B (2:51)

That's right.

A (2:53)

So a large number of students end up going to small institutions which we call colleges. We call them colleges. They're not universities. They are affiliated to universities. That's how India works in higher education. The big name, well reputed group called good institutions are few in number. And secondly, because they are fewer number, they are either fiercely competitive for getting admission or they are very expensive. Now once it becomes fiercely competitive, that is also an expense. You have a lot of money to get there. So today, any education today in India, if you want a good reputed college to give you an undergraduate degree, it's really very expensive. So in our terms you would say at least 25, 30 lakhs is what you spend. Lakh again, you keep converting. I'm talking in rupees again. So that's pretty high for India. So. And a lot of interesting other things are there. Even though IITs take very few undergraduate students, they are very heavily funded by the Ministry. Ministry funds the IITs very heavily. So all this was okay, it was going on for quite a while. But then we reached different stages of economic growth in the 90s. We boom through it and all that. And a lot of colleges, a lot more colleges came up for engineering particularly. And then now I think there is the second sort of stage where maybe AI and Data Science and ML are going to dominate. And we need many more engineers who know how to deploy these systems, understand these systems, etc. So that seems to be the next stage that we are at. So maybe at this point there is an opportunity for the IITs to deliver a program that maybe is not so expensive. When I say expensive both in terms of the competition, the intense competition that students go through, the intense competition itself, if you look at Kota for example, and even other places, it's become very inhuman in some sense. Yes, very competitive. Students losing their childhood and then just, just for getting into college and then your brain is not developed properly and what do you do? What do you do after college? Right, that's just, that's a big question mark. So we really wanted to break this triangle hold we have because of numbers and resources and limitations. And can we do a full fledged UG degree for under 5 lakhs? That's like under $5,000. Can we do a good, highly reputed, good quality UG degree? Students who come out are really skillful, employable, they're going to go out and make a name for themselves, put the IIT brand behind it. Right. So you can do a very low cost degree, not put a heavy brand. So can you put the IIT brand behind it and still deliver a $5,000 less than 5 lakh, let's say, right degree in India? Is that possible or not? Was a great question that came up four, five, six years ago. And a bunch of us is me that I'm talking about, a group of us sat down and thought about it and debated it. And I think today we have created something like that. And after five years I can say that and give you more numbers as we go along. But after five years, looking back, looking at all the numbers, looking at where the students are coming out, I think we've really created a low cost, highly scalable, very high quality, rigorous undergraduate degree program which I think is the need of doctor in the country.

B (6:23)

Yeah, I saw some advertising a similar program at IIT Jagpur. Is this idea spreading now?

A (6:33)

Yeah. So Guwahati has a program like this. Jodhpur is starting a program like this, Pakna starting a program like this. But I think it's still early days in the sense that I think IIT Madras, because of the pioneering nature of many other things we have done in online education that earlier we had a bit of an advantage and I think we know a lot of secrets and a lot of things about how this works, how this model works. And I think our model is working really well today. The other institutes are learning and I think if they are interested, they may do. I'm really still waiting for some of the older, more famous IITs to act in this regard. So maybe if they do and if and when they do, that would be a really huge next.

B (7:12)

Yeah, well, ultimately, given that it's online, it's not geographically fixed, it could be a combination of the top IITs.

A (7:22)

So given our nature, like I mentioned, because of our earlier experience in online education, in our program we have teachers from all across the country. I see many other IITs, the faculty from other IITs teach in this program. So they contribute in so many interesting ways. So that that way it's become a very unique and very interesting program.

B (7:41)

Yeah, two things struck me in my interview so far, one was that a significant number of IT graduates are unemployed. I'm not sure why. That's what I was told. That's last year.

A (7:58)

That is entirely true I think. I don't, I won't say unemployed. I don't think. See if the way we put out our fraction of students who actually got a job and all that is a little complicated, you should really understand those numbers. Well, but I wouldn't say a large fraction of IIT graduates are going out. Definitely not maybe through the regular placement on campus. There is a smaller number that's there and I think there is a small number that still ends up looking for a job when they're about to graduate. And that is probably true but I don't think at least in the older IITs the, the number is that large. Alarming or something like that.

B (8:37)

Yeah. And then the other thing that struck me was in the AI mission Mr. Singh was talking about, there remains a huge gap in the number of qualified and now engineers, whatever you want to call them and the projected need to execute the plans that they have. So I imagine is that in your mind as well?

