Pod Insights The Brand Called You Unlocking Predictive Analytics: Alok Aggarwal, CEO & Chief Data Scientist of Scry Analytics | TBCY Podcast About this episode In this insightful episode of "The Brand Called You," host Sandeep Tyagi sits down with Alok Aggarwal, the CEO and Chief Data Scientist of Scry Analytics. Tune in as they discuss Alok's journey from setting up IBM's research lab in India to co-founding EvaluServe and leading Scry Analytics. Discover the challenges and opportunities in the knowledge management systems space, the limitations of Large Language Models (LLMs) in B2B scenarios, and the potential economic impact of AI. Alok Aggarwal sheds light on the development of Auriga, a sophisticated knowledge management system tailored for complex industries like investment banking. He also reflects on the transformation of the KPO industry and offers his thoughts on current issues, including climate change and technology hype cycles.
Transcript 0:08 Speaker 00 Hello, I am Sandeep Tiwari, your host for the brand called You and the brand we are speaking with today is Alok Aggarwal. Alok is the CEO and Chief Data Scientist of Scry Analytics. In December 2000, he had also started and co-founded the company called E-Value Serve, which grew to 4,500 people in many, many countries. And prior to that, he established IBM's Research Lab in India. He's also worked at MIT and IBM Watson Research Center. So he's a man of many talents.
0:48 Speaker 00 He's got a BTEC from I T Delhi, which I share with him, though he was many years my senior. And he's got a PhD from Johns Hopkins. He's got over 100 research papers and eight patents. So we are delighted to have you, Alok. Welcome to the show.
1:07 Speaker 01 Thank you for having me. Really happy to be here.
1:10 Speaker 00 Let's get started with your professional journey. Tell us about Scry Analytics. First of all, what does the name actually mean? And I was reading about it a little bit. You've talked about AI and I've heard a lot of stories, but maybe you could tell us a really interesting implementation or actual impact that you're having using AI at Scry.
1:37 Speaker 01 Sure. So first of all, about my story, I would start by saying I'm an accidental entrepreneur. And it all started, as you rightly pointed out, because I started IBM India Research Lab during 1998 and the year 2000. We started in January 1998, grew it to 30 PhDs, 30 masters. And it started in a very interesting way because my boss at that time, who was head of research, Dr. Paul Horn, he said, here is, I will act as your VC, here is $3 million. I expect you to use it in the best way to grow IBM India Research Lab and make it famous. So that was my first foray into actually starting something new.
2:22 Speaker 01 With a different goal, I mean, the goal was not to increase revenue for IBM, but more to increase branding and marketing. But then, as you rightly said, in December two thousand, January two thousand one, we started Evalysur with the co-founder Mark Wohl and Wider, who had moved from McKinsey Zurich to McKinsey in Delhi. And we met in April two thousand because our kids were going to the same school, American Embassy School in Delhi. And our wives met and they put us together. So that's how we started. And then I quit to start Scry, which we will discuss with more.
3:01 Speaker 01 But the reason I call myself an accidental entrepreneur is because when I graduated from Hopkins with a PhD in basically AI in theory of computer vision, I never had ever imagined that I would become an entrepreneur. In fact, I wanted to be a professor. So I joined IBM Research to do a lot of research and publish papers so that I can get a good tenured position in a college or in a university. Happened to be on a sabbatical at MIT in nineteen ninety, nineteen eighty-nine, and then I realized that I don't want to be a professor.
3:39 Speaker 01 And because it requires a lot of being on committees, writing NSF proposals and a lot of stuff which we were not doing at IBM Research, so I came back. So life has changed every five or ten years for me, and that's why I don't ask people when I'm interviewing, what will you do in five years? Because everything has changed.
4:01 Speaker 01 So that in short is, I mean, how it all started. The word Scry actually is an English word. You can use it in travel, convert cry to spy. It means crystal ball gazing or fortune telling. So Nostradamus is considered the biggest cry because he wrote all the futuristic narratives and so on.
4:22 Speaker 01 And given that we are doing, creating products which are in, with respect to predictive analytics, we are predicting the future. Whether an excavator will die in the next sixty days, whether a person will return or a company will return the money in next sixty days or ninety days, whether the supply chain will be broken in the next few months or few weeks. So a lot of what we are doing, a lot of what P I is about, is about predicting forecasting the future by looking at the past data. So, uh, uh, so SRI actually essentially is somewhat based on the model of IBM Research or Bell Labs or Xerox PARC.
