such time
3:25visible is
3:28in addition to innovation that is the
3:31underlying networks
3:34and the underlying cloud centers, data
3:37centers and today they are becoming one.
3:40The boundaries are blurring. In fact
3:42there are no boundaries. Today the
3:44agentic AI when it takes over it travels
3:48across the networks and traverses and
3:51talks to agents start talking to each
3:53other and make things happen and they're
3:56making things happen in real time much
3:59more efficiently than what we humans
4:03could do same as calculator does the
4:07calculations much faster.
4:10So the world is changing and I thank you
4:13and welcome you on behalf of uh DOT
4:16department of telecom once again and
4:21thank these speakers. In these speakers
4:24we have word of experience. Uh we have
4:27people who are managing this change,
4:31people who are actuating these change,
4:33people who are looking at global picture
4:36what this change is bringing and Believe
4:39you me all of us have to understand that
4:44unless there is an underlying network
4:47all this agentic AI will not mean or all
4:50these AI innovation will not mean what
4:53they are required to mean and that is
4:56where we come through and that is where
4:59this topic of generative generative AI
5:02and the future networks. So happy
5:05listening and once again welcome to you
5:07all.
5:12Thank you Anil sir. Uh yes he has set
5:14the tone for the session to start. I'll
5:17request my colleague Supati to take it
5:19forward now. Thank you.
5:25>> Thank you sir. Good morning everyone.
5:28Our first speaker of the day is a
5:30distinguished IT officer with extensive
5:32experience across Indian telecom
5:35ecosystem. Over his illustrious career,
5:38he has held key leadership roles in
5:40MTNL, the department of
5:41telecommunications, where he has
5:43contributed significantly to carrier
5:45services, data services, and telecom
5:48policy research. Sir has also served at
5:51tribe as adviser and principal adviser
5:54on network spectrum licensing. His deep
5:56expertise in telecom policy and spectrum
5:59management makes him a highly respected
6:01voice in the sector. Ladies and
6:03gentlemen with a huge round of applause
6:05please join me in welcoming Shri S Abbas
6:08senior DDG TEC uh department of
6:12telecommunication. Thank you sir.
6:30Distinguished co-panelist
6:34Esteemed audience,
6:38ladies and gentlemen,
6:40a very good morning to all of you.
6:44So, it's my privilege to be here in this
6:46India AI summit today with this topic of
6:51generative AI and future telecom network
6:55technologies.
6:58So we are having distinguished panelists
6:59from the industry and they are going to
7:02be telling more about how the AI is
7:06going to shape the telecommunication
7:09network and how the use of generating AI
7:13will benefit the telecom network with
7:16the network optimization and to the
7:19consumer services.
7:22So
7:23I have to just uh give a brief about how
7:29the traditional AI system were working
7:33till now and how the generative AI is
7:36going to transform the things in the
7:39coming time in the coming future
7:40generation technologies.
7:44So
7:48the traditional AI system that is
7:51already working with the telecom network
7:53and it primarily analyze or classify the
7:56data and uh gives the output which is
7:59currently now required whereas the
8:02generative AI system that generates new
8:05outputs based on the pattern learned
8:08from the large data sets. So these
8:10models of the generating AI they rely on
8:14advanced machine learning techniques and
8:16are trained on massive volumes of data
8:19to understand context generate coherent
8:22responses and perform complex reasoning
8:24task.
8:26So what is the difference between
8:27traditional AI which is currently being
8:29used and the generative AI which will
8:32bring some additional features also. So
8:34the traditional AI has already proven
8:37highly effective in telecom environment.
8:40It performs well in structured well-
8:42definfined tasks such as anomaly
8:45detection, KPI forecasting,
8:48traffic prediction, churn estimation and
8:51predictive maintenance.
8:53These use cases rely on statistical
8:55learning, supervised or unsupervised
8:58models.
8:59And uh for many operational scenarios
9:02also those requiring realtime
9:04reliability the traditional AI remains
9:07sufficient and in some cases preferable
9:10whereas generative AI becomes
9:13particularly powerful when the task
9:15require contextual reasoning synthesis
9:19of information or interactive
9:21intelligence.
9:23For example, in root cause analysis,
9:26traditional AI can detect correlation
9:28and identify likely fault domains.
9:31Whereas generative AI on the other hand
9:34can go a step further. It can interpret
9:37multissource logs, generate human
9:40readable summaries, propose step-by-step
9:42remediation workflows and even simulate
9:46alternative scenarios. Also similarly in
9:49customer support also the traditional
9:51chat bots rely on predefined scripts and
9:54intent trees whereas generative AI
9:57powered systems can handle open-ended
10:00queries maintain conversational context
10:04and generate personalized responses
10:06also.
10:08So generative AI can generate synthetic
10:11data sets for rare scenarios, create
10:14digital twins of network environments
10:17and simulate network behavior under
10:19future demand conditions.
10:22So this relationship between generative
10:26AI and future telecom networks it is
10:30going to be further broader particularly
10:32in 5G advance and 6G because these are
10:35the co-developing as mutually enabling
10:38technologies.
10:40In earlier generations, AI was
10:42introduced as an optimization layer.
10:45Whereas the emerging 6G vision
10:47increasingly described network as AI
10:49native and this means AI is not simply
10:52an add-on tool. It becomes embedded with
10:55the architecture itself.
10:57AI models may influence control plane
11:00decision, spectrum allocation, resource
11:02management and service or orchestration.
11:06And these generative AI applications
11:08they will be useful in radio access
11:11network as well as core network and
11:14provision of services. Also in the RAN
11:17domain generative AI can generate
11:20synthetic radio data sets, build digital
11:23twins of radio environments and enhance
11:26channel state information prediction.
11:29The wireless physical layer foundation
11:32model of generative AI. This will enable
11:34the large scale pre-training on radio
11:36data sets improving beam hopping massive
11:38myo organization and interference
11:41mitigation and this reduces the reliance
11:44on costly drive testing and allows
11:46network to adapt more quickly.
