If I Get Another Chance, I’ll Buy Google Chrome — Perplexity CEO Aravind Srinivas x Gobinath

Yes, the most wanted Arvind Srinivas is here, CEO and co-founder of Perplexity AI. Can you define who is Perplexity? Is it my friend or my teacher or companion or a soul mate? It's meant to be a teacher.

It's a marriage of Google and Chargipiti. Okay.

This population is trying to break the monopoly of Google yes Parallel me your answer engine which is a Ability company a chroma walk on Google. They have been it was an opportunistic attempt to encourage more competition What about a feature that my buddy in the mariana or the soul a live and die? Roma, I will see me was one of the cheaper and a person Google will work under a giggle a If we ever see this again in the future, Arvind Srinivas will try to buy Chrome.

100%

You are a big shot in Google.

No, competition is important.

I love this attitude.

Chrome is already there, Firefox is there. So, there is no need for a Comet browser.

Donald Trump can use only one iPhone. But Donald Trump's life His assistants His team will be there to set up his life That's not you

So we are building that now Do you watch Tamil movies? Sometimes Who is your favorite actor?

Rajinikanth

Your favorite cricketer?

Sachin

Who are the people entrepreneurs you follow?

Mukesh Ambani, I met him once He told me a quote Content is king, but distribution is God

Recently you met our Prime Minister Mr. Narendra Modi What was the discussion?

His fear was...

One piece of advice to the students

Our curiosity and creativity have limits That is the one human skill that AI cannot exhibit Being curious

I am Aravind Srinivas Enum Nan Title Sponsors Appaswamy Real Estates Dr. Sri Varma's Wellness Supported by NexCura Powered by Gen AI Healthcare Vishar Gurukulam

Powered by Pissupilladha Sunland Danish Ahmed College of Engineering

Stratos Alteza on OMR by Appaswamy Real Estates Powered by Pisi Pisi Pilladha Sunland, Danish Ahmed College of Engineering. care. Knowledge, opportunity and community. Welcome to Vishabh Gurukulam Music School, where music learning is redefined.

Pisi pisi piladhu sunland refined sunflower oil. Sunlandle samachaa shakthi thaana varu. Danish Chennai, a school of intellectual powerhouse.

Anivarakum vanakkam, welcome to Aravind Srinivas Enum Naan. This podcast is very important not only for me but for everyone watching it When I put up interviews on technology and science on my channel, especially on artificial intelligence I get comments within 5 minutes, asking when I am going to interview him Gopi, do an interview with him. He has given time for his busy schedule. I am in San Francisco now in his office. Yes, the most wanted Arvind Srinivas is here. CEO and co-founder of Perplexity AI.

The answers are perfect. You said that answers are perfect, a world is my dream. What do you have to say about that? It seems like there is a philosophy behind it.

When teachers are teaching in our school, the student who knows the answer all the time, everyone will say, oh, he is the smartest person in the room. We build up an incentive system over time. In an exam, the person who knows more about the subject is the smartest in the class.

So, it is deeply integrated at a system level. But reality now is AI can answer almost anything. Almost accurately. Mistakes will occur, but every year the progress is so much that the long tail of mistakes is also reducing. In such a world, we should create incentives

where you are encouraging people to ask questions, not know things. Oh, okay. You have to move from the era of knowing answers to everything to asking the right questions. Professor students don't like it when I ask them a lot of questions. Other students don't like it either. Sometimes students will ask questions for no reason.

But when they ask authentic questions, the flow of the lecture will get disrupted. But we should go back and think, why should a lecture be taught in the same way? We had horses, then we moved on to cars. We had people manually doing arithmetic on paper,

we moved to calculators, computers. Lot of the way in which our world works has changed. We used to write letters, now we send WhatsApp texts. But the way we teach has remained the same. It's in the same classroom, there are benches, students sitting, there's a board, there's a teacher,

and there's a linear lecture going on. So, when you want to learn a new topic, you go through the same kind of lecture that happened 50 years ago. Why should it continue that way? Namah undhu like, ithaar ureya hai. So, obviously perplexity, chachi piti, you go, how do you start there? You start with a question.