5:05 Speaker 01 My view, when I spent about sixteen years at IBM Research and my view at the end became that if IBM Research or Xerox PARC or Bell Labs or now Google DeepMind had their own consultative salespeople, they would be by themselves trillion dollar companies. They had to rely on salespeople from IBM Division and who did not really understand, who would rather sell hardware, whether it was A S four hundreds or supercomputers, did not really understand what IBM Research does and therefore in spite of the fact that there is a lot of innovation in these areas and in these companies, divisions particularly, it was not getting sold.
5:46 Speaker 01 So we basically, fundamentally, I would say we are an R and D company, but we use R and D to create products. We have six products out of which we are selling six product lines, out of which we are selling three very, very strongly. And we have consultative salespeople. So we started in February two thousand fourteen. So I won't call it a startup anymore. We are about three hundred and twenty people, out of which around forty are in, in San Jose, in Silicon Valley, where we have the largest, we have a complete R and D center. We only hire masters and PhDs. And then in Gurgaon and in India, Pune and Hyderabad. Again, we have another forty who are deeply into research.
6:28 Speaker 01 Um, another sixty who are in advanced development and the remaining managing various products and upgrading.
6:38 Speaker 00 Wow, that's a sizable team for deep research or um so give us some examples. What are the three products that you have built up? And and maybe take us a little deeper into why? Why is it? Um, how is it connected to the cutting edge research that's been done?
7:00 Speaker 01 So, so the first one is Colatio. All of them have effectively Latin names because we wanted to get trademarks. So Colatio in Latin, it's pronounced as Colatio. Colatio is same as collation or putting things together. But it's all about taking data which comes in PowerPoint presentations, SharePoint, PDF documents, uh, scanned documents, you name it. I mean, you can X B R, L, X M L and so on, converting it into electronic data. And then doing something with it. Now, let me take the example of scanned page or a picture. Let's say you have a picture you want to convert it.
7:40 Speaker 01 Now, this has been around for at least last thirty years on optical character recognition. Now, and we've been improving optical character recognition and its accuracy quite a bit. Currently, most of the companies who are in this business will give you ninety to ninety-two percent accuracy. So you're given, let's say, a table or in fact, let's start with text. You're given text and if it's a bad document, not that legible, then it'll probably be more like eighty-five percent. If it's a good document, like ninety-five. I, you are, let's say, an insurance company and I come to you and say, I can do a, a, or OCR or optical character recognition very well.
8:24 Speaker 01 And my accuracy is of my system is ninety percent. Your immediate question is, that's great. I have documents which are three hundred pages. Now, would you be able to tell your system be able to tell me as to where is likely to be wrong? Because I read the whole, the whole three hundred page document and then you're not saving me that much time or that much time. But if you could, even if you're ninety percent accurate or your system is ninety percent accurate, you can even tell me twenty percent of the places where you would be wrong, but the eighty percent are all correct. Then I'm good to go because I only have to look at twenty percent and therefore you have right of the bat, you'll save me eighty percent time and therefore money.
9:07 Speaker 01 So, Colleccio basically is an intelligent document processing and reconciliation system.
9:14 Speaker 00 Hmm.
9:14 Speaker 01 Could say upfront as to where it is likely to be wrong.
9:19 Speaker 00 Hmm.
9:20 Speaker 01 So, and where it has been able to reconcile. So, it reconciles in many different. One of them, for example, in a table, let's say income statement. If, if attribute A plus attribute B plus attribute C equals D. That's what you see like in to a revenue, you would see revenue from products, revenue from services, revenue from other in, uh, other sources, total revenue. So, on the fly, just like a human, it understands that there is a formula there. And that there is a formula A plus B plus C equals D in this particular document and every document is different. In this particular, there is a formula there.
10:01 Speaker 01 And then it uses that to see whether the formula, actually the numbers actually add up or not.
10:08 Speaker 00 I see.
10:08 Speaker 01 Addition, subtraction, multiplication. For example, in W two's, which in the US you get at the end of the year. Uh, you'll get a, a person will get the salary, let's say it's hundred thousand. And social security tax on that salary should be six point two percent.
10:25 Speaker 00 Sure.