11:49In the field of core network, generative
11:52AI supports automated policy generation,
11:54intelligent network slicing, fraud
11:57detection and advanced traffic modeling.
12:02And
12:04there is a transition from there will be
12:07a transition from traditional
12:09self-organizing networks which depend
12:12heavily on rule-based mechanism towards
12:14predictive context aware and self-arning
12:17system and further collaborative
12:20inference architecture which models are
12:23distributed between devices and edge
12:26nodes help manage computational demands
12:28while maintaining a strict latency
12:31targets. Such distributed AI processing
12:34will be critical in 6G environment.
12:38And so that is why TC has initiated a
12:43technical contribution also to the ITUt
12:47study group 13 on benchmarking framework
12:50for generative AI in telecommunication
12:52network.
12:54So the future network will accelerate
12:57the proliferation of gen AI also. So AI
13:01will be helping the network as well as
13:03network network also will be helping the
13:05generative AI to process the data faster
13:08because 5G advanced and 6G provide ultra
13:11reliable low latency communication,
13:13massive bandwidth and integrated edge
13:15computing. So these capabilities allow
13:19generative AI models to be deployed
13:21closer to the users enabling realtime
13:23immersive application autonomous systems
13:26and intelligent industrial solutions. So
13:29network slicing will provide dedicated
13:32resource for AI workload. So as
13:35communication, computing, sensing and AI
13:38converge, the network will become
13:40pervasive platform for delivering the
13:43gen AI services at scale. And at the
13:47same time, the security and trust
13:49challenges must also be addressed.
13:52Threats such as prompt injection, data
13:54poisoning, model inversion, adversarial
13:57manipulation and privacy leakage can
13:59have serious implication in telecom
14:01network. Compromised AI decision could
14:05impact service continuity or even
14:07emergency communication. Therefore, a
14:09strong governance framework and secure
14:12model update mechanisms are essentially
14:15required for safeguarding the networks.
14:18So in conclusion, the relationship
14:21between generative AI and future telecom
14:24networks is deeply interdependent and
14:27transformative. Future networks,
14:29particularly 5G advance and 6G, will
14:32provide the high capacity, low latency
14:35and distributed computing infrastructure
14:38necessary for generative AI to operate
14:41reliably at a scale. So the success of
14:446G will therefore depend on how
14:46effectively we integrate generative AI
14:50into the network architecture while
14:52ensuring performance, assurance,
14:54interoperability,
14:56security and trust. Thank you so much.
15:07Thank you sir. As always your insights
15:10are second to none. Our next speaker
15:13began his career at C DOT and later
15:16joined the department of telecom as an
15:17IT officer before moving on to the
15:20industry where he held senior leadership
15:22roles in network planning, engineering
15:25and technology strategy. Sir has also
15:28served as the CTO of Levara group in
15:30London and was a member of GSMA's
15:32executive management committee. Ladies
15:34and gentlemen, with a huge round of
15:36applause, please join me in welcoming
15:38the president and group CTO Mobility of
15:41Reliance Gio, Shri Sham Prabhakar
15:43Madika.
15:51Thank you.
16:15Can we have the presentation please?
16:19Good morning.
16:21Thank you.
16:23Uh Abas Gi Janet Chadi
16:29and the gentlemen from
16:32uh It's a pleasure and privilege. Anil,
16:35thank you so much for hosting the DOT
16:37session. I think uh we do as a as the
16:42bandwagon of AI, bandwagon of compute,
16:44bandwagon of IT and semiconductors kind
16:47of get bigger and bigger and bigger. I
16:49think we also need to understand and
16:51highlight the pivotal and fundamental
16:53role that telecom as an organization
16:57kind of uh as an industry plays into it
16:59because this becomes the baseline on
17:02which uh the whole uh picture is getting
17:06evolved. What I want to do over next uh
17:0910 minutes max is is actually take you
17:12through some fundamental paradigm shifts
17:14that are happening and how are we as an
17:16industry, how are we as a country and
17:18how are we as a planet actually uh kind
17:21of looking forward to it and how is this
17:23going to change the way
17:28how are we fundamentally going to use
17:31this as the next vehicle on which uh not
17:35only the citizen interact interactions,
17:37enterprise interactions, government
17:39reaching to its citizens, everything is
17:41going to happen. So let's start with the
17:44fundamental forces of this this nature.
17:46We are talking about two fundamental
17:48forces. They are coming together now.
17:50Connectivity which is
17:53growing by leaps and bounds. We talking
17:55about a really incredible and
17:56unprecedented velocity at which
17:58connectivity is growing. Every 10 years
18:01you see a new generation coming in. We
18:03started in India almost now four decades
18:06back with 4G then with 2G then came 3G
18:09then came 4G now 5G is something which
18:12is which is out there which is getting
18:14advanced and we are right now on the
18:15annual of of of moving to 6G satellite
18:18is there knocking at the door fiber is
18:20getting much more pervasive so
18:22connectivity is indeed kind of moving at
18:25a really unprecedented velocity that is
18:27the first force the second force is
18:29compute and if you look at compute it
18:32has also Thanks to Moose's law,
18:34processing power doubling every year,
18:37which is almost a thousand times every
18:3910 years.
18:41How does it scale up from very basic
18:44capability of of of microprocessors to
18:46group processors to serialized chains of
18:49processors to now GPUs and TPUs and
18:52things like that where uh the ability to
18:56synthesize, process and understand data
18:58is actually going again at an incredible
19:00velocity. And when these two forces are
19:02moving it at at these great velocities,
19:05the only currency they are using is is
19:07is data. There's just one currency. In
19:10fact, network and connectivity moves
19:13data from one point to another. And
19:15compute processes and synthesizes that
19:17data for output, for insights, for
19:19actions. Now is the time when these
19:21forces are actually converging to to use
19:23this common currency cost data called
19:25data to take it to the extremely uh
19:30intuitive level extremely u uh aware
19:33level where anybody or everybody be it a
19:36human being or a machine or a process uh
19:39is able to understand the data process
19:41the data connect it to the nearby
19:43adjacent and foreign systems and draw
19:46insights and actions which were actually
19:49in the realm of science fiction as as as
19:52early as 5 to seven years. I mean before
19:54covid we could not have even imagined
19:56the way AI has used connectivity and has
19:59become this pervasive in in our our
20:02world of of industry.