Question, yes. You don't say, teach me and just watch a video. You start with a question, then you start another question, another question, keeps going, keeps going. How you uncover the knowledge through the unit of question as a first class citizen is the fundamental change. Now, how do we change teachers into education teachers? That is an open question. I'll tell you one experiment we did. MIT, Massachusetts Institute of Technology.

Introduction to biology class. That professor reached out to me and said, Hey, I want to teach this semester only with perplexity. What do we think about biology? Do we have to study a lot about biology? Yes. In fact, we have to we should study biology a lot. In fact, we have a word in Tamil, we should study a lot.

Right? We have a thing called memorizing. But that got changed entirely, where both the teaching assistants and the students were in one interface, perplexity. And they tracked what questions the students asked, and rewarded them based

on that. Assignments were done using perplexity like you would do something called deep research in perplexity and you would generate an answer or report and then you would iterate on top of it. So that and then the professor told us that experiment was so so good and he wants to keep doing it more and more. That tells me that like you know the previous era of education system needs to completely change. Students should be rewarded for asking questions. Good questions should be given incentives.

We have come to a place where we give incentives to good questions, not to answer them.

Correct. Because AI can do better than us and it will keep getting better. Yes, we should obviously know and understand but the way you understand, there's a famous scientist called Richard Feynman and he says you do not truly understand something that you cannot create. So the way he came up with that is he said once somebody asked him, do you know about this bird? I said yeah, I know the name of this bird. This is so and so bird but then he said, yeah, I know the name of this bird. This is so-and-so bird.

But then he realized, okay, I only know the name of the bird, I don't know what that bird is. I don't know anything about its origins, its ancestors, like, you know, how it came to be, why it exists the way it is, none of that. So there's truly a difference between knowing the name of things and actually knowing the thing. And how do you get to know the thing?

You ask questions, you go deeper, you go deeper. So that's why our, you know, now the company in the product launch funding, like deeper is the mission of the company. But at some point you need to come up with it because company is not just meant for a product, it's meant for a purpose. When I came up with it, there was one quote I read, it's actually from an Indian person, Geetha Iyengar and she said,

knowledge has a beginning but no end. So that's why we made our our mission statement where knowledge begins. Perplexity is where knowledge begins. We think there is no end to it. So let's say you want to learn about artificial intelligence. You can ask like what is the latest in AI? What is this thing? What is this thing? You can keep going deeper and deeper and deeper.

And that's why like we think there is no limit to learning. And learning through questions instead of learning through lectures, instead of learning through like books, materials, that is a fundamental change. It's much more natural. And I'll tell you why, when I met PM Modi last year, he actually said this thing that I didn't even realize.

He said, when we are kids, how do we actually learn? We just ask questions. Correct. We don't read books. In fact, many parents have problems and they don't know how to answer the questions that their children ask. So, now there are tools like this. Obviously, we can do it. But, simple questions, there are deep physics, correct? So, why is the sky blue

or like why is the ocean blue? All these questions will be asked. Why is the sun in the morning time and not in the evening? Why is the sun in the moon white? When we answer all these questions, we will know our own limitations and our own knowledge. Einstein know? Yes. Einstein is usually says like if I don't know if it's truly his quote, but it's attributed to him that If you cannot explain it to your grandmom, you do not truly understand. So, not answers, but questions. As we hear more questions, our depth in the topic will come out and the simplest questions often have the deepest answers and the greatest discoveries in science and technology engineering

have all come from someone deeply questioning something

So, the one who asked questions will be the one who finds out the truth Why did this happen?

Newton was the authority of the subject in physics Newton's laws even now. But Einstein did not take it for granted that Newton's physics works in all scenarios and question what happens when you travel at the speed of light. That led to the invention of the theory of relativity, both general and special theory of relativity. And then people

could have taken Einstein for granted but they said okay what if we go the subatomic particle theory of relativity, both general and special theory of relativity and then people could

have taken Einstein for granted but they said okay what if we go the subatomic particle level, what if light is a wave or a particle that led to the invention of quantum physics which Einstein disregarded for a long time. There is a famous quote of God does not play dice with man because Aurobindo probabilistic nature of particles and he didn't believe in it but some people questioned it and disregard what he said, not in a disrespectful way but more like truly in a truth-seeking way.

