- 6 weeks ago
India Today AI Summit 2026: MIT Prof Raskar says everyone will have AI agent soon
Category
🗞
NewsTranscript
00:00So let's try and make this interactive now.
00:02How many of you have Aadhaar cards?
00:06Raise your hands.
00:07Just about everyone?
00:09Is there anyone missing who doesn't have an Aadhaar card?
00:11Or an ID?
00:13So here is someone who is proposing
00:16an AI for each and every citizen of India.
00:19Please welcome Professor Ramesh Raskar.
00:22He's an associate professor at the MIT Media Lab
00:25and leader of the Camera Culture Group.
00:27He's an Indian origin innovator from Nasik in Maharashtra.
00:31He pioneers computational imaging, AI for health
00:35and now decentralized AI.
00:37And Professor Raskar is wearing a fascinating t-shirt as well.
00:41It's called Doodh.
00:43Let's see it.
00:44Okay, so let's talk about Doodh, Professor Raskar.
00:47Maria, wonderful to be here.
00:49As I said, every citizen has an Aadhaar card.
00:54What if every citizen has an AI agent for themselves?
01:02Sounds very futuristic.
01:03Yes.
01:04But, you know, many of us who can afford,
01:08you know, we usually have somebody to assist us.
01:10But imagine, you know, a 70-year-old Sita Bai in rural Bihar
01:16and she wants to get the same kind of services
01:19that, you know, you and I can afford.
01:22And if Sita Bai can just pick up her phone and say,
01:25you know,
01:26mujhe toh Kumbha Mela jana hai Nasik ko, Nasik mein kya karo?
01:31If you have to go to Kumbha Mela,
01:32you'll just ask somebody to take care of everything.
01:34But imagine for Sita Bai,
01:37an AI agent that's running on her phone
01:39can do everything for her.
01:41You know, her travel, her food,
01:44you know, she's pre-diabetic,
01:46so she wants a certain kind of a homestay
01:48and access to a particular type of vegetarian food and so on.
01:51All those things are taken care of by this AI agent
01:54as long as the folks at Kumbha Mela
01:58and all the vendors there
02:00also have their own AI agent
02:03so they can talk to each other.
02:04Okay, so Sita Bai's agent is talking to the agent
02:08of one of those who are there,
02:11perhaps among the organizers in Kumbha Mela.
02:14That's right.
02:15And how do they connect?
02:16So, you know, you're familiar with Flipkart
02:20or, you know, platforms like that
02:21and ONDC, you know,
02:24is trying to make it more democratic.
02:26They're saying if you're a vendor,
02:27if you're a consumer,
02:28come on this decentralized platform,
02:31anybody can list
02:32and anybody can buy products from them.
02:34So systems like ONDC,
02:36which is part of the DPI,
02:38Digital Public Infrastructure,
02:39is already doing that.
02:41So we just need ONDC to run at Kumbha Mela
02:44so all the vendors can register their AI agents
02:47on this platform
02:49and any one of us,
02:50our consumer or pilgrim agents,
02:52can talk to them.
02:53And so the vendors are saying,
02:55you know,
02:55I was at Prageraj last year
02:56around this time
02:58studying, you know,
02:59how the Kumbha Mela works
03:00and I had a, you know,
03:01it's beautiful infrastructure
03:03but experience and access is challenging.
03:05So, you know,
03:06I had a headache
03:07and I was just trying to look for some,
03:08you know,
03:09some tablets.
03:10But I couldn't find,
03:11among all the vendors,
03:12who it would be.
03:13I would walk, you know,
03:14various places.
03:14So there's no discoverability of resources.
03:18So imagine if the vendors are,
03:20you know,
03:20very excited to register themselves
03:22and their AI agents
03:23don't have to pick up the phone every time
03:24saying,
03:25hey, Ramesh is calling,
03:26he's looking for,
03:26you know,
03:27headache medicine.
03:28But I just,
03:29my agents say,
03:32and it's kind of,
03:33you know,
03:34polling hundreds of vendors,
03:37hundreds of pharmacies
03:39that might be selling that.