20:06Now when these two forces by itself are
20:09so fierce what happens if one starts
20:11complimenting the other? If you look at
20:13what compute does to connectivity,
20:15compute makes connectivity aware which
20:16is one of the biggest single largest
20:19shifts which is happening in in these
20:21new generations we are talking about 5G
20:24to 6G. The network is aware of what is
20:26happening. The network is is
20:27self-healing. The network is
20:29selfoptimizing. The network is
20:31personalized and the network is real
20:33time. So the pipes were always there.
20:35But what connect compute has made them
20:37realize is these pipes can now be aware.
20:40These pipes can now decide. These pipes
20:43can now predict. These pipes can now
20:45automatically
20:47drive actions. And that is where the
20:49multiplier effect of compute and
20:50connectivity has come. Other way around,
20:53how is connectivity kind of uh driving
20:56the compute capabilities. So while
20:58compute makes connectivity aware,
21:00connectivity is making compute flow in
21:03the way it has never before. It is now
21:05flowing in your vehicles, it is now
21:07flowing in the devices, it is now
21:09flowing in the edge. getting closer to
21:11the customer. Again, inference is real
21:13time,
21:14not only computed real time but also
21:17transmitted real time but also acted
21:18real time and that is where connectivity
21:20has aided compute to drive it to the
21:22next level. The intelligence is now
21:24distributed and this distribution is on
21:26the hands and legs of of connectivity
21:28which is driving this apart. So if you
21:30look at the forces one impact on another
21:33another impact of other actually they
21:35are not adding it is not 1 + 1 they are
21:37multiplying it is it is it is a 10x
21:39syndrome that is happening the moment
21:41you see how connectivity is impacting
21:43compute and how comput is impacting
21:45connectivity going forward and that is
21:47where this amalgamation of of of compute
21:49and connectivity these two forces on the
21:52currency called data is is going to make
21:55us live the next decade make excel in
21:57next decade and make the life totally
21:59different and totally pervasive of
22:01intelligence going forward. It is
22:03everywhere. Now you're talking about
22:05individuals. We are talking about
22:06assistants that understand the context.
22:09We are talking about translation things
22:11real time. We are talking about
22:12accessibility real time. Uh smallest of
22:15jobs about booking, planning your
22:18travel, booking your air tickets.
22:19Smallest of jobs around uh getting
22:22assistance, schedule your things,
22:24getting assistance, write email for you.
22:26It is right now available to every
22:28individual and it is only this is just
22:30the tip of the iceberg. I mean the
22:33ability of individuals to use this AI in
22:36a connected world is only limited by
22:39their own personal imagination. Sky is
22:41really the limit or or or even beyond.
22:44Now that's the individual context. If
22:46you move to the enterprise part
22:48enterprise and we've been hearing this
22:49about now 3 four years on on this
22:52industry 4.0 concept. How do you
22:56brutally and deeply automate
22:59every action that an industry does? It
23:01could be a manufacturing industry. It
23:03could be a health industry. It could be
23:05education industry. The amount of value
23:08ad this intelligence does on top of all
23:10processes, all capabilities, all supply
23:12chains, all assembly lines is
23:15unimaginable. And imagine this aided
23:18with market intelligence, aided with
23:21customer awareness for that particular
23:22enterprise. We are actually talking
23:24about workflows that will self-generate
23:27themselves. Workflows that will self
23:29create themsel and drive the fulfillment
23:31cycle to a totally different level of uh
23:34uh
23:36satisfaction to the end customer.
23:39So we're talking about individual, we're
23:41talking about enterprise, but most
23:43importantly how does government and we
23:46we have Abbas GI talking about how
23:48government is is is kind of drive this
23:50whole uh forum today and and and this
23:53week is talking about how AI is going to
23:55help us as a country. But imagine the
23:58tools, imagine the value it it gives in
24:01the hands of uh government. There are
24:03the digital infrastructure is already a
24:06humongous success story in the country.
24:08Digital payments for us. Digiatra is an
24:11example. Aadhaar card is one. Your
24:13banking is all online now. I think these
24:15capabilities I think we are any which
24:17way almost a decade ahead than rest of
24:20the world even on developed countries of
24:21the world. Now imagine
24:24putting intelligence on top of it.
24:26Imagine putting awareness on top of it.
24:27Imagine making it inclusive for the
24:29citizens. Imagine making it available
24:31and accessible to the last man in the
24:33last village in the last grand panchayat
24:35of the world. How is that going to kind
24:38of scale up the reach and the decisive
24:41capability of the government to make the
24:43citizens life better? So if you look at
24:45all the three vectors, you and me as
24:47individuals, you and me as uh corporate
24:50citizens who are driving our own
24:52respective industries and companies and
24:53finally you and me as as as Indians, we
24:56are taking it to the next level. It is
24:58going to multiply. it is going to drive
25:00it to the next level. The networks
25:02definitely have come off age and I think
25:04I mean one of my favorite stories is how
25:06about these G's came about. You're
25:09talking about every decade when we
25:11started with voice moved to messaging
25:13moved to data moved to video. Now we are
25:15getting into a level which is so
25:17inclusive which is so intuitive that
25:19there is a human AI harmony in the next
25:20generations to come. But this is not
25:22only the networks that have come of age.