10:25 Speaker 01 But it figures out, after seeing a few W two's, it figures out that look, the formula is six point two percent. And if it actually shows that the social security is sixty two hundred. It has figured out that indeed, uh, both of those numbers are correct. Or Medicare, which is one point four five percent. So, I just gave you some very formulas, but it could be with names. Like Alok Aggarwal or John Doe and so on, it could be addresses. It uses many different ways to reconcile. Both within the document, across the document.
11:00 Speaker 00 Yeah, yeah. These, um, I guess heuristics or these rules for how to validate something. That revenue may be broken down or address has to this be this form or social security is six point two percent. Do you end up using a lot of rules like if then statements or these are somehow deciphered from the documents themselves?
11:25 Speaker 01 These are deciphered from the documents. We humans are so creative and we'll maybe talk about artificial general intelligence, which I don't believe in that we are anywhere close. The reason is we humans are so creative.
11:37 Speaker 00 Hmm.
11:37 Speaker 01 We love to break all rules, even for well structured documents which are following international financial reporting standards, IFRS or GAAP. Even then the accountants will change things around, will add things up, down in different ways and so on and so forth. So you cannot, a business rule system by nature will fail automatically. So the whole area aim is just like we train our children, you train machines to learn the same things in a different way so that you can then test them and they learn how to decipher them. So this.
12:17 Speaker 00 So, so does it work with a version of truth? You need to have a sample set and say these ones are accurate or how does it extract and know that these things, you know, how does it differentiate between something which is a balance sheet versus something, let's say, which is a list of telephone numbers?
12:38 Speaker 01 Yeah, very good question. So it basically figures out from a balance sheet, it figures out using natural language processing that the key words or key terms are those related to a balance sheet as opposed to those income related to an income statement. For example, a balance sheet is likely to have key terms and key words related to assets, liabilities and equity, whereas income statement would have revenue and profit. Now they may not be calling it revenue, they may have a different name for it and that's where natural language processing and lot of semantic searching comes into picture because the syntax may change but the semantics usually does.
13:20 Speaker 00 So are you training these models or you using any of the large models like chat?
13:27 Speaker 01 Yeah, so these models are mainly trained models. We have trained over the last since two thousand and sixteen. So they are very strongly trained in figuring out and that's why we have trained it on about one sixty different document types. Now, because we have trained it on one sixty different document types, new training for a new document types becomes easier. The problem using large language models or GPTs is twofold. One, they're not, they've been trained on four hundred million to about a billion pages of of words, mainly words. Very few tables, very few formulas and so on. So they're not able to decipher formulas. In fact, they go wrong on tables.
14:12 Speaker 01 They even go wrong on tables by. So if a table, if the human is really creative and creates an interesting table, they will be only seventy percent.
14:23 Speaker 01 And the important part of Polatio is table recognition, table extraction and table reconstruction. And similarly for graphs and charts, so suppose you have a bar graph, vertical bar graph or a horizontal bar graph, you want to extract all that.
14:39 Speaker 00 Very interesting. So you have trained your own models. Are these gigantic models with millions of parameters?
14:45 Speaker 01 I mean gigantic, gigantic is all relative, right? So we used to think they are all gigantic. Now you don't think so because they are about two gigabytes in in total memory. And when you look at a typical Llama three point two, which is a open source GPT, it's more like ninety six than ninety two to ninety six gigabytes in. So from that perspective, I'll say it's a small model. But again, there are small switch models also which are two to four gigabytes. So it's reasonable. We've trained it on about, I mean, we've, we can't afford to and we haven't trained in four hundred million pages.
15:25 Speaker 01 But we've trained it on about a hundred, on about a million pages worth of important data that is important to us. And therefore.
15:32 Speaker 00 That's very impressive, very impressive.
15:36 Speaker 00 Very impressive.
15:37 Speaker 01 Yes, but we're getting there. We're getting to the human level accuracy. So within the company, and I'll talk about this in a minute, we call the minimum accuracy required, which is ninety percent. Goal level is ninety-five percent and human level is ninety-eight point four. It's not that we are always accurate either. In fact, you give a really bad microfiche or a typical Indian fax from nineteen nineties or two thousand, I mean, most humans won't be able to understand what is written.
16:10 Speaker 01 And so we are trying to get to a ninety-eight point five percent accuracy. Human level within the next year, year and a half with the hope that the remaining one and a half, we can show actually these are the five percent where humans should be looking at. So it becomes a maker checker problem that we see all the time in various business process outsourcing. Somebody is doing the invoice, I mean converting invoices, somebody is checking it. So we want the maker and the AI system and checker to be human and the checker only spends five to ten percent reviewing everything.