03:40So I don't,
03:41both parties don't have to continuously
03:42be on the phone
03:43or on WhatsApp.
03:44These agents are doing everything for them.
03:46You know,
03:46you have said,
03:47and I'm going to quote you
03:48from one of your recent views,
03:51that future will not be controlled
03:53by a single system
03:54but will be created
03:56by millions of empowered agents
03:58working together.
03:59This is essentially
04:00what India believes in
04:02in terms of decentralized AI.
04:05Will Dood figure in here?
04:07Absolutely.
04:08I mean,
04:08if you remember before UPI,
04:11first of all,
04:12you know,
04:12financial access was challenging.
04:14But if you did,
04:15you had to use a credit card
04:16by a company like Visa.
04:18And,
04:19you know,
04:19it was highly centralized.
04:20You know,
04:21all the shopkeepers
04:21had to be in the Visa network
04:23and all of us
04:24had to carry a Visa credit card
04:25to do that transaction.
04:27But UPI made it,
04:28you know,
04:29very decentralized
04:29because you can use
04:30any bank
04:31to any vendor,
04:33to any consumer,
04:35can connect
04:35through this decentralized platform.
04:37And think of the AI agents
04:39that are emerging
04:40through platforms
04:42like Dood and others.
04:43It's a similar thing.
04:44It builds on top of
04:46Aadhaar and UPI
04:47and DigiLocker
04:48and ONDC
04:49is just a layer
04:50above that.
04:51So it's not really
04:51that magical
04:52if you think about it.
04:53But there are
04:55very few countries
04:56that can make it happen.
04:58And those countries
04:59need
04:59a very strong
05:01identity
05:01and payment rails.
05:03As far as I know,
05:04there's only one country right now.
05:05And that's us.
05:07That's here.
05:07That's here.
05:08Okay.
05:09Another quote of yours
05:10before you
05:10have a presentation
05:12for us.
05:13We'll be playing that
05:13in just a bit.
05:14But you say
05:16that the current
05:17AI era
05:19is factory phase
05:21and you want
05:23it to have
05:24bazaar phase.
05:25This is again
05:26decentralization
05:27or something
05:28which is being seen
05:29as democratization
05:31of AI.
05:32What will be
05:33bazaar phase here?
05:34And is factory phase
05:36largely dominated
05:38by those
05:39mega tech giants
05:40who are talking
05:41about trillions
05:42of dollars now?
05:43Marie,
05:43you're absolutely right.
05:44I mean,
05:44right now,
05:45we are kind of
05:45the factory phase
05:46of AI.
05:47In fact,
05:47some folks
05:48call it AI factories.
05:49And you need them
05:50in the beginning.
05:51You know,
05:52billions of dollars
05:53of investment,
05:54you know,
05:55centralize all the data,
05:56centralize all the compute,
05:58you know,
05:58you have to build
05:59nuclear power plants
06:00to power them
06:01and so on.
06:01It's like
06:02building a factory.
06:03And you need that
06:04in the beginning.
06:05But as you know,
06:06in every society,
06:07in the beginning
06:08you build factory
06:09and everybody else
06:10is just using
06:11what's built
06:12in that factory.
06:12That's what's happening
06:13in AI right now.
06:14Some big companies
06:15build AI
06:16and we just use it.
06:18But very quickly,
06:19we're going to switch
06:19from the factory era
06:20of AI
06:21to the bazaar era
06:22of AI
06:22where every one of us
06:24can create
06:25our own AI models,
06:26train our own AI models.
06:28There'll be a Maria agent,
06:29there'll be a Ramesh agent
06:30and we'll be talking
06:31to each other.
06:32Our agent will talk
06:33to each other.
06:33So when you go to the bazaar era,
06:35the economics changes
06:36right away
06:37because now our agents
06:39are participating
06:39in the economy
06:40and it's based
06:41on what we do.
06:42So you know,
06:42usually we wake up
06:43in the morning,
06:44we take care of our messages,
06:46take care of our email
06:47and we feel like
06:48we have got some work done
06:49in our digital life.