A (9:04)

Yeah, definitely, I think. And, but, but I think if you look at it, look at that point also a bit carefully. Right. So I think the number of CS based degree programs around computer science, the number of degree programs they have today has skyrocketed. Almost every college that I spoke about today the engineering colleges pretty much teach only computer science and mechanical civil and all this of died down onto very small number. So the number of people with computer science oriented ML or AI or yes degrees is increased and. But then the difficulty in the Indian education system is it's. We're not somehow very skill focused from, from school also there is like I think maybe it's a societal thing. I don't know. Somebody has to study this more closely. But I think generally the. There is this knowledge and other things are valued and people who work with their hands get things done. It's maybe a societal tendency from long ago. It's like we don't look at that as respectfully as we would look at us look at someone who's sitting down and thinking. Right. So I think so. So. So our education system is geared towards knowledge and not so much towards skills, the hands on skills and that shows quite often. So you'll have a lot of people graduating with BTech in Computer Science and you'll be shocked to see that they can't get a piece of code to run on a computer which you would think is like the essential basic skill. They may write the code up in a piece of paper and give it, then what good is that? Right? So I think this is a recurring theme across all the entire Indian educations. So maybe the people in the Indian AI mission, really what they mean is we may have a lot of graduates, but employable graduates who can be deployed into projects and can execute on the ground, that is a big issue. So in our program, for instance, we have done something very different. In fact, I can confidently say very few degree programs in the country do that. So we do a foundation part which happens in the beginning, about a year or so, eight months to a year. And once students finish that, we do a diploma part. We call it explicitly a diploma part. And it is focused completely on skills. They earn quite a few credits in that part and it's very demanding. There are four projects that they have to execute hands on, build two apps, show that the app runs, build an ML model, show that they understand what it is and do that. Then one more ML oriented. So two ML projects, two application development projects. They have to do hands on in demonstration. So we put that as part of it first. Then after that we learn the theory in the later levels a little bit more in detail. So. So we've done a good combination of foundations and skills and focused a lot on skills. We want to have skill based outcomes as essential part of this degree product. So that is one another innovation that we did which is not talked about so much, but it's I think an important part for why, why this program is, I think, likely to succeed in the future.

B (12:03)

Yeah. And when did the program start?

A (12:06)

First batch of admission was in 21J. And the conceptualization, all of that was late 2019. Formal approval was 2020 in the IITM Senate. First, you have to understand for an IIT, IIT to offer a degree, undergraduate degree without GE, it's like, what is a good comparison? I think I had one faculty tell me, you know, Stanford will not do community college.

B (12:29)

Right.

A (12:29)

So I think it was, it's like, I don't know how to describe it. I think that it is a very significant move for them to do something different and that is responding to the societal me in a positive way. And all of that. There is this big ivory tower that people have built and put the IITs and talk of. So. So we had to really talk to a lot of people. So that took a lot of time in convincing everybody that give us a shot. At this point, we'll show you how to run a program like this, which doesn't, doesn't have a sharp filter at the input, but it will have a funnel and the funnel will operate. I, I'll show you how the funnel operates. We admit more people keep it fair as long as they spend enough time, pick up the skills, let them funnel through. Not everybody's going to make it. Let's have the, let's have the strength of the exit. Let's have a strong filter at the exit and not at the entrance. And I think today we would see that all the numbers show that they show that this is really hard.

B (13:21)

Yeah. How long is the program?

A (13:23)

Oh, that's a tough question to answer. So nominally four years. The number of credits as a four year program, the entire BS degree. But we have a lot of intermediate exits. I spoke about the diploma exit in the middle. Right. So that's particularly important because when you keep, when you want the financially less privileged folk to come in and do a program like this, they can't put away five years of their life and then study and then go to. Doesn't work for them. It's just this period. Even, even if we give a 5,000 degree, it doesn't work. Wasn't as serious issues at home.

B (13:55)

Sure.