16:48 Speaker 00 But to identify where the confidence level is low, do you use a separate expert system? Or to kind of like evaluate what to, whether this would be?
16:58 Speaker 01 This is all, I mean, and that's probably happening with humans too, that there are independent algorithms which are working in conjunction with each other. That look, this doesn't make sense. And so there are independent algorithms, models which are there to check for the confidence level. But because just saying that the OCR confidence is low is actually not a very good measure because by nature, these deep learning networks are so confident and they're overconfident actually that they mess up. I mean, that's why we have hallucinations because they, they think they know everything.
17:36 Speaker 00 That's great, Alok. So, so what else? You said six products. Tell us about one more, which is where you have done something interesting.
17:44 Speaker 01 So the second one is what we call Origa. Origa is a constellation of stars as old, I mean, from human perspective as Mesopotamian race. Because it's, it's a very, very well-known constellation, pretty much all societies have it in their culture. And, and to us, what we're doing is with Colatio, we're actually creating knowledge management systems for companies. Now, L L M's, G P Ts are all trained on internet data. They're not trained on the company data because no company will give them the confidential data or give us the.
18:25 Speaker 01 So the whole point in, in L L M's, if you get out of the B to C domain, because I, I can ask questions, of course, we are all excited by it. You can ask questions, I can say, look, I have a ten-year-old son, he, he has a birthday party tomorrow, can you tell me what I should do? And he'll give a good answer and then I can then give a hint, hey, but I want to get a clown, is that a good idea? And then he'll tell me all, the L L M will tell me all about the clown. Or I can ask questions about a culture, in Indian culture, on a birthday, you put a thilla. These are effectively hints that you're giving to an L L M. It's very similar as if you were training a child and then you were testing a child
19:08 Speaker 01 and giving hints to the child when the child is not able to, to completely go in the direction that you want. Works very well for B to C questions. There are some hallucinations where it just goes completely wild and answers wrong. It happens with humans, happens at least hundred times more with, with these machines. The problem is this only stops with, because I'm giving hints, which are often called prompts. And we don't like to call them prompt engineering because there is less engineering, more, in my view, creativity. Because I want it to go in a certain way, so I have to give the right hint, whether it's a human or it's the system.
19:47 Speaker 01 An L L M. So but it doesn't work in a B two B domain because now it has not been trained on that data. And now you are suddenly putting a data, let's say, three hundred thousand pages worth of bank data into it. And you want to ask questions. Hey, can you tell me, let's say it's an investment bank. And an investment bank or equity research group wants to compare the second quarter report to the third quarter report for Wells Fargo, right? They can say, hey, can you tell me what is the second quarter report, second quarter revenue versus third quarter revenue for for the for Wells Fargo, for the company? For most likely, L L M will be able to answer because that's a pretty simple question.
20:32 Speaker 01 But then I suppose I ask a question, what is the CASA deposits for Wells Fargo versus in Q two versus Q three, second quarter versus year? Now even if I really understand what CASA may mean, actually it is checking account plus savings accounts deposit. And now this is a, this is a complex question because the report may not have, Wells Fargo may not have given an answer of CASA or even checking account plus savings. It may have given checking account and savings account separately. So you need to now combine. I mean, that's where human intelligence comes in somewhat L L M. So now somebody has to give that particular problem.
21:16 Speaker 01 That look, you may be going in the wrong direction. I want checking account plus savings. Now if you do all of this prompt creativity or prompt engineering for the L L M for an internal data, the analyst will get frustrated and say, look, I might as well look at it myself rather than giving all these prompts. So what Origah does is it actually, just like Quilesio tries to tell you as to where it could be wrong. Origah is an automated prompt generation system using subject matter expertise. So it's basically will be different for
21:56 Speaker 01 just like humans are different for accounting and you have different humans in insurance, you have insurance and the writer. So Orinda is getting has a semantic search layer which is very, very specific and tailored with subject matter expertise to the particular industry vertical that you are.
22:16 Speaker 00 So you're creating various personas with its own guardrails and its own prompt hints to say, do this, do that.