25:24The AI now in the coming years is is
25:27really really coming of age from basic
25:30compute to shared resources. Now from
25:32recognition to intuitive it is now
25:35actually merging and where these two
25:37roads meet the connectivity and the
25:39compute with AI and with the next
25:40generation of network. Uh the only
25:43people to kind of gain is is all of us
25:46as as citizens Indians and as proud
25:49members of this industry which is
25:51shaping the world. Thank you so much.
26:01Thank you sir. I am sure your insights
26:04uh really carry a lot of value for
26:06everybody in the hall. Uh also uh Shri
26:09Abbas has not only joined us in his
26:11capacity as senior EDGTC sir is also the
26:14CMD of TCIL.
26:17uh and a request to the speakers. We are
26:20expecting the honorable minister at
26:2111:30 this very hall. So at request if
26:24uh the speakers can be a little mindful
26:27of time. Our next speaker is based in
26:30the United States. He's the founder and
26:32CEO of digital twin labs which designs
26:35and deploys strategic digital platforms.
26:38Previously sir was the CTO of IBM North
26:41America for blockchain IoT and cloud.
26:44Sir was also nominated to the
26:46prestigious IBM Academy of Technology.
26:49Ladies and gentlemen, please join me in
26:51welcoming Shri Gurinder Singh Alwalia,
26:54CEO, Digital Twin Labs.
27:01Can you hear me? Okay.
27:06Thank you.
27:08Um, there was a little bit of audio
27:11difficulty sitting in that corner. I
27:14hope everyone can hear me okay including
27:16my co-panelists.
27:18Um the most profound change of AI
27:23is
27:25in the cost of predictability which is
27:28coming down.
27:30But even more profound and really the
27:32driver behind AI
27:35is that everything
27:38everything can now be represented as a
27:41language. So our definition of language
27:44has been reinvented by AI and I'll pass
27:48through a little bit about that.
27:51My thanks first to the department of uh
27:54telecommunications
27:56to the entire team at uh CEO AI and then
28:01particularly
28:02to General Cocher who might have stepped
28:05away at the inopportune moment. Um and
28:10to brigadeier son, thank you so much for
28:12inviting me as a speaker over here.
28:17Continuing in the theme of imagine,
28:21I will have you imagine something
28:23different.
28:25Imagine a child
28:29flying a red kite
28:32on a green field
28:35under the blue sky.
28:39A child
28:40flying a red kite
28:43on a green field under a blue sky.
28:48You heard my words.
28:50You heard my language.
28:53But what it created is a picture in your
28:56mind, right?
28:59You turned my words into pixels.
29:05I can represent a picture through
29:07language.
29:09Similarly, you can represent anything
29:12through a language which is what AI is
29:14doing.
29:17I'm a computer scientist by training and
29:20I come to you as a practitioner of
29:22deploying disruptive solutions which is
29:25what I've done for the last 30 years
29:27mostly in the US but globally.
29:32And even though I'm a computer
29:34scientist, I lean on economists that
29:38might be amongst you to explain
29:40technology because technology cannot
29:43explain disruption.
29:46And these are three
29:49corpus of knowledge which I personally
29:53lean to and I find it useful to many in
29:55my audience and in my circle and that is
29:58the theory of the firm which basically
30:01explains why transaction costs which is
30:05an economic phenomenon
30:07has advanced disruption and advanced
30:09technology of which we have illustrous
30:13scope panelists over here from the
30:15cellular industry and we all know what
30:18that has done that has done to the call
30:19of communic to the cost of
30:21communications and e-commerce and so on
30:22and so forth. The second author explains
30:27that these disruptions not only happen
30:30but they are actually predictable.
30:32It's a dry book but if you'd like you
30:35can take a look at uh Carlott Perez a
30:39Argentinian professor and the third one
30:42is also a Nobel laureate like the first
30:45one explaining something called
30:47institutional economics and this is
30:48actually very powerful and most recent
30:52it says that you can organize without
30:54organizations
30:58right you can organization you can
31:00organize without organizations so The
31:02army and I know there are many army
31:05soldiers over here even though it is a
31:08very strong institution and organization
31:11at the field when it's detached needs to
31:13organize organize its formations without
31:17central organizations in some in some
31:19cases.
31:21So the emphasis is on organizing
31:25as opposed to the organization.
31:29If we look at a systems a pro a few
31:31progressions we've been through
31:33tabulating systems we've been through
31:35programming systems and we are now into
31:36intelligent systems I speed through some
31:38of the slides in the interest of time
31:41this one is a little bit more
31:42interesting where we had a little bit of
31:44data and it was simple it was calculable
31:47right and we could do it mathematically
31:49through equations then it became more
31:51complex but it was still calculable and
31:53we continued to do it through algorithms
31:55and programming now algorithms and
31:58programming
31:59programming has run its limits. If if
32:02you are designing an autonomous car
32:04connected to a network that has to make
32:07onboard decisions,
32:09you cannot possibly compute the number
32:13of permutation combinations and the
32:16complexity becomes infinite. And that's
32:18why we now lean on data to find data,
32:21data to build the algorithms. And we
32:24lead into a a world that is not
32:26programmatic. It's declarative,
32:29right? My simple example on a quick
32:32tangent on declarative is if you want
32:34chole, right? You just want chole, you
32:38don't have to define the recipe of how
32:40to make the chole, right? So you want to
32:43go from A to B. You put in the GPS
32:46address and it just takes you there.
32:48That is not a fig. That's not science
32:51fiction. It's basically how cars are
32:53beginning to drive now. So uh I'll skip
32:57this around
32:59uh except for the for the end which says
33:02given the agent meaning given like chat
33:05GPT or grock you're given an agent all
33:08it has to do is it has to look for the
33:10tools the edges of the graph and then it
33:13figures out what nodes to traverse
33:17through in order to meet your command.
33:20Right? Human behavior is by definition
33:24declarative on what what you want and
33:27the technology should serve that to you.