So you have to always question things, you have to always like not assume that whatever is currently taught to you is the right way and that will only come, you can only do that if the system incentivizes you to do that and earlier it was difficult because you had a grading system, you wanted people to know, have the skills and the best way to judge for that is whether they knew the topic, knew the subject well but now you can do your work with the help of an AI

it will answer your question

So you can ask as many questions as you want

So why should we be in the old system? Why should you know everything? You should know something You can ask a good question only if you have the base knowledge But in every topic, we need to know the depth and detail. You along with AI is the new entity.

In that kind of world, our curiosity and creativity have limits. Because that is the one human skill that AI still cannot exhibit. Being curious. Being curious. Yes. Your perplexity, that is the meaning right? Correct. Curiosity. Correct. Because Einstein had those questions

and AI answered them, maybe we can find out the relativity theory very soon. But on the original spark of curiosity, what happens if we travel at the speed of light, that was his own imagination. That is the limit. That is what they say. If we go to AI, we can go to a bigger parameter like Maria Queston's cake law.

But on the initial thought of to go and ask a question about something, of being curious about something, that is deeply human and that and a lot of the things we humans have done art, poetry, writing, scientific discoveries have all been not because, oh I want to make money, I want to be famous, I want to like change the world. No, people were just fundamentally just expressing their curiosity in so many different ways. You're in the art industry, so you know a lot of the things you just don't do because oh I'll get a million views, 10 million views, you don't do that. You're just deeply curious about something.

And then because you're doing it in a way that's very authentic, people appreciate it and then you make a name for it. So I think that's sort of like why we we believe at least in our company that we should celebrate human curiosity. Everyone is afraid of AI. AI will do everything we do. How can we earn salary? Then what is that one advantage you have over the AI?

It's curiosity. So curiosity is the key. Yes. If you want to be successful in AI era, talent, skill, communication or curiosity, which one will you put first?

If you want to capitalize AI era, it is human curiosity. Talent is all relative. What are you talented? What am I talented? At some point remembering phone numbers was considered a measure of intelligence. At some point doing three digit multiplication with another three digit

in your head was considered a sign of intelligence. Why do you care about it now? There is a question, Arvind. The kids don't even remember their phone numbers. They don't know anything. They don't remember anything. Their brains are getting dry. They don't know why they are like this. How do you see this? Why is it necessary today?

If you ask a generation to tell the 16th opportunity, they are not going to tell. How do you see this? Artificial intelligence is thinking independently. It is thinking going to say. How do you see this? Artificial intelligence is thinking on its own. It has killed all of these.

How do you see that? Most of the famous scientists,

they have dyslexia. Einstein struggled with getting good marks in some exams. Some things he just ignored. Did it matter? We celebrate, we consider him a genius. His brain is still functioning.

He is still functioning. He is still functioning. getting good marks in some exams, like some things he just ignored. Did it matter? Like we celebrate, like we consider him the genius. His brain is still studied as like, how did it have more connectivity patterns than the rest? The brain will wire itself to do the tasks that you want to do it, right?

Of course, there's an intrinsic wiring that you're born with. Over time, there is an intrinsic wiring that you are born with. Over time there is some flexibility to adapt. The task that we have to be good at multiplication, memorize poems, remember social science textbooks, history textbooks, that was in our Marx exams. But now if we go and recite that again, about French Revolution, if we recite that again, it will not work. But where the details as

important as taking the core point of why things happen, no. Now if you go to Paris let's say, a tour guide will give you a better lesson of the history than your social science teacher would have. Because it's much more interactive, they take your questions, they encourage you to actually learn. So what is the best way to teach history? You have to take people to a place, walk them through and have them engage and ask questions and try to connect what happened there

to something in the present day that they could relate to. So that's why documentaries are hits. So it's all about the way you teach. It's less about like expecting current generation students to do the same thing we did because that's not a sign of being smart.