22:24 Speaker 01 Yeah, we can create prompts because I don't know what you as a user will will ask, right? So we can't create prompts and we don't expect you to give prompts as a user because then you'll get frustrated. You will say, I might as well do it myself. I might just review what what the UBS Fargo quarterly reports say. So it has to, the semantic search engine has to come up with automated prompts, fifteen to twenty automated prompts to answer your question. Now, if there are hundred prompts, just like with the human, the choices are too many, the human will, the child will get confused. Hey, you're giving me too many hints. Which one should I take? If you give too many, too few hints, you may not hit the thing in the right thing. So the semantic
23:07 Speaker 01 search engine has to give, I mean, we've seen fifteen to twenty is typically a good number.
23:13 Speaker 00 Very interesting. So this is then created for a specific use or specific use case.
23:20 Speaker 01 For investment banking, we have a very specific use case. There are areas which are much easier because you don't have subject matter expertise as much there, like call centers or somebody's asking about, hey, I want to visit Dubai. Can you tell me what is interesting places to check? Some of it is already in the internet. But so there are areas where you don't need so much subject matter expertise and therefore you don't need an automated hint processing system or a prompt creation system. But then some areas are very complicated and the most complex, of course, is healthcare, which we are not touching yet.
23:59 Speaker 00 Okay, got it. Changing gears, let's talk about your previous avatar where you were an evalue sir, founded it, ran it for fourteen years. Tell us about that journey, how was it in Gurgaon in that early time? And what do you see the impact of what you're doing now and what's happening in Gen AI on that whole industry where people were just going offshore to do a bunch of knowledge work, in which you guys were pioneers and lots of other people have been doing that work since then and making it into a big industry?
24:35 Speaker 01 Yeah, so that's a very interesting era and I think it's about every industry and every company that it goes through its growth, it flattens off and then external factors come in and it begins to go down or it needs to be revitalized. That is true about even Gurgaon, I mean, Gurgaon was a farmland, people, Americans often ask me about Gurgaon, they're hearing a lot about it and a lot of multinationals have moved there. And I said, yeah, in some sense, it's very similar to Silicon Valley. In nineteen sixty-five to nineteen seventy, Silicon Valley was also farmland and then Hewlett-Pickard came and it began to grow with Fairchild and then with Intel and so on.
25:19 Speaker 01 So Gurgaon, it was thirty, thirty years later. So that itself was a very interesting period for where we started the company. The knowledge process outsourcing industry was also interesting because here we were, instead of going with IT industry or business process outsourcing, which was getting very famous at that time, doing call centers from India, setting up companies which would, which would clean up your invoices, begin to figure out your financial statements. We were talking about doing research and analytics for investment banking, for McKinsey type strategy companies, Boston Consulting Group and the like. So it was a very interesting area.
26:02 Speaker 01 One thing I learned was, which I actually write in the book also that I published about ten months ago, is we were, we as inventors, innovators, entrepreneurs get very excited and we think because it's such an interesting technology or an interesting innovation, it's going to catch fire, it's going to sell like hot. I wrote, I'm guilty of it. I wrote at least four, five research papers. How much K P O will grow, knowledge process outsourcing will go in India during the next ten years. By twenty thousand eleven, I think I said eleven billion. And it grew by only two and a half or three.
26:47 Speaker 01 This is a typical problem we see, and this is a problem with G N A I today. People are talking, McKinsey report says that total A I will have an economic impact of anywhere from eleven trillion, not billion, but trillion dollars to twenty two trillion by twenty thirty. Now, if you look at the total G D P of the world by twenty thirty, it's currently hundred and ten trillion, it's likely to be one fifty six trillion. So you're talking about ten percent of the world's G D P will depend on A I. Right now, our analysis shows that there is no more than thirty billion of total worldwide that A I is making a change in.
27:32 Speaker 01 So this implies that you're doubling, you're more than doubling every year. And we know doubling is easy when you're starting from zero or one, but doubling becomes very hard when you're in the billions. And definitely very hard when you're in your trillions, almost impossible.
27:49 Speaker 00 Right.
27:49 Speaker 01 So, I mean, more recently, somebody said it'll be hundred trillion dollars. And I wrote a small post on it saying that, look, the U S, the world economy will be one fifty six. And you're saying hundred trillion dollars will means two third of the world economy will depend on G N A I. I'm not blaming these guys because I was myself part of the, part of the group, which at that time got excited. I mean, this excitement always comes about. Then investors get excited, then the general public gets excited, then the hype comes, goes up through the roof, goes to outside Milky Way, and then everything falls back down.