06:52But, you know,
06:53a year from now,
06:54you'll actually wake up,
06:56look at your agent
06:56and say,
06:57how can I make my agent better?
06:59It's almost like me
07:00talking to my teenage daughter.
07:02And I'll say,
07:03you know,
07:03I'm going to train this agent
07:04till yesterday,
07:05my agent was responding
07:06in this way,
07:07but now I learned
07:08something interesting
07:09from Rajdeep
07:09and from Maria.
07:10I'm going to edit
07:11a little bit more,
07:12add some more thoughts,
07:13also just say that
07:14I'm going to be very busy
07:15in March and April,
07:16but maybe next year
07:17I want to go to Kumbh Mela.
07:18I can kind of modify
07:19my agent
07:20and my preferences
07:20and my knowledge
07:21and my expertise
07:23and then launch it again,
07:24right?
07:25And that takes care
07:25of the majority of my life.
07:27But there's also economy
07:28around it.
07:29You know,
07:29I might say,
07:30hey,
07:30I'm a professor at MIT
07:32and as you can imagine
07:33the most common question I get,
07:35hey,
07:35my son is in high school,
07:36how can he get into MIT?
07:37Right?
07:38So I could just answer
07:39all those questions.
07:40My agents can just answer
07:41those questions.
07:42But also,
07:43you know,
07:43joking aside,
07:44also some serious questions
07:45about my research
07:46or my companies
07:47and so on.
07:48So I might start charging,
07:49I don't know,
07:50100 rupees per quiz
07:50over time.
07:52But at some point,
07:53you know,
07:53your agent might be better
07:54than my agent
07:55and your agent is charging
07:56200 rupees per query.
07:57I was like,
07:57why is Maria's agent
07:58so much better?
07:59So I'm going to send
08:00my own AI agent
08:01to the IIT
08:02of agent schools
08:04and IITs
08:05will run schools
08:06for agents.
08:07My agents
08:08will become smarter
08:09over time
08:09and it will start
08:10competing with your agent
08:11now for,
08:12you know,
08:12for business
08:13and so on.
08:14At some point,
08:15you know,
08:15my agent will get rusty.
08:16I might have to send
08:17my agent to a repair shop.
08:18So there will be
08:18agent repair shops.
08:20But, you know,
08:20at some time,
08:21my agent will also
08:21start doing mistakes.
08:23So I might have to buy
08:23an insurance
08:24for my agent
08:25and so on.
08:26So you could imagine
08:27there's a whole
08:27Asian economics,
08:29a whole Asian ecology
08:31that's going to emerge
08:31and we can't even imagine
08:34all the new businesses,
08:36all the new
08:36socioeconomic structures
08:37that we'll have
08:39in the next few years.
08:40Okay.
08:40Since you are saying
08:42that the impact of AI
08:43will be beyond technology,
08:46it will be in the domain
08:48of welfarism,
08:50happiness,
08:50and opportunities,
08:51then what does
08:53human interface mean then?
08:55Yeah.
08:56I think this is the challenge,
08:58right?
08:58Like,
08:59who's here
08:59really excited about AI?
09:02And who's really worried
09:03about what AI can do
09:04for jobs
09:05and misinformation
09:06and fakes and so on,
09:07right?
09:08So as you can see,
09:08there's a lot of fear
09:10and excitement about AI
09:11and that's because
09:12we're in the factory
09:13mode of AI.
09:14We have no agency,
09:15we have no control
09:15over how AI
09:17is going to get delivered
09:17to us.
09:18But when we switch it
09:19and every one of us
09:21has our own AI,
09:22our own micro AI,
09:23our own AI agent,
09:24we can participate,
09:25you know,
09:26as first class citizens
09:26into this.
09:28And that's going to bring
09:29a lot of safety,
09:29lots of checks and balances.
09:30It's more democratic.
09:32Just like in democracy,
09:34you know,
09:34even if an individual
09:35seems powerless
09:37in the beginning,
09:38because of a democratic process,
09:39even the most powerful
09:41have to listen
09:41to every individual.