For example, a lot of the younger kids play Roblox, Minecraft. They build interesting things inside that world. I don't think I could have done that as a kid. So, the dimension of knowledge has changed now. more moving towards fluid intelligence. What is fluid intelligence is when you don't even know something ahead of time, how fast can you pick it up. IQ tests are there for a reason. IQ is a controversial topic, you know some people believe in it, some people don't. But truly what an IQ test is meant to measure is if you are given new patterns on the fly, and you quickly reason and make sense of things but if you practice 10,000 IQ tests and then take the 10,000 first obviously you will

do well that's not really fluid intelligence so what kids of the next generation should be good at is fluid intelligence okay okay now I'm going to teach you a new topic, it is not in the textbook. It is just a lecture. The exam will be after 60 minutes. Based on what I have taught. Who will score well in that? Assume no student has already read that before.

That is one example of fluid intelligence.

So, now it is fluid intelligence. The methods we have studied will not work anymore. We don't have to see it as a game.

Another idea I have had, sometimes I wonder, what if you incentivize a student to go teach the class. Like that is their, maybe 12 chapters in the book, that book is for the entire semester and each chapter will be taught by the student and they are graded based on how well the other students understood the topic. That's it.

Recently you met our Prime Minister Mr. Narendra Modi. What discussion happened?

Actually, I am very very impressed with how hard he works even at this age and he has been PM for many years. So, he can just sit back and relax. But no, he actually is still, I met him on a Saturday evening like 6 PM or so, still in his office and working. He said that kids when they grow up, they will ask questions to their parents in their

own language and they will ask through voice. Very natural way in which they learn. So he emphasized the importance of AIs to be able to answer in voice, the voice to voice experience and in the language that they grew up with and he emphasized how India has so many languages, so don't just focus on being good at like Hindi or like even dialects are important according to him not just the language. So, he knew about perplexity. So, he knew that the core value out of perplexity is

citations referencing the sources. So, number perplexity the key difference to other AISs. It will go and search the web, full resources and then whenever it gives you an answer, it will tell you where it was taken from. It will tell you where it was taken from. It will tell you every sentence. So, what is his interest is, lot of western literature, science, some of it was actually known even in

Ancient Indian texts like Aryabhata Bhaskaracharya. They all had really good Mathematical knowledge and so what even algebra was known to us So our kind of interest in a number if we could go and scan all the ancient Indian texts, digitizing it, indexing it, and seeing if there are correlations to what is in the western and Indian texts using this referencing technology. He was interested in that.

So he discussed that with you.

Yes, he discussed that. He also talked about the voice based multilingual. Then, his fear was that AI is very powerful. So,

his hope is we shouldn't become more lazy because of AI. Like if all the students are just using it to cheat on their homeworks or exams

That's what

Online exams, cheating on course certification, digital course certification When you work, you upload the PDF and answer it to AI We have a tendency to shortcut everything So how to avoid that? I told him the same Fundamentally you have to encourage curiosity how to avoid that? he asked so i told him the same

fundamentally you have to encourage curiosity they have to without wording the output of the work have them come up with new tasks new questions new projects

so that they don't feel excited about the actual filling of the forms working they feel excited about what other projects can be done So that AI can do all this But I can create new projects and create new value in the world

So if that curiosity comes, the cut and paste work will be lost You think they won't do it

Correct They will do it, but AI will do it But cut and paste will do it If an AI can do that work If we keep on saying that you can do it cut and paste If an AI can do that work, we can't say that you have to do it, you have to do it You are just fighting the inevitable

You need to actually think Now that AI can do all this better than you, what should you be doing? Why don't you take up new projects that you can use AI's and create more

revenue for the company upskill them upgrade them challenge them make them feel excited to come into work the reason they are delegating all this work into AI is because they don't like the work they're doing or they want to get it done faster and go back home and relax so now they are giving all this to AI for example a project, a detail, a pdf

they don't want to do all this so they are giving it because they want to, for example, a project, a detail, a PDF.