33:31Satya Nadella um you know the beloved
33:35CEO that India produced and is there in
33:39Us
33:41but I think he's around here now
33:43basically said that the application
33:45layer is collapsing into agents and the
33:48seller industry and the application
33:50industry saw as the infrastructure
33:52shifted the application and the nature
33:54of applications also shifted.
33:57What does this mean? It means it's
34:00moving from imperative to declarative
34:02right graphs are moving to be
34:05non-deterministic flows
34:08and generation is retrieval. When we
34:11talk about generative AI
34:14it is generative as opposed to what? It
34:18is generative as opposed to retrieval
34:22like retrieving from a database. This is
34:24not something that exists in the
34:26database which is how the past errors of
34:28computing this is being generated. Uh
34:31now the blue ones is is of a little bit
34:34uh attention. The logic shift is from
34:38predictive
34:40to predictive neural networks to
34:42predictive machine learning and then to
34:45reasoning and reasoning is then
34:48packetized for those in the network
34:50world. similarly as tokenization.
34:55So the flow shift is from data to tokens
34:59to process and then to predictive. So
35:04language is no longer just a
35:06communication.
35:07It is a representation of the physical
35:10and the digital world. It is a
35:13representation of our declarations.
35:16It is an expression. It is a thought.
35:19And more important to those of us in the
35:21networking and then the computer world,
35:24it is computation. Language can be
35:26computed.
35:28So
35:31I advance the
35:33u the sentence that everything is a
35:36language and language is everything. If
35:39we begin to understand this, we begin to
35:41understand that the era pioneered by
35:45Sand Microsystems. Some of you might
35:47know that company and I actually had the
35:50privilege of working for Sun
35:52Microsystems. They coined a phrase
35:54called the network is the computer.
35:57The network is the computer and they
35:59pioneered that phrase in the 80s
36:03and it is increasingly every few years
36:05and decades become even more true. So in
36:09that sense the language the network is
36:12also now the language. Text is already a
36:16language. Music is a sequence of MIDI
36:18events. Audio is numerically represented
36:21already. Images are represented as as
36:23grids as pixels. Video is images plus
36:26the sound. So on and so for so forth.
36:30You can take it to 3D models for DNA,
36:32molecular research, pharmaceutical and
36:35so and so on. So
36:39I now come to how I started which is
36:43when you visualize a child flying a red
36:46kite on a green field under a blue sky.
36:50You identified a who you identified a
36:53what you identified a field. You
36:56identified when it's a blue sky and you
36:59identified details of the color. We do
37:02that in our human brains and machines
37:04are beginning to do that as well.
37:06only because we are able to represent
37:08these all these artifacts not just text
37:13but all these artifacts
37:16as a language. All right. So uh that's
37:20pretty much it. Everything is a
37:22language. Language is everything and
37:25because all seeing is reading. So I
37:29tried I tried my best not to talk as a
37:31computer scientist and we begin to now
37:33ex see the impact of this expression of
37:37language which has to be computable not
37:40just not just by computers but by the
37:42intervening networking
37:45um intelligence. Thank you so much.
37:54>> Thank you sir. Uh ladies and gentlemen,
37:56we're expecting the minister here any
37:58moment. I'd request everybody to be
38:01seated even after the speeches are over.
38:04Uh our next speaker, he's a business
38:07technology leader with 30 years of
38:09experience across Nokia, Erikson,
38:11Seammens, Reliance, Korean and RFS. In
38:15his current role as the CTO and head of
38:17strategy at Nokia, he shapes technology
38:20vision and strategy for the India
38:22region, driving 5G and 6G adoption,
38:26nextgen technologies in telecom spectrum
38:29readiness and ecosystem development.
38:31With a huge round of applause, please
38:33welcome Shri Manoj Gurani.
38:42Hello, Namaste everyone and good
38:43afternoon. Um,
38:46time is limited.
38:48When we talk about AI, the first thing
38:50which comes to your mind is algorithms.
38:53And when you talk about AI in India, the
38:56first thing which comes to your mind is
38:57the scale. Scale of 1.4 billion
39:00population, 1 billion broadband
39:02customers,
39:04diversity, 22 languages. So I mean India
39:08is like a sandbox. So something if works
39:11in India is going to be a guarantee that
39:15it's going to work very successfully in
39:17all parts of the world. I'm going to
39:19touch upon few points which are related
39:22mainly to the networks because that's
39:23the area where I come from. So the first
39:27thing is connectivity which is pretty
39:29obvious with so much of AI tsunami and
39:32the digitalization which we are all
39:34witnessing. The good news is that
39:37network will be at the center stage.
39:39There is huge focus on the connectivity.
39:42The connectivity of course will
39:44transform from connecting people in the
39:462G to the broadband to the connecting
39:49devices to what we call it as connecting
39:52intelligence.
39:54There are huge amount of trends which
39:56are emerging.
39:58We believe first of course is going to
40:00be the volume of traffic. The volume of
40:03traffic with the onset of AI is going to
40:06be humongous. I mean we forecast that
40:09the impact of AI alone will lead to a
40:12bump up of almost 30 to 40% of the
40:14traffic in the mobility side alone.
40:17That's the first impact and are we ready
40:19to actually uh cater to this kind of a
40:22network. The second is this kind of
40:25network is highly bursty,
40:28highly unpredictable
40:30and from the download heavy it becomes
40:33more upload heavy. So these are the
40:35trends which the networks will have to
40:38really focus on and therefore the broad
40:41philosophy in the connectivity is that
40:44you have to come to the paradigm where
40:46networks can sense, networks can act and
40:50networks can adapt.
40:53So that's the first part on the
40:54connectivity side.
40:57The second aspect which is pretty
40:59obvious is the autonomous networks.