A (13:55)

So feed somebody. It doesn't work at all. So they have to start working. So that's why the skills first is very important. So at the end of the diploma, many of our students get internships, small jobs, some coding, maybe back end work, not so great, but still pays them some and then that helps them survive and go. So there is a diploma exit, there's also a three year degree exit. Then there's a four year degree exit. Nominally four years but what we're seeing on average the foundation is about one year. Students are taking on average one year. The diploma, which we thought would be two years, is actually going to nearly three years, two and a half to three years. So students are struggling. Like I said, the Indian system is not focused on skills. We don't push the skills part too much and the diploma tests the skills so it does the exact same thing. It's harder for people to do so they take longer projects, are incredibly strong finances. They really funnel up those who can sit down and do stuff with their hands. So I think it's.

B (14:52)

That's important.

A (14:53)

And then after that, once they finish the diploma, they find the degree level smooth sailing. They finish it off in a year. So the faster students, really bright ones are finishing it in three and a half to four years. And the more Average level seems to be four and a half. Four and a half.

B (15:10)

Right. But it's. There's no cut offs like no, we

A (15:14)

don't cut them off. We cut them off at eight years. So people are going to graduate. We had more than thousand graduates at the base level already in the first few batches. And every term we're expecting about 500, 300, 400, 500 types. It's increasing slowly.

B (15:29)

Yeah.

A (15:30)

And I think we will graduate about thousand thousand five hundred people in the four year degree every year.

B (15:35)

Yeah. And the prerequisites for coming into the program wouldn't those.

A (15:40)

So those also we thought about capsule once again a focus was we want to reach the financially less privileged kids who didn't go to great schools. See that's I think Even in the U.S. that's true. If you are in the city, you go to a great school district and you see and then everything is on offer if you go to one. I mean even in Atlanta that was true. I know from experience go to a urban place like Atlanta and go to a normal government public school there, many things are not there. You know, you're not taught and I think then you don't get AP classes, you don't get any of these things. And this is true in the US and doubly true in India. You don't get everything you want if you don't go to a grad school. So we wanted to make sure that we can reach anyone. That's number. The second thing is given the importance of data for society going forward, we cannot have only engineers do data science. We need lawyers, we need doctors, we need humanities folks, we need everyone doing data. So we wanted to reach that crowd. So we decided we will do a strong foundation and assume what only math and other things at what we call the 10th standard levels like the sofomo year of school. And after that students tend to specialize a little bit in school here in India. So after 10 standards. So till 10 standard it's common everywhere. And we said we will assume only that the what we need after that small data science. We will develop a span of a program. So we're not relying on the fact that kids went to grade schools and all that or they studied physics, chemistry or something. In 12th standard we don't rely on any of that. They can study whatever they want. We will build it here when they come in and that's how they will.

B (17:13)

Yeah. And the so right now at any one time how many students are in a program?

A (17:22)

So right now I think There are more than 40,000 active students across various levels. But you have to understand, I think there's a funnel operating. Sure, funnel. I'm giving you the total count. It doesn't mean like 40,000 will graduate. Yeah.

B (17:35)

And you were saying you graduate each

A (17:36)

year, what did you say it looks like? Best projections right now is about thousand to fifteen hundred a year in the four year bsc. But in the diploma level we'll be much larger. It'll be in the few thousands. Few thousands will graduate in the diploma level, but in the four year degree level, probably thousands to thousands.

B (17:53)

Yeah, I've taken, I took Andrew's Coursera course and it relies heavily on TAs of teachers, assistants to interact with the students. How do you scale that here?

A (18:07)

So that's an important part of our program. Right. So we have faculty at IITs are incredibly busy. In fact, IITs end up attracting very good quality faculty. So they end up having large research groups and they contribute a lot in so many different process. So they can't be handling large cases, not at this level, not online and all. So what we do, which is similar to what many US universities do. So a lot of US universities have research, tenure track faculty and then they have teaching. So we have hired people whom we call instructors. These are our full time staff. Many of them are PhDs, masters, graduates from IITs and other institutes, very top institutes. Maybe they didn't want to go into industry, they want to be in teaching, but they can't. Maybe they're not that great that they could get into tenure track research positions. So then what can they do, you know, if they want to be a good teacher? So we found so many good people like that and we've hired them as instructors. Instructors actually interact with students and help translate the content in a way that students can understand, sit down, solve problems with them. The foundation level for every course. There is eight hours of live sessions every week that these instructors do so students can come in, ask questions, get their doubts clarified. The numbers are large, maybe your voice is not heard, but you can, you know, paid eventually. Eight hours is a long time to come in. Sure.