28:29 Speaker 00 But you know, I think Bill Gates wrote this line many years ago in his book, that technology, like many other of these changes, they change lot less in the near term than people expect. But they change a lot more in the longer term than people also expect.
28:48 Speaker 01 Absolutely. And that was my, I mean, chapter one of the book, The Fourth Industrial Revolution and Hundred Years of AI. I bring that out, that each revolution had its height. Railway was hyped in eighteen forty. Lot of capital went in, railway tracks were put in, gigantic railway stations were created. The masses did not start using it till most of the railway companies collapsed. And then it became so cheap that you had not only people using it, you had freight moving. And suddenly Procter and Gamble, which was producing gambles, candles, soap and some perfumes in Cincinnati,
29:29 Speaker 01 realized that railway freight was so cheap, they can sell it in New York. There are unintended consequences. We would have never thought in nineteen ninety-six, sitting in IBM Research, that one of the consequences will be social media coming out in two thousand five to two thousand ten. Telemedicine. I mean, so absolutely Bill Gates is right. And I bring that out, that actually hype may be bad for the investors who lose their shirt. Maybe bad for innovators like us who become despondent when it goes bust. But it's actually good for the society because lot of capital and lot of passion goes into it.
30:08 Speaker 00 Sure.
30:09 Speaker 01 So going back to the last aspect of your question, I do believe that these L L Ms and what we are doing with Oregga will create a K P O two points here. Because you will not be again able to take away humans, that's not the aim right now. But again, the humans will become checkers rather than fleet, just like what explained. And there is, it won't happen overnight, it won't even, in my view, won't even happen in the next three years. But I think over next ten years, the Indian outsourcing industry in general, but in particular, Indian IT industry and KPO industry and BPO industry will see a lot of change. KPO particularly because of Gen I and LLM, because a lot of KPO is about search and explore.
30:54 Speaker 01 And, and asking questions about your data and about external. So I do believe that that's going to happen. I'm still, by the way, involved only peripherally with the value service, I still have quite a few shares. But, but because I'm running Scry, I can't, it's impossible, I'm the only founder, it's impossible for me to run.
31:15 Speaker 00 Sure, I understand. Well, let's change gears, talk a little bit about your personal stuff. Where were you born and what was your childhood like?
31:24 Speaker 01 I was born in Delhi in nineteen fifty nine. I, my childhood was, I don't think there was anything very different about, I mean, very out of the ordinary about my childhood. I, I was born in a middle-class family. My father was an advertising manager in a newspaper called The Statesman, probably the oldest newspaper along with The Times of India. Used to live in Karol Bagh in Delhi, went to a private school. Most people, they are go to anywhere, especially at that time, but even now. Went to IIT Delhi, I didn't want to go to any other ITs because home was closed. I could go back to on Fridays, home, get some really good food.
32:08 Speaker 01 And that, so I, I, I mean, other than at that time, my passion in high school was cricket and maths. And it continued for a long time. But other than that, I wouldn't say it was anything out of the ordinary or remarkable in anything.
32:23 Speaker 00 You said your dad was an advertising executive. What do you think you have taken from him? What part of his personality, his value system, his beliefs, his way of doing things do you find reflecting in you? And what from your mother?
32:42 Speaker 01 So, most of it comes from my mother. And I think that is probably not only true about me, but most, many other people because mothers are closer. They speak much more than dads do. I mean, men versus women issue. But also if they are homemakers, mothers are homemakers, they obviously spend more time with their kids. So, I think I got a lot of, a lot of personality or things that I do, I actually probably got it from her. One thing which I got it from him, he was very, very, very risk averse. But, but he was actually, he was a master. I had a master's in English and a master's in Urdu.
33:26 Speaker 01 And he would always tell me that you should learn more language. And I actually tried to avoid, which was a very big mistake. I could have learnt Sanskrit in high school and I blew it off. I could have learnt German because it was required in my PhD by Hopkins. I blew that off. And I, in retrospect, I thought that was a very big mistake. But the thing he told me was a particular verse from Urdu, which I'll say it out and then I'll translate it. Khud savar hi maidani jang mein girtay hain. Woh tefla kya girege jo khutno ke bal chalte hain? That it's only riders, horse riders who are fighting the war will fall down.