So you want the working environment to be more exciting than this?

Yes, we want to challenge and incentivize. So they are going to do it there. So beyond that, you think curiosity is the core If not, we can't do anything If we are not interested in anything fundamentally, our life will be useless

On the other hand, I feel that curiosity and artificial intelligence will stagnate I am asking N number of questions and N number of answers. If I ask you about my work, you keep asking me about the tabular column, how to do it, what design do you want. I keep asking you so many questions.

My curiosity never ends. I keep getting answers.

Major rewiring of the brain is that Earlier, we had a lot of questions The answers would take an hour or a few days Some people would go to a country to meet an expert to find answers to a question That has gone now. The tool you use is the same. Jensen Wong, the NVIDIA CEO uses the same tool, Perplexity. So there are no limits.

The one who is more curious, the one who uses it and asks more questions, they have the

advantage. So that is what we do in the product. After the answer, we suggest a few more questions to ask. We did that so that we help you ask the next question too. product learn the Pandora after the answer we suggest a few more questions to ask oh yeah we did that so that we help you ask the next question too we don't just give you the answer to us an answer is the beginning of the next

question that's why you say knowledge has a beginning but no one we are trying we are going to help you to ask more questions we're going to suggest new questions to ask there is an entire team in Perplexity whose only job is to improve the next set of suggested questions.

How? Okay. So, Perplexity is doing a great job of suggesting the next set of questions. Correct. Okay. So, you have the answers to that.

The answers will work the same way. But the questions still shouldn't be a human limitation. Even if you don't click on the questions we suggest, you will still think, okay, we can ask another question.

Yes, it's there.

Secondly, you will get an idea of what to ask next. That is a journey. I am talking from personal experience. Since I am mostly a CEO, I don't get the time to do technical work Before, if I wanted to learn a new subject, I would go to a YouTube video or listen to lectures There are a lot of online lectures, now I don't have time for that I don't get time for 1 hour or 1.5 hours I am not getting time

So what I do is I open the YouTube video There is a comment box for perplexity There is an assistant on the side Transcription Not only transcription, there is a transcript

You can also pull sources from other parts of the web Just basically Imagine, if you have perplexity on every website, how would it be? I would go to that and ask questions. What is this? What is that?

After I get a decent understanding, I would go to this part of the video and explain this to me. I would go to that part of the video and explain that to me. I would consume that video in a non-linear way. Without consuming it in a linear way That would be the major unlock for me

2 hours worth of content would have my context, what I need I would pull that, personalize it and get the answer I think that capability when I was preparing for IIT I had that power I capability to prepare for IIT

I was very good at preparing for it What happened at that time was I didn't understand certain concepts when I was studying physics and chemistry I would call my friends and ask them what I didn't understand It would take a long time to understand that and properly learn it. And how much can a professor interrupt in a lecture.

Especially if you are a top scoring student. If the question you get in a new lecture is a very simple one, you will feel ashamed. So with the AI there is no judgment whatever you want say no one's looking at you you can ask the simplest question and if the AI gives you an answer you still don't understand it you can keep asking keep asking that I truly

felt was wow if I had this that's a big freedom yeah it's so freeing. No one, why should anyone judge me for the questions I ask? So now, sometimes I want to go back and ask questions about simple physics stuff I might have forgotten or still not on top. I love it. These tools are amazing. If someone else asks you, you'll say, Arvind Srinivas doesn't know this. You can talk there. Particularly in the business, I am not a businessman. I am more a technologist, scientist.

But when you start a company, you have to learn how business works also. And who is there to teach you? I can't even study MBA. Do I have time for that? First of all, I think that system of work has to change. Doing a company itself is the best kind of MBA you can get. Start a company and build a product, get it in the hands of people,

and monetize it, make revenue, raise some capital. This is an MBA lesson for you. In fact, when I started the company, many people told me not to do this, you are a scientist, you are a researcher, find something new, just get a stable job and don't try to take risks.