41:03The good news is that when 5G came and
41:05when 6G arrives tomorrow, the data
41:08availability in terms of the structure
41:10of the data, the quality of the data,
41:12the volume of the data that's already
41:13kind of available
41:16and the AI assisted operations are used
41:19in a big way already for the past 3 four
41:21years. beat your traffic forecasting,
41:24beat it at your network operations which
41:27uh Sham already talked about uh beat the
41:30normal autonomous uh uh actions and the
41:32use of AI agents in the network
41:34operation that's already happening in
41:36pretty much a big way but what will
41:38actually happen is with so much of
41:41autonomy you will actually face the
41:43challenge of what we call it as black
41:45bodies syndrome which means that if you
41:48provide so much of autonomy to the
41:50networks a stage will come where you
41:53don't know that for any action do we
41:55have any accountability I think that's
41:57one part where we have to be very
41:59cognizant of
42:01uh especially in the mission critical
42:02networks and telecom today is not the
42:05telecom network it's actually a critical
42:07infrastructure so autonomy is going to
42:09play a big role simply because you
42:12cannot do the operations the traditional
42:14way the volume of traffic the plethora
42:18of technologies like from 2G 3G and
42:21going all the up to 6G. The amount of
42:23spectrums which we have, the amount of
42:26diversity which we have, all this is
42:28going to make it very very difficult. So
42:31the a use of AI in operations is
42:33something which is absolutely very
42:34mandatory.
42:36But to watch scale the autonomous or the
42:39autonomy can be achieved is for us to
42:41see. I think I'll leave with the uh last
42:45message which says that the autonomy you
42:48should actually have the manual
42:49override. I think that's very very
42:51important. I think the human angle, the
42:54strategic oversight, I think that should
42:56not be left.
42:59The third aspect which I'd like to talk
43:01about will be sustainability
43:03and as telecom
43:05the the uh telecom as an industry
43:08contributes maximum after the aviation
43:12we would have done great in terms of
43:14doing developing the technology in
43:16trying to reduce the energy per bit. We
43:19have made rapid strides. Lot of work has
43:21been done. But despite that the CO2
43:24emissions have increased and the two
43:27reasons are the volume of traffic has
43:28increased, the densification of the
43:31networks in terms of plotting more base
43:33station has increased. So when 6G
43:36arrives,
43:37we have taken a very ambitious target
43:40that hey can we at least at the network
43:42level reduce the energy consumption by
43:45half. That's a very ambitious target. So
43:47on a baseline of 2019, can we reduce the
43:50energy at the overall network given the
43:53volume of the traffic, given the
43:54densification, can we reduce it by half?
43:57I think that will be uh a real uh
44:00contribution which the technology has to
44:03do. Uh the good news is that AI as an
44:06application is being used in a pretty
44:08successful way. I think there are energy
44:10savings applications which have yielded
44:1215 to 20% of energy savings. But this AI
44:16is sitting at the top of the network.
44:19Going forward, we have to move the AI
44:22native way. Which means we have to deise
44:24the products and the technologies to
44:26make them AI native. In true sense,
44:30in the end for a country like India, I
44:33would say that there are four things we
44:34would should be doing. We should
44:37actually look at devising responsible AI
44:40frameworks and I think rapid strides
44:42have been made in that direction.
44:44The second thing is that we have to have
44:47robust data governance system. Again, I
44:50think the government has taken very very
44:52good steps in that direction.
44:54The third is to have a solid
44:56collaboration between the industry, the
44:58academia, the startup and the tax side.
45:02That's very important. We should
45:04actually move into a very very
45:05harmonized world. With the current
45:08geopolitical scenario, there is a huge
45:10danger of fragmentation that should be
45:11avoided. Telecom is successful because
45:15we actually follow standards and I think
45:17that's something which we should also do
45:19when we are leading uh when we are
45:21designing the AI frameworks
45:24and in the end for a country like India
45:27we have to think about India first and
45:29we have to come up with the models which
45:31think about the region which are more
45:34vertical specific which have more
45:35context.
45:37So with that uh thank you very much.
45:45Thank you sir for sparing your time and
45:47your valuable insights. Our next speaker
45:51was commissioned into the Indian Army's
45:53core of signals in 1986.
45:55He brings over four decades of
45:57distinguished experience across defense,
46:00corporate and academic domains. Ladies
46:02and gentlemen, with a huge round of
46:04applause, please join me in welcoming
46:06the program director of Tech, Colonel PK
46:10Chri. Sir
46:19uh good morning ladies and gentlemen.
46:22There has been a series of very
46:25informative talks by most imminent
46:27co-panelists who are here. Uh my
46:30presentation or my discussion with you
46:33for next 15 minutes is slightly
46:35different. It's a purely communication
46:38theory and information system
46:40researchers perspective
46:42on the employment of generative AI
46:47on network optimization.
46:49Because of the constraint of time the
46:52topic given to me was generative AI and
46:54networks. So in 10 15 minutes nobody can
46:57address that vast canvas. So I selected
46:59a small part of it where AI is getting
47:04massive traction in the global computer
47:07science research community and that is
47:09the how to use generative AI fruitfully
47:15for network optimization.
47:18So basically I am working on a survey
47:20research and some some parts of that
47:23research is going to be presented to you
47:25in next 10 odd minutes. So can I have
47:28next slide?
47:31Can I have next slide please? Yeah. So
47:33my aim is Yeah. Okay. Thank you. So my
47:37aim is uh two forth to acquaint you with
47:41what is happening latest in the global
47:44research community about the network
47:46optimization by using generative AI
47:49which we have been using so forth so far
47:52using conventional mathematical models.
47:55where does the generative AI fit in and
47:58the second part is while doing so we
48:01will see a taxonomy of the tools and
48:03methods being employed. These are my
48:05twin aims of next 15 minutes of this
48:07course with you.
48:11So we have been doing network
48:13optimization so far in all our wine and
48:17wireless networks for three things. We
48:21estimate the environment. For that
48:23estimated environment, we allocate the
48:25resources and after we have allocated
48:27the resources, we control and monitor
48:30it. This is what we do in optimization.