34:11 Speaker 01 But if you're just crawling on, on the floor, it's almost impossible for you to fall down. So, you have to take risks to become a horse rider and you will fall down, you have to just assume that. And that, I think, has reverberated with me for, I mean, pretty much for my life.
34:32 Speaker 00 Very interesting.
34:34 Speaker 00 Um. So even though he was risk averse, he gave you something which you have used to take some risks in life.
34:40 Speaker 01 Exactly. He was extremely risk averse and in retrospect, when he passed away, um, a few years ago, then I always would, I would think that, look, he, it was very interesting that he was so risk averse and yet he would speak, I mean, he would say this probably once a month, at least, if not more.
34:59 Speaker 00 Interesting. And from your mom?
35:02 Speaker 01 From my mom, I learnt a lot. I mean, one I learnt was that she, unfortunately, because women were not supposed to, to be educated. I mean, when she grew up in nineteen twenty five, she was taken out, but she was extremely, extremely intelligent. And she, even though she had just passed eighth grade, uh, after that, she continued reading newspapers and so on. And what I learned from her was reading at least half an hour to an hour a day, it actually keeps you on your toes. I have actually fortunately continued, I still try to browse through uh, five different types of um, news media.
35:45 Speaker 01 I don't have any accounts on Facebook, et cetera, because that's just too time consuming.
35:53 Speaker 01 Too much of a time sink, but five uh, like CNN, MSNBC, Yahoo, a few others. And then five in India, in the Stan Time Statesman, just browse through the most important ones. Um, takes me about an hour and that is something I learned from. But then there are many other things I, I learned from. She was actually a risk taker as opposed to my dad, so.
36:16 Speaker 00 Very interesting. Okay, well, the next section, I'm gonna ask you some quick questions, no right, wrong answer, whatever comes to your mind, that's the right answer. What's your favorite cuisine? What's your comfort food?
36:29 Speaker 01 I think Indian cuisine would be my favorite, followed by Chinese.
36:34 Speaker 00 What's a favorite book? If you were meeting a nephew who hadn't met and you wanted to give him some advice or her some advice and wanted to gift a book, which one would you give them?
36:47 Speaker 01 I, over the years, I've become very, very fond of Hindu culture, not Hindu religion, but Hindu culture. And even though it is hard to understand, I think it is worthwhile spending the time and effort. So actually, I would give him Bhagavad Gita, not any translation, I mean, translation in the, sure, but not trans, not anyone's views. And I would say, just, I want you to just read chapter one and chapter two. Because I think, I don't look at Krishna as a god or Lord Krishna. And I don't even know whether he spoke Bhagavad Gita completely, or whether it was entirely
37:28 Speaker 01 an imagination of the writer Vedvias, who was writing it to explain things about, about the universe, about Indian society. So I would like to, I mean, and I go back, back and forth, and that's something which I would like to give to a person and tell him that, look, read it and actually analyze it. Don't go, go to anybody else and listen for their views, get your own views.
37:59 Speaker 00 Yeah. What activity besides your work and family gives you joy?
38:05 Speaker 01 So I, I, as I said earlier, I'm more of an accidental entrepreneur. I'm more of, I consider more of as a builder, being engineers that you and I are. So one of the things which I have enjoyed, and again, it happened by chance, happenstance, that I started building houses. So in New York, I built about sixteen houses, and I was also, my arms got twisted, fifteen or sixteen, to build a Hindu temple upstate New York. And then I built my own house here in Silicon Valley, and then one other house. So. So I actually, how do I build it? I designed them quarter inch to a foot scale by hand.
38:47 Speaker 01 And then I gave it to, originally I would give it to a person to draft it into a blueprint. That was in two hun- uh, two thousand up to two thousand, two thousand five, it became all computer aided design. So now I still give it to somebody to put it on a computer aided design. Then I'll hire successful contractors, not not give it to a general contractor and then uh. So that, to me, it's a passion, it's not a hobby because it takes too much time. And I haven't done it frankly since I started Cry, because Cryonics, I'm the only founder, so I don't have time to build houses. But I.
39:23 Speaker 00 Very interesting. Okay, maybe I'll call you up, I'll look and I wanna build a house, get a design from you.
39:32 Speaker 01 Design, I think engineers' designs are much different than architects, by the way. Engineers by nature are less artistic and more, hey, does it work or doesn't it work? Does uh, it's in good, functional, exactly.