09:43It's the same thing
09:44when you democratize AI.
09:46When every one of us
09:47has the ability
09:48to participate
09:49in the AI ecosystem,
09:52suddenly there is
09:53a bit of equalization there.
09:54And that's how
09:55the socioeconomic aspects
09:57are going to change.
09:58And, you know,
09:59people might be worried
10:00about their jobs
10:02and so on.
10:02That's because
10:02we're in the factory era.
10:04But in the bazaar era,
10:05there will be
10:06so many new opportunities.
10:08All of us will be busy
10:09creating new businesses,
10:10as I said.
10:11And how soon
10:12is the bazaar area?
10:13That's a good question.
10:14I mean,
10:14if somebody reminds you
10:16that Chargipity
10:17is just about three years old,
10:19you know,
10:19how much has changed?
10:20We take it for granted now.
10:22You know,
10:22how many people
10:23are doing their homeworks
10:24with Chargipity?
10:25You don't want to raise your hands.
10:27How much has it changed?
10:28So you can only imagine
10:29how quickly.
10:30Is it going to take
10:31ten years,
10:32five years,
10:32maybe one year?
10:33And that's why
10:34the next year's Kumbha Mela,
10:36which is happening in Nasik,
10:37is, you know,
10:38our first flagship example
10:40of having,
10:41you know,
10:42millions of people
10:43having their own
10:44AI agent
10:45as a pilgrim.
10:46Okay, so
10:47what do you have for us?
10:48Let's take a look
10:49at the presentation now.
10:54Yes, right here.
10:56So first of all,
10:57Doot,
10:58which as you know
10:59means a messenger
11:00or an envoy,
11:01not your master,
11:03is a personal AI agent
11:06for every citizen,
11:08right?
11:09Just like there's
11:09Aadhaar and UPI
11:10and so on.
11:12And, you know,
11:12Doot is not trying
11:13to control you.
11:14It's your messenger.
11:15It's not taking decisions
11:16on your behalf.
11:17And the way
11:18we are doing this
11:19is a technology
11:21called NANDA.
11:22Okay?
11:23Right.
11:23And NANDA stands
11:24for Networked AI Agents
11:26in Decentralized Architecture,
11:28which is a mouthful,
11:30but NANDA means joy
11:32and also name of my sister
11:34who lives in Nasik.
11:37So the NANDA technology
11:39is what's really
11:40playing a role.
11:41And as I said,
11:43if you think about
11:50what's the internet
11:51of AI agents,
11:53you know,
11:53we are kind of
11:54in the mainframe era
11:55of AI,
11:55where things get built
11:56in big factories,
11:58switch to PCs,
11:59then the intranet,
12:01then the internet
12:01of AI agents,
12:03but the internet
12:04was about connecting machines,
12:06the World Wide Web
12:07was about connecting people,
12:09and NANDA
12:10is about connecting agents
12:12that are representing people.
12:13Right?
12:14So that's the key here.
12:15So NVIDIA speaks
12:17about factory phase,
12:18you are speaking
12:19about the bazaar phase.
12:20That's right.
12:21And the bazaar phase
12:22comes how many years
12:24after the factory phase?
12:25I mean,
12:25it's already happening.
12:26It's already happening.
12:27What you may not realize
12:27is that the reason
12:29why we have to win
12:30the factory phase
12:30is this models
12:32have one trillion parameters.
12:34They cost billions
12:35of dollars to build
12:36and so on,
12:38but now the models
12:39are becoming so small
12:40they'll run on your phone.
12:41It'll take less power
12:42and less compute
12:43than watching
12:45a YouTube video.
12:46So imagine, Maria,
12:47if, you know,
12:48all your information,
12:49all your degree,
12:49all your health,
12:50all your travel,
12:51which is maybe
12:52a couple of megabytes
12:52of information,
12:53is running on your phone
12:55and doing the work
12:56for you all day,
12:57all night,
12:57like having your own assistant.
12:59So that's where
13:00the bazaar phase
13:00is already happening.
13:01But then what's needed
13:03is the network aspect of it.