48:32Needless to say, there are very imminent
48:35experts who run mobile network. It is
48:37real time, very dynamic and so forth.
48:39But the methods which are used are both
48:42non-convex,
48:45B level and stockistic processes.
48:49Putting it very simple term a non-convex
48:52optimization is like what how do you
48:54design a aircraft wing it has number of
48:58multiple optimal solutions. You see a
49:01mountain if there is one valley it's a
49:04convex solution optimization. If there
49:07are number of passes to cross that
49:09mountain it is non-convex. When we are
49:12talking about generative AI we have to
49:14see how this thing is changing.
49:18So
49:20there are largely two type of generative
49:22AI models which the computer researchers
49:24throughout the world are employing and
49:29while we are very familiar with the
49:30large pre-trained models like chat GPT
49:34or cloud A or anthropic what we what we
49:36use but let us not forget the first one
49:39which we have written is the generative
49:41diffusion model it's like a artist
49:43making a painting he puts some broad
49:45brushes on the canvas it may appear
49:47appear random to you but ultimately it
49:50translates to a very good masterpiece.
49:52So here what we do in diffusion
49:54generative AI we put some random noise
49:58make it gossian and after we then we d
50:00noiseise it to extract the object of
50:04interest. So both of them are being used
50:07by using two methods. One is either they
50:12are people are doing oneshot
50:13optimization or a marov decision
50:17process. Put it very simply in oneshot
50:20each decision is made independently.
50:23And in marov we have a sequential
50:26decision making process based on the
50:28last probability state. Both of them are
50:31being used. So two type of generative AI
50:34diffusion model and large pre-trained
50:37using both oneshot and markoff process
50:40we will come to this is what is
50:42happening right now and gradually people
50:45are making way for generative AI models
50:49to do such tasks. We will try to
50:51separate hype from the fiction. There is
50:54a lot of hype that generative AI is
50:57going to get deployed and to do
50:59wonderful things but there are real
51:00challenges and real challenges is a
51:03understatement. People are struggling.
51:05So I will I will come to that in some
51:07time. But what has brought a urgency to
51:09the entire process of changing over from
51:12conventional mathematical models either
51:15oneshot or marov to generative AI is the
51:17advent of 6G standards where it has been
51:20mandated that the optimization process
51:24will be a IMLdriven. So there is no
51:26choice we have to get into it and from
51:29the hardware heterogenous interfaces
51:32come to a softwaredefined world. So this
51:34is the urgency of doing so.
51:38I will skip the general part of the
51:41general diffusion model and large
51:42pre-trained. You must be aware of the
51:44many differences which are there.
51:47What we do is the four things which is
51:49at the bottom of this slide. We optimize
51:52some variables like say traffic demand
51:57using some input parameters like latency
52:00or throughput to achieve a objective.
52:05to achieve a objective using some
52:08constraints. So coming to the solution
52:11which is emerging in the present world
52:14is twofold. Generative AI in network
52:17optimization is being used as two-fold
52:20models. One is as a solution generator.
52:23It reads the entire data and comes up
52:25with a one short solution or it is
52:29coming as a uh deep learning
52:33re deep learning reinforcement model
52:36where it takes sequence by sequence. Now
52:39both are able to do optimization to a
52:41large extent but there are two
52:42challenges. The DRL policy in diffusion
52:47model is very successful but it has got
52:49latency
52:50because iterative steps are being taken
52:52to solve it. The the the large
52:57trained pre-trained models are having
53:00lot of computational issues because it
53:02is performance cost is not matching up.
53:04So researchers are trying to solve it.
53:07What is ultimately happening as late as
53:106th of February uh this uh This great
53:14researcher who has got more than 1
53:16million citations that MIT scientist uh
53:18Kiming Hay has published a paper which
53:21has created lot of buzz. He has been
53:24able to solve the diffusion model
53:26latency problem by using something
53:29called drift
53:32model. So why I am saying I will
53:35conclude by saying there are lots of
53:39promises and lots of challenges. The
53:43research community is heavily invested
53:45in solving the challenges. AI is fast
53:48emerging. What I say today may not be
53:50valid one week after now because new
53:52models are coming up. Such a heavy
53:54investment has been made in uh global
53:58market in AI that the failure is not an
54:01option and the research community will
54:04emerge. Network optimization is the most
54:07viable likely application and we will
54:10succeed in doing so. With this I will
54:12close and thanks a lot to organizers for
54:14giving me a chance to share my thoughts.
54:24Thank you sir. We'll be joined by the
54:26honorable minister any moment now. I'd
54:28request everybody to please remain
54:30seated.
54:31And uh last but absolutely not the least
54:34our next speaker. Her role entails
54:37providing public policy support for
54:39mobile network operators across the
54:41globe on telecom on data privacy,
54:44spectrum policy, IoT, 5G and competition
54:48policy. She is a former British telecom
54:51director where she held senior roles in
54:53the areas of strategy, portfolio
54:56management, product management, mergers
54:58and acquisitions and IT. May I request
55:01she Miss Janet White to kindly join us
55:05with a huge round of applause, please.
55:11>> All right, I know we're very short on
55:13time and that a lot of the things that
55:15maybe we I was going to say have already
55:17been said. So, let's try and keep my uh
55:20comments and uh short. First of all,
55:24though, I must say it's wonderful to be
55:25here in Delhi, a place where it really
55:28feels technology alive. When you walk
55:30through the halls here, you can feel it.
55:33It's absolutely amazing.
55:35So, we meet at a time when the world is
55:37trying to understand what generative AI
55:40is and what it means for society.
55:43And what we see across markets,
55:46industries, and especially across the
55:48region is that AI is no longer something
55:51happening around our networks. It's
55:54something that's happening through them.