B (19:28)

When you say live, do you mean online or.

A (19:31)

So the recorded lectures are done by the faculty and then the instructors do it online, like

B (19:39)

for scaling class. Because as long as you could find enough instructors to have all those students, you could scale.

A (19:48)

Yeah, yeah. So I think there's lots of interesting dynamics here. The first thing is students in an online program. We imagine they're isolated and they're sitting in front of the Computer and all that. So we wanted to break that also a little. So we do a lot of student government activity, even though it's an online program. There are students are put into houses. We have 12 houses named after the forests of the country. And then the houses have their own internal structure. They're elected leaders and all that. And then there are clubs and societies and the students form. And these clubs and societies organize meetups in different parts. Country students in the nearby areas come and meet up there. And every year we do an on campus event which is called Paradox Paradox, because it's an online program that has a campus event and the students run that program. So it's a week's program. Students, if the program from all across the country coming to campus because it's the summer and students are not here.

B (20:44)

Sure.

A (20:44)

So they come into campus, they do sports events, they do cultural events, they do technical events. And that's the place where they meet each other and then they go back and they stay in touch. And then there are student groups, all sorts of interesting student groups that get formed and then they talk to each other. They help each other study also. So that dynamic is. Is something that we actively seeded in some way, but we can't control every aspect of it because that's something that they have to do on their own. And that's very vibrant and alive. And I would say that's a crucial, critical. It's not like every instruction is coming from us and that doesn't work. Anyone who has done teaching will know that this belief that what I'm saying is going directly into their hearts and mind just doesn't work. We have to go outside of the class, talk to each other, interact, debate and discuss. That's when things get solidified. And that's happening. That's happening in a very interesting way. Student driven way. That's also very important.

B (21:42)

Yeah, yeah. Georgia Tech has, I think, the first online master's. Little interesting aside, I applied and was rejected.

A (21:52)

Okay.

B (21:53)

I was shocked. I've never been rejected from anything, but I didn't have the necessary computer science background. But have you. Do any of the students go on to that program or other online master's program?

A (22:10)

Not necessarily online. Many of them even go to campuses for master's programs. So. So last. Last year, in fact, no, this year there is a admission for the master's programs at IITs and IAC, you know, top institutes. It's through an example gate. And this is a national level exam. People all across the country, right this last year, out of the top 10 ranks. 3 ranks by students on sport. Yeah. So and what the first ranker, All India rank one as we call him, he's actually a doctor, so he did MBBS from ames, which is the top medical college in the country. After two or three years he figured out he doesn't want to be in service, he doesn't want to do doctor stuff. So he wanted to go back to tech. But then he didn't have a way back. It's not easy in India to do these things. And he took this program, he did so well in it. CG is so I believe. Then he wrote this GATE exam, he topped it. Now he's doing M Tech in AI and Indian Institute of Science. So his career and lives have changed and I think something similar is happens. Those are the sort of amazingly interesting stories. The student stories of this program are just unbelievable. That there's a filmmaker in Kolkata. He was actually a filmmaker and he's a young kid only this filmmaker, he did our program. If initially he didn't get a high cgpa, but he understood enough, he made a lot of friends. Now he runs a company which does AI production and all this stuff for movies and he runs a company, it's sitting right here that is doing all that. So I can tell you stories like this about these students and I think that's also a part of it. So coming back to your question. Yes, many students are doing PG programs in India and abroad. Many have gone to the us, uk, Germany, all of them.

B (23:53)

Yeah. With this doctor or the. You said the filmmakers were young, but do you have older students coming back mid career?

A (24:01)

Yes. There's no age. The oldest student we had for a while was 80 plus. Actually, I don't know, it was really hard to support him, but I think, I don't think he finished it. But quite a few in the 60s and 30s and 40s also is quite a few. But of course the large majority is under 30. But 15% of the people in the program are actually working in companies. So that's a pretty large number as well. So that's another interesting dynamic. So these student groups I spoke about invariably has one who's 31 member who's 30 years or 40 years old. Then there are others who are younger and that person is working in a company and that group and the dynamic it has, I think has no equal anywhere else. Even if you go to a physical university, you're not going to find a peer group which has an older person and a younger person. And then there is you say something and that person says, you know what, that's not how life works. Eventually things change after you. And all these conversations are also happening. So that is also a very interesting group, by the way. Yeah.