13:05How do you build a network?
13:06Yes.
13:07Because, you know,
13:07it wasn't enough
13:08that everybody has cash.
13:10What we needed
13:11was something like UPI,
13:13so there's networks.
13:14So anybody can talk
13:14to anybody
13:15and, you know,
13:16buy insurance
13:16or buy a new phone
13:17or, you know,
13:18get credit instantaneously.
13:20So the network
13:21is more important
13:22than the AI.
13:23So do I have to instruct
13:25my agent
13:26to speak to,
13:29say,
13:30the agent
13:30who is sitting somewhere
13:32and then they communicate?
13:34Or does that happen
13:35because it picks up
13:36from my mention
13:38during one of the conversations
13:40that I may have
13:41with intent
13:42that I want to do that?
13:43That's a good question.
13:45Maybe in the beginning,
13:45it's like raising your child.
13:47You know,
13:48in the beginning,
13:48we handhold,
13:50we are very prescriptive
13:51and as they start,
13:52you know,
13:52raising,
13:53it's like,
13:53you know,
13:53now you're independent.
13:54By the time they're 18,
13:55they go to college.
13:56We're like,
13:56beta,
13:57now you're on your own.
13:58Right?
13:58And the same thing
13:59with agents.
13:59In the beginning,
14:00we'll be very controlling,
14:02very prescriptive,
14:03but at some point,
14:04the agent will be so good,
14:05Maria,
14:05it's like,
14:06you know what,
14:06just go ahead and do it.
14:07I'll be fine with it.
14:08Okay,
14:09so let's move on
14:09to the next slide.
14:12So we don't have to go
14:12through all the slides.
14:13Yes.
14:14But I think I just want
14:15to get this idea across
14:16that in the beginning,
14:18we have to build
14:19a very solid foundation.
14:20You know,
14:21just as you move
14:22to a new city,
14:23move to a new country,
14:24you need a passport,
14:25you need an address,
14:27you need your Aadhaar card.
14:28You know,
14:28every agent also needs
14:30to have their own Aadhaar card.
14:31Right?
14:32Every agent needs
14:33to have their own bank wallet
14:34and so on.
14:35But very soon,
14:36you know,
14:37agents will start
14:38helping in our jobs,
14:40they'll start making money
14:41on our behalf,
14:42but agents will also
14:43create their own stock markets.
14:44They'll also create
14:44their own banks.
14:46Unfortunately,
14:46they'll also create
14:47their own casinos.
14:48So there'll be some good stuff,
14:49there'll be bad stuff as well.
14:50But soon after that,
14:52you know,
14:52when we move to a new city,
14:53it's not just about
14:54settling down
14:55and finding a job.
14:56We start creating clubs.
14:57We start hosting events
14:59like this.
14:59We create universities.
15:00We create justice system.
15:02We create sports teams.
15:03And agents are going
15:04to do the same thing.
15:05They're going to create
15:06their own justice system
15:07because what if my agent
15:09didn't do the right thing
15:10that Maria agent
15:10asked you to do?
15:11So you're going to sue my agent.
15:13So we need our own
15:14justice system
15:14for it as well.
15:15So that justice system,
15:17again,
15:17will be agent-like.
15:19The agents are just
15:20interacting with the agents.
15:21There'll be agent judge
15:21and agent clerk.
15:23And they're all interacting
15:23with each other.
15:24Every once in a while,
15:25we might have to go
15:25and intervene
15:26to see if everything
15:27is going fine.
15:28But most of the time,
15:29they're just talking
15:29to each other.
15:30It's like high-frequency
15:30trading today
15:31on stock market.
15:33You know,
15:33somebody's writing algorithms
15:34and stocks are being traded.
15:36You know,
15:36the fund managers
15:37barely take a look at it
15:38just to make sure
15:39things are right.
15:40But the trades are happening
15:41very, very fast.
15:42Maybe that's the closest analogy
15:44to how agents
15:45will be doing this fast trading
15:46with each other,
15:47not just for commercial purposes,
15:48but for entertainment,
15:50for social,
15:51for justice,
15:51and everything else.