55:57Let me share a simple example. In one
56:00Asian market, after the latest wave of
56:03AI assistance became widely available,
56:06mobile operators saw a tenfold jump in
56:10related traffic in just one month. An
56:13astonishing acceleration.
56:16This is not an incremental change. It's
56:18a shift in how people interact with
56:21digital services. And this is not an
56:24outlier. GSMA intelligence estimates by
56:27the end of this decade AI services could
56:30drive up to half of all additional
56:33traffic fueled by multimodal assistance
56:36generated content and richer inter
56:38interactive applications.
56:40Put simply, AI is shaping the rhythm of
56:44our digital lives and our networks are
56:47already feeling it.
56:50Nowhere is this uh transformation
56:52transformation more exciting than here
56:54in India. India is emerging as one of
56:57the world's most dynamic digital
56:59economies. And that's not an empty
57:02phrase. GSMA analysis uh shows that
57:07Indian enterprises are projected to
57:09invest nearly 11% of their revenues in
57:13digital transformation through to 2030.
57:18far above developing markets average and
57:21India ranks among the top five countries
57:24globally for digital transformation.
57:28India has shown the world what can
57:30happen when digital infrastructure and
57:33public uh public purpose align where
57:37through infrastructure through uh Adahar
57:40UPI or the broader DPA DPI ecosystem
57:45and now as India turns its attention
57:47towards AI the country has an
57:50opportunity not only to adapt the
57:53technology but to help shape its
57:55governance its standards and trajectory.
57:59So what do we need to get right? Four
58:01things.
58:03Infrastructure that matches the moment.
58:05AI is pushing networks into new
58:07territory. Higher uplink demand, tighter
58:10latency expectations, unpredictable
58:13traffic patterns.
58:15Future networks from 5G advance through
58:18to 6G must be built not just for
58:21capacity, but for intelligence,
58:23automation, and real time
58:26responsiveness.
58:27Secondly, we need policies that support
58:30investment across the whole digital
58:33ecosystem.
58:35India and other countries has taken
58:38important steps to support digital
58:40infrastructure including incentives that
58:43have helped accelerate data centers and
58:46cloud expansion.
58:48These efforts have strengthened the
58:50compute side of the ecosystem and c
58:54created momentum for AI adoption.
58:58But the network layer often does not
59:00receive the same level of targeted
59:02policy focus even though it's a network
59:06that ultimately carries the AI
59:09intelligence the data and the
59:11experiences that AI enables AI enables.
59:16This isn't about favoring one part of
59:18the ecosystem. It's it's about balance.
59:21A future digital economy requires end to
59:25end support. Compute, cloud, data
59:28centers, and the mobile networks that
59:30connect them. Thirdly, trust and
59:33resilience. This was mentioned earlier.
59:35Uh people will only en embrace AI if
59:38they trust the systems behind it. Trust
59:42depends on cyber security fraud
59:44prevention, responsible data practices
59:47and governance frameworks that protect
59:49users while empowering innovation. Trust
59:52is not a nice to have. It's the
59:54foundation of a digital progress.
59:58And finally, openness and collaboration.
1:00:01AI networks, devices, clouds, and data
1:00:04flows are all intersect. And we we've
1:00:06never seen them before. No single
1:00:09organization or a country can navigate
1:00:12this landscape alone. Progress will come
1:00:15from partnerships across sectors, across
1:00:18borders and across digital ecosystems.
1:00:21Whether through shared infrastructure,
1:00:23interoperable APIs like those supported
1:00:26by GSMA open gateway or collab
1:00:29collaborative approaches to security and
1:00:31innovation.
1:00:33So in closing, as we look ahead, I
1:00:36believe India has the potential not only
1:00:38to participate in the AI era, but to
1:00:41define it. The world is watching how
1:00:44India scales AI responsibly, how it
1:00:48strengthens its digital foundations and
1:00:50how it ensures that growth is inclusive
1:00:53and impactful
1:00:55as and as generative AI converges with
1:00:585G, 6G and cloud, India stands at a
1:01:01remarkable inflection point with the
1:01:04opportunity to lead in a way that lists
1:01:07the region and sets a global benchmark
1:01:09for the future of intelligent networks.
1:01:20That was an exciting session I would
1:01:23say. Various aspects got covered I
1:01:25think. So quite a breadth of it u a
1:01:29global perspective also. Uh thank you
1:01:31all for your speakers for your u
1:01:34insightful presentations. Uh and
1:01:36obviously thank you all for keeping it
1:01:38brief. We understand you would have
1:01:40thought to be going on and on but so uh
1:01:43we'll request uh director general COI
1:01:48left journal Dr. SP coacher to come on
1:01:51the stage and felicitate the speakers.
1:01:54So
1:02:00uh I think so Shri Abas G has already
1:02:02left us. So we'll start with Sham Mikar
1:02:05GI. Thank you very much sir.
1:02:16I'll ask Gurindra G to please.
1:02:21Thank you very much sir
1:02:24for giving us a global perspective.
1:02:30Next I'll ask Manoji to please uh
1:02:34thank you very much sir.
1:02:44Thank you very much sir.
1:02:47And yes
1:02:50Janet ma'am please thank you very much
1:02:53for giving your presentation ma'am.
1:02:59Uh we can have a group photograph
1:03:01please.
1:03:06Sir,
1:03:09so group photograph
1:03:22a huge loss of applause for everyone.
1:03:24Yeah, thank you.
1:03:30Thank you. And kindly be seated in the
1:03:33hall itself. The minister is about to
1:03:36come at any moment.
1:03:38Uh we will have the pleasure of having
1:03:41interaction with him also.
1:03:43We'll also request the participant of
1:03:45the first session also to stay back.
1:03:48We'll have a photograph with the
1:03:49minister.

About this transcript

This page contains the full transcript of India AI Impact Summit 2026: Session on Generative AI and Future Networks Session by Software Technology Parks of India, generated from the public captions YouTube serves with the video. The transcript has 8,409 words across 1,368 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.