39:46 Speaker 00 Yeah, yeah. What is your biggest concern about the state of the world?
39:52 Speaker 01 My, I think I'm not the only one. I do believe that we as a society is not putting enough e- effort to on things related to climate change. And I think it'll come and hit us fairly hard and when it hits us. Uh, it uh, it will, uh, there will be millions of people whose livelihoods will be destroyed. And some may die or may get really, health wise, get hurt very badly. And I think we are putting too much effort on AI and job losses and jobs gained. Which again, as I said in uh, when we were discussing KPO, I don't think that is going to happen anytime soon.
40:34 Speaker 01 And by the time twenty fifty comes along and you have jobs losses and jobs gained. The human population would have stabilized and would like, would start declining. Because fifty five countries would have a declining population, India would have a stable population at that time. So we are literally be begging AI to be doing the work for us because we won't have enough people who are working age group to do the work. So I'm not actually worried about AI, I'm more worried about climate change. And it reminds me of what happened in nineteen sixties, which of course a lot of people either don't know or have forgotten about it. That we have a massive famine in India and China during those periods.
41:17 Speaker 01 And everyone thought that India and China will just deteriorate and go down. And I think India lost something like five or six million people because of starvation. China, nobody really knows, but the stories are awful. Probably they lost much more, many more people. But then in the green revolution came. So the good news is humans, once they put their mind to it, they can solve even the most complex problems. Big problem. The bad news is that it takes a lot for human society to realize that this is an enormous problem and the human race, Homo sapiens, are at stake at this point.
42:00 Speaker 00 Right.
42:00 Speaker 01 And that is what my biggest worry is.
42:03 Speaker 00 Okay. We are way past our time, but we close out with this most important question. What is Bran Alok Aggarwal in your own words?
42:16 Speaker 01 I would say there are five or six characteristics I would like to have and I would like to portray. Passion, innovation, humility above all, because we always rise on the shoulders of others. Integrity and grit. Grit is a hard work to achieve your passion to. And I often call it Pi Hig, for in short, passion, innovation, humility, integrity and.
42:46 Speaker 00 Wow, integrity and what's the last one?
42:49 Speaker 01 Grit, G I R I T, grit.
42:52 Speaker 00 Okay.
42:52 Speaker 01 So the ability to literally turn your passion into reality. I mean, that's the perseverance always reminds me of a very, I mean, I start the book on the first chapter with the following interesting episode. In early nineteen hundreds, Edison's very good friend, Robert Melloi, visited Edison in his lab. And he says, for all the work you have done for so many years in creating a nickel iron battery, you have nothing to show for. And Edison's quipped by saying, why, man? I know thousands of ways in which I can't make nickel iron battery in.
43:35 Speaker 00 There you go.
43:35 Speaker 01 That's the grit, right? That's the grit of creating an incandescent lamp bulb or a nickel iron. That is what I think we innovators require. That is what human society requires is to convert passion into something you need extreme amount of. Because you will be drawn from the horse every night.
44:00 Speaker 00 Thank you so much for joining us, Alok. It was wonderful to catch up with you and record this. And there you have it, people, the brand for today, Alok Aggarwal, the Pyg man. I think that that should be likely a title of your book. What's the book called, by the way?
44:16 Speaker 01 The fourth industrial revolution in hundred years of A I nineteen fifty to twenty fifty. It's on Amazon. We are actually converting into an audio conversation, thirty five conversations of approximately half an hour each, which will put on Salman Khan Academy and Khan Academy and Porsera for your charge.
44:37 Speaker 00 Excellent. There you have it. Thank you so much, Alok, once again.
44:41 Speaker 01 Thanks, Sandeep. Thanks for having me.
44:45 Speaker 02 Thank you for listening to the brand called you videocast and podcast, a platform that brings you knowledge, experience and wisdom of hundreds of successful individuals from around the world. Do visit our website, W W W dot T B C Y dot in to watch and listen to the stories of many more individuals. You can also follow us on YouTube, Facebook, Instagram and Twitter. Just search for the brand called you.
Previous Unveiling Integrity: Haykuhi Harutunian, Former Chair of Armenia’s Corruption Prevention Commission Next Data, AI, and Empowering Women | Asha Saxena, Founder & CEO WLDA; Author, ‘The AI Factor’ Page updated November 15, 2026 · transcript and summary are machine-generated