15:52So if there is a jury
15:53also for agents,
15:55then are we then heading
15:57to a situation
15:58then perhaps in future
15:59summits of this nature
16:01will have agents of ours.
16:03We may not be there
16:04in the room at all.
16:06No, no.
16:07The whole point is
16:08the more the technology
16:10there is,
16:11the more time we'll have
16:13to spend with each other
16:14in the real world.
16:15You know,
16:16just because a computer
16:17can't be chess
16:18didn't mean we stopped
16:19playing chess.
16:20Just because there's
16:21a bicycle or a car
16:22that runs faster than us
16:23doesn't mean we stopped
16:24running or exercising
16:26or running Olympics
16:28for running.
16:29Right?
16:29So I think the more
16:30technology there is,
16:31the more human
16:32it becomes our
16:33day-to-day activities.
16:34So I think we'll be
16:35spending more time,
16:36like just think about
16:36running this conference,
16:37how much your team
16:38spent time doing
16:39all the stuff
16:40that happens behind us
16:41since I know
16:41how amazing your team
16:43is trying to coordinate
16:43for me to be here,
16:44but imagine if all of that
16:46is happening,
16:46you know,
16:46semi-automatically
16:47and then we can come here
16:49and just focus on,
16:50you know,
16:50what we really enjoy
16:51which is exchanging ideas
16:52and thinking about the future.
16:54Okay,
16:54so your final thought
16:55and this is my question,
16:57what's your most optimistic view
16:59of AI reshaping
17:02India's economy
17:03and society
17:04say by
17:052027 and beyond?
17:07So let's start
17:08with the potential
17:10dystopian world.
17:11The potential dystopian world
17:12is that
17:12we remain in the
17:13AI factory era
17:14and,
17:15you know,
17:16they build
17:17the most amazing
17:18math teacher for the world,
17:20you know,
17:20the most amazing
17:21mental health worker
17:22for the world
17:23and they'll be
17:23exactly math teacher
17:25for the whole world,
17:26you know,
17:26exactly one accountant
17:27for the whole world
17:28and everybody else,
17:30all of us,
17:31will just be
17:31intelligent slaves
17:32given by this master AI.
17:35That will be
17:35pretty soulless life.
17:36We won't lose jobs
17:37but jobs of most of the people
17:40will be just using
17:41these AI models
17:41given to them.
17:42That's the dystopian world
17:43we are walking into right now
17:45but once we move
17:45to the bazaar phase,
17:47what's going to happen is
17:47the math teacher
17:49is training their own
17:49AI agents,
17:50the students are training
17:52their own AI agents,
17:52they're talking to each other
17:53about many interesting things,
17:55you know,
17:56if you're an artist,
17:57if you're a creator,
17:58you know,
17:58if you're a leader,
17:59you know,
17:59you're,
18:00so it becomes
18:01a creator economy,
18:02you know,
18:03we're all going
18:04to become artists,
18:05we're all very proud
18:06of an agent we're creating
18:07because our agents
18:08are representing us
18:10in this parallel economy.
18:11So I think the,
18:13you know,
18:13the glorious decade ahead
18:15is that
18:16it's a beautiful world
18:17where all of us
18:18are training our agents,
18:20we spend a lot
18:20of physical time
18:21things that we
18:22enjoy doing
18:23and many of the
18:25pesky situations
18:26that are out in the world
18:26like health
18:27and education
18:28and travel,
18:29livelihood
18:30and insurance
18:31and climate,
18:33all these issues
18:34are being solved
18:35in a democratized
18:36Asian-thick world.
18:37That's my hope.
18:38All right.
18:39On that positive note,
18:41or should I say
18:42dood for all,
18:43so maybe
18:44in some years
18:45from now
18:46all of you
18:47and all of us
18:48will also have
18:49an AI agent
18:51for every citizen.
18:52That's a great idea
18:53certainly.
18:54Thank you so much
18:54for joining us
18:55and thank you
18:56Professor Raskar
18:56for this conversation.
Comments