B (25:02)

And what about international?

A (25:05)

So that number is not very large. I think about 350 or close to 400 students are outside of India. A large group among them are Indian experts from India who are living abroad in various space. A small number is actually a foreign citizen outside the country. I think the US is maybe around 25 or 30. I'm not sure what the number is. Not very high. It's more under 50.

B (25:33)

And how are those people discovering you? And also international. So yeah, 30 or so from the U.S. do you have Chinese students? Do you have Japanese students? Do you have Thai students?

A (25:45)

I don't know. I don't know the spread. I don't think it's that spread. International is really very small. Like I said, large majority is in the Gulf. Yeah, Gulf and Singapore is a lot of Indian folk there. And that's how it is. And we don't reach anybody else ordinary with our publicity efforts. We don't do serious publicity as such. Definitely not international. We are not focused on that.

B (26:07)

Yeah. And in terms of growing, are you at capacity or.

A (26:13)

Ah, that's a good question. I think we get about every year these days we get about 40 to 45,000 applications. And out of the applications there is a small process by which people get in and then all of that. I think those applications can grow up to 100,000 and we'll be okay. I don't think that's going to worry us too much after that. Beyond that, maybe we'll reach some sort of a limit and I have to have to wait and see. But I think we really have a team. When you see what we have here, you can see there are some people in the team you can meet some incredibly capable people we've managed to attract into the program for various reasons. I think there's a social cost that pulls people in. It's okay monetarily also we compensate them quite well. But then there is this overwhelming pool that you can make a big difference to society as a whole that brings in a lot of fantastic people. And those people have no limits, can tell you they're not afraid. Many of them are women. They have no fear at all. Academics like me, you'll be afraid of these numbers in projects on 1 lakh and all that. And they'll stand up and say, just show us the numbers. We don't we don't care. Yeah, seems to be.

B (27:19)

Yeah. At the beginning I was talking about how you came back ultimately to teach and that there's a dire need for India's AI advancement and the diaspora come back after the PhDs. Was there a specific reason or was it just personal preference for me.

A (27:43)

Okay. So see my family, almost my entire family is in India. Nobody has gone abroad there. In fact, my parents didn't go to college, me and my sister. So little bit more, much more tied to India than anything else. Well, that's not necessarily the reason. But for me, coming back and teaching in India was always a good income and undergraduate basic was clearly there. So for me it was an easy decision. I think quite a few people come back. It's not that they don't come back, but it's a complex issue right at that time and that age to decide whether you want to be here. So it's a complicated decision to make and I think, who knows, I mean today the world is a complicated place as well with the politics and, and how people are reacting to the bride is also changing.

B (28:29)

Yeah.

A (28:29)

So who knows, maybe there may be a few more years where you want to come back. We'll see.

B (28:33)

Yeah. The other thing just generally AI, India's AI mission to ultimately, according to Mr. Am, catch up with the US in China. I could have interesting analysis from Professor Maussen at IIT Delhi. He was talking about China because China's higher education system was completely destroyed in the 60s. I started going to China in 1980 and it was I think 79 is when they reopened the universities and the first people were going back. So their higher education system was incredibly weak throughout the 80s and 90s. And then the government at a certain point had the resources to start investing in research and attracting talent back. The Chinese national assume moved abroad, which is fine, but they were so far behind everybody in AI certainly. And then generate AI happened well first neural networks, but then generative AI in quick succession 2012, 2017. And by then they had enough of a population of qualified researchers that they could jump on that and leapfrog them.

A (30:05)

And

B (30:07)

the time that they had lost and now they're competitor the United States. Do you think India has the chance for that kind of, you know, if you can build up your.

A (30:23)

The human.

B (30:24)

Well, you have the human resources, if you can invest in the infrastructure and as AI is inevitably going to go through another breakthrough and you'll be at the top of the game.