Then one of my angel investors told me, this is a free MBA lesson for you. So, starting a company and running it. And sir, I have on the fly learning is like Fluid intelligence Taking a new task Yes, we don't know

But we can learn as we go And everything around us Someone did not know it already for the first time But they've figured out

It's interesting actually Because you are a tech guy and you are a scientist also, most of them get into technology, this thing and all. But from that, you become an entrepreneur, a tech giant.

I'm not a tech giant yet, but I understand what you're saying.

But building a company like that within a short notice How did you see that transformation happen within this short notice?

First of all, a lot of luck A lot of luck involved If I look at it, Google is worth 3 trillion 3 trillion is 3000 billion So relative to Google, we are not even 1% of it. So, one day Google's stock will move 1% or 2% up and down.

On this one day stock move, the entire company's valuation is so...

But you said you made it in a shorter time. 20 billion compared to this thing, compared to Google or NVIDIA is smaller.

Very small. I would say almost irrelevant.

But how did you create it in such a short time? You could have been working on a product or a model. But how did you do this transformation correctly?

I didn't do anything to get the valuation and market cap Even now, I don't know how far I can go from 20 billion But, what I feel is that you have to keep building a great product This is what Steve Jobs said Fundamentally, why do we have this company? Why do we set up offices and hire people?

It's okay if we get paid, but that's not the purpose of the company. Because to put in this much effort and build products and experience, you need to have motivation Intrinsic motivation I have that drive We need to build a product that we can use That is the core motivation

Since the beginning, I have been interested in asking questions Learning, very knowledge centric upbringing is more than money scholarly recognition is what my parents always encouraged so that is a

deep value there is no product like that actually, google is the only closest product you can say to something that you can actually just feel like it was built for learning. Wikipedia, a few products existed, but still learning on Google or Wikipedia

is still not easy. You can't ask questions, you can't do what you're doing on perplexity. And so every time I feel a limitation in whatever existing products there are, I try to think what can I take and solve there and make Perplexity do that. So, we took out the limitations of Google Chrome and added new things to it and built it. We did that in the Comet browser as well. What was your question? What made you want to create Perplexity? and how to make it a reality The first question is We started a company

and we had to search for social graphs and how to search for content on twitter who are Gopi's followers who are Gopi's followers and who are my followers we couldn't ask these questions

This was the first demo we built and after some funding we hired at least one engineer as many founders as we could After I built my business and got some funding, I needed to hire at least one engineer I didn't know how many founders I could hire So, the engineer asked me a question

How do I get health insurance? In America, it's not like India, where the company is the only employer I'm a company, I don't have any business experience I don't know anything about health insurance I didn't come to see it I am building a product

My wife is saying I need to get health insurance otherwise I won't sign the offer Ok I will figure it out Just give me 3 days and then I tried to ask about health insurance Can't find anything

because insurance is an ad category for me there are like 20 insurance providers who keep competing for that keyword so that their webpage gets ranked up and you click on it and buy their products but what I wanted to evaluate was what insurance should I take for employees

what are the plans what are the relevant things in the startup stage? This is my question. But I could not ask that question anywhere. At that time, there was no product. So, I am taking what is available, and I am finding out the answer to my question. Correct.

And I called the investors and asked for insurance. In the middle of the lecture, I was told not to think of make it a dumb question in the middle of the lecture So I didn't ask anyone Then I thought of a tool that we can use ourselves We can answer whatever question we get At least we can build that

First we need to build a product like ChatGPT We can answer the product like ChargeBT The question we ask will tell us what the AI model is

My question is the same There is a confusion here ChargeBT is already answering the questions On the other hand, I have a search engine called Google In between this, there is an answer engine It comes in a segment called AI answer engine

In what way is this different from Google and ChatGPT?

It's a marriage of Google and ChatGPT. Or Wikipedia and ChatGPT, whatever you want to call it. So, when we read Wikipedia, there is a footnote that principle and the chat gpt interface of chat gpt combined the data, whatever data

is there on the internet it will ingest using google's principle if you build a product that is perplexity so chat gpt