A (30:40)

I think it's possible, at least from my point of view. At the undergraduate level we are producing good people. Now definitely the numbers are much larger. But of course for something like Deep Seq to happen, I think, I think yesterday we were having a debate in our department about this and somebody was saying that Deep Seq there was a chance that it could have happened even in India. Right. So I'm not sure if I've spoken to Mitesh Kafra who's here. He was named and times AI list the top hundred influencers, etc. So he runs a pretty big AI mission here in IIT Madras, this company called Sarvam which is trying to build a foundation model or something like that. And it's just a, it's, it's a sequence of random things you have to tie down together to have a successful company and a product on there. Right. I think we're very close. I think there's lots of work that's happening. Will it all line up and to result in that even the next big thing happens, we'll quickly get there. I think it's very possible. So let's. I'm very hopeful. I think we have the right ecosystem at this point. But I think comparing India with China is a bit complicated. And you said you know China very well and I think way money flows, way investment flows there is very different from how money flows here. Our regulations, codes and structures and weights we have and all of these enter the picture.

B (32:09)

And

A (32:12)

having said all of that, I'm still hopeful the next big breakthrough that happens, we'll have a product over there which we compete with. Yeah.

B (32:20)

And a lot of it depends right now on building a data center capacity because that means that. Right?

A (32:28)

Yeah, yeah. So I think those things happen. That's where the trick is. Right. And will we put up a nuclear power station, power couple of data centers very quickly and very quickly is very hard in India. You try anywhere in India. Now if you start the discussion of having a nuclear power plant, this could be huge protests and.

B (32:47)

Yeah.

A (32:47)

And it's very hard to convince people either way that silver battle work and all of that. So that's one issue. I don't know how that's going to get solved. Maybe we'll solve it. Maybe we'll use solar, maybe we'll use something else more interesting that needs to come up. And then who's going to invest in that data center? I think is it going to be the Googles and the open AIs of the world or it's going to be an Indian grown company? It's going to do that. Is it Is are we going to be able to act together in one unified way? Is there some tipping point? Let's wait and see.

B (33:20)

Yeah. Well, so, but you're an optimist about this India AI mission, how that's being worked. Yeah.

A (33:29)

Reasonably optimistic about it. Let's see.

B (33:32)

Yeah. And where do you see more broadly, not talking about undergraduate, but where do you see Indian strengths in research? Just looking across the research landscape here,

A (33:48)

that's a tough question. Maybe I'll answer specifically to IIT Madras. Right. So if you look at, if you look at IIT Madras, I think every department has a few researchers who are well known global. There is, there's no doubt about it, they have seen as some of the leading lights in the global research societies in that area. No doubt about it. In every department you find one or two for sure. I think the same is probably true if you compare, let's say the US system, Chinese system, just that the number and the size of the departments are much larger there. And by that, by the same extension they have more people who succeed at that level. And I think we are at a point where we are seeing one or two such good people in a department of say 30, 40, maybe when the department sizes grow, maybe even grow even more, we will have little bit more of a critical mass of such people. And I think after that we may compete with the US and China and maybe, I don't know, US and China are hard. At least the European countries will start competing a little bit better, I think.

B (35:04)

Yeah. You mentioned work on local foundation models. Who's doing that work?

A (35:11)

So there is a company called SAO which is trying. It's not, I don't know if they have a foundation model that's out for anybody to see, but they have the ambition to do that. And let's see, let's see how that works. So I think they work closely with the India Mission as well. The India Mission will know that.

B (35:29)

Yeah. And you mentioned also a professor here,

A (35:32)

Mitesh Kapra, so you can check it out. He was named in the Times AI influencers list, the top hundred. Yeah.

B (35:41)

Although as a journalist, I'm always skeptical because they're journalists.

A (35:46)

But I know him, he runs a pretty big project called AI for Bharat. AI number four. Bharat.

B (35:52)

Yeah.

A (35:53)

And it's pretty interesting. So he's trying to create data sets for Indian languages. Oh, and very interesting things of that sort.

B (36:01)

Yeah, yeah. And hopefully, as I said, I spoke to IBM in Bangalore, the top tech powers, commercial companies are have operations here or hiring people. Do they Give that could be a two edged sword because on the one hand they give support to, to research, academic research institutions. At the other side they're taking people away from the academic institutions without saying anything controversial. What's your view on that?

A (36:35)

Yeah, so it goes both ways. I think at some level as a researcher, if you want to be in a university teaching you have to like interacting with students. So you have to be able to really enjoy being around students, enjoy the process of transforming them into something. That process you have to enjoy. And if you don't then university might be a good place to do research but you won't enjoy it to the fullest extent possible. And I know many who have applied to even iitm our own department and then ended up in Google Research or Microsoft Research was. That's. That's definitely more lucrative than being an academia even though the salaries these days are not too bad in academia as well. Reasonable. Not that naturally a skill case. But I think that critical missing element is do you interact with students? Do you enjoy that interaction? Do you see that as a positive aspect for your life that is not there. Then I guess people go there. So at the. I don't see them necessarily negatively. That's okay. It's. I think it's a two way street at some level. Some of them quit jobs there and come here as well. That happens too. Yeah, yeah.

B (37:42)

No, that's interesting because in a way India, as you said during the IT phase became the back office or the world's corporations.

A (37:54)

Yeah. But will we take the lead in any new research and build a cutting edge product which is out there? So that's something that's trying to think of examples in other areas that we've done it. I think we've done it. If you think of EVs for instance, there are Indian companies which put out cars which look at least to me it's an outsider as good as maybe many of the competitive products out there. So will we do that in this area? As a. It was a great question.

B (38:21)

I think.

A (38:22)

Let's see. I think people more powerful than me are working on it also. I know. Let's hope something will come to do.

B (38:30)

Yeah. And in terms of the. In the AI mission or the National Program to Advance AI, how involved are the IITs in that?

A (38:43)

I think they're quite strongly involved. I think Mitesh, for instance, I keep mentioning his name. I think he's involved quite a few people in our data science department and others are involved. I'm not directly involved with them. I got to know about them. But recently in the context of something else I'm more of a. The more on the AI education side, not so much on the AI development side. Yeah, yeah. So.

B (39:05)

Well, another thing, speaking of AI education, are you using much AI in the program?

A (39:12)

Yeah, so that's the context in which I got to learn about all these India AI mission. So we do have some bots on the. On our portals. We're cleaning to experiment with them. A coding assistant. I think there's an interview bot which interview students and all that beginning to experiment with that. And I think that will also help us scale better with the support that students need to study remotely. I think a bot is a very effective way of doing it. We're experimenting a little bit without it with it. But there is some the issue I see for large scale deployments when I have to Deploy it for 10,000 users, 20,000 users. It's not as cheap as Internet or something, right? Yeah, it's a little bit high cost. And the second thing I see is it makes quite a few errors. I think the fraction of times it makes errors goes to the. I mean it's 10 to 15 seems to be the number these days for critical tasks. Like for instance one task. I think some of the critical tasks we tried was we have coding assignments. So students quote right a piece of code and we can always evaluate it with test cases. But then it'd be nice to have an AI read that code and then give some comments on what was right, what was wrong, what could be changed, etc. And we found in our experiments 10 to 15% of the times it makes mistakes. So quite often it's right. I mean for 90% it's right. It's pretty high. But when you want to deploy it and scale for 10,000 people that 10% will hurt you. So that's the limitation. So these I think are two limitations. One is the cost and I think the energy it consumes, it's really high. Then the error rate, I think it has to come down below 5 I think to my knowledge below 5% like a. Yeah.

B (41:00)

Have you. Do you use any of the commercial API?

A (41:07)

Yeah. So I think the. We don't do it directly. We have companies that help us out with it and I'm sure they are using foundation models or possibly open AI or Gemini. Anything else? Foundation models, API calls and they do some fine tuning on.

B (41:24)

I interviewed Arvind Srinivas, the founder of Plexa. Is he involved at all in.

A (41:31)

He's an IITM grad.

B (41:33)

I Think it.

A (41:35)

Which Madras.

B (41:36)

Madras. Is that right?

A (41:37)

Yeah. So he offered Perplexity for free for all IITM students, including the BS Data Science students. Really?

B (41:44)

Yeah.

A (41:45)

So students are using Perplexity quite a bit.

B (41:48)

I use it very heavily. Yeah. Well, that's wonderful. But even perplexity, the error rate on code review is too high.

A (41:57)

Last time we checked, it was that maybe we should do that.

B (42:00)

Yeah, everything's advancing.

A (42:02)

It's advancing. So we did it about maybe two and a half years ago. So we'll try it again now. Yeah, but I think it's not zero. Definitely. If it comes to around 5%, we'll be very happy. But we are doing specific tasks. So there is a coding assignment. This particular question. Some students have to write code for it and we want to give feedback on that piece of code to the student.

B (42:25)

Right.

A (42:26)

That is a very specific task. And then those kind of things. When you do it repeatedly, the error. Like we are interested in getting the data.

B (42:35)

Yeah. Although I've been traveling, but it seems OpenAI is announced some advancement on reducing hallucinations was.

A (42:46)

Okay, okay.

B (42:48)

The. Yeah. This whole idea of AI for education fascinates me because in the us which is what I think of, but certainly in India, there's so many students that don't have access to quality education. And if there were really a system that was reliable, certainly most students aren't going to stick with it, but multitude students will. Do you think about that? Is that something.

A (43:19)

Yeah, so I think that's something we're definitely looking for. Looking to build some sort of a teaching assistant type bot which should be very, very useful. That's something. I don't think there's a product like that out there. I don't know if we can make this thing. We can certainly use it, we can pay for it. I don't know if we can develop it on our own, if we should do that.

B (43:40)

Yeah. Well, maybe you can just talk generally about how you see AI advancing in India and impacting Indian society. This is my first trip to India. I've read a lot about India, watched a lot of movies and videos about India. But it's different being in a place and it's a very complex society and being here after coming from a world where I'm listening to AI pundits, pundits and researchers who are always looking five to 10 years ahead, you get very excited about the future. But when you come back down to the ground and you look at society, you wonder, yeah, AI can speed up corporate processes, no doubt, but how is AI going to help all of these people struggling to make a living in this very complex society. Do you have any thoughts on that?

A (44:49)

It's a great question. It's a very sort of global question. It brings.

B (44:54)

Yeah, that's right. It's way beyond India.

A (44:58)

So I think in the Indian context there's a lot of fear that the IT backend jobs will go away. And the IT backend jobs lifted a huge section of our society out of poverty. And maybe not even if you said out of poverty, at least got them to a much higher income bracket than was previously thought possible. I think that happened a lot. But there is that fear that those jobs will go away. But there are also people reassuring everyone that other type of backend jobs will come in AI systems. Somebody has to maintain the AI systems that are there else everywhere and we will get there and we get educated, etc, maybe we'll, we'll move to that. That's one part of it. Yes, I agree. The corporate processes will become better. I think a lot of other issues in the country itself, if you look at India fundamentally, one, I'm not sure if you read about it, but the south versus north divide is a pretty significant. So historically the south was a little isolated from all the things that are happening there and that's one part of it. And secondly, I think the early, immediately after independence the early chief ministers had the political. They had, they were really visionaries and they invested a lot in spreading education grad schools level education started a lot of schools, came up with a scheme by which people got food for free schools, etc. Why am I saying? Because the situation in the south is generally dramatically different. You can take any number. You can look at maternal mortality ratio, infant mortality ratio, you can look at gross improvement ratio, any significant statistic. The southern states, Kerala, Tamil Nadu, Karnataka, Andhra, Maharashtra, including those states, way better than many of the northern states. So the key northern states are a little bit behind, so not significantly behind in many of these numbers. Why am I saying that? I'm saying so. So now some of these systems have to get replicated elsewhere. It's education. You can take education.

B (47:01)

Yeah, it's.

A (47:03)

And you need a very good working school system. Right. I think people and people have to believe in that system. They should believe that you go to school, you get good education, come out, then you go to college, you have to actually learn, then do something there and then your life becomes better. So that kind of belief in that system has to come. That is in my opinion, not there in the, in many states in the country in large numbers. There is pockets where it's there, but it's not spread out. And that's where you see the big difference. Not like the south is like advanced significantly beyond the wealthier states, but still it is noticeably better. I'm not sure if moment shift travel, but you'll see that it's noticeably better. So I think the advancements of many of the states can get short circuited if we deploy AI wisely, correctly, and if AI actually advances in making that better. Like for instance, a lot of work that we're doing in doing AI for Indian languages, doing AI and Indian languages, how to convert. How can some farmer living in some small village ask a question and get answers from AI? All of these kind of things is very interesting. So if those things get deployed really in an intelligent way and kids get educated better, farmers do their jobs better, I mean they can grow things better. And more importantly, I think the justice system, the courts in India famously clog all sorts of crazy reasons. If the justice system gets better because of AI, then I think fundamental societal change can happen, alter many things. So that's a dream as such will happen on all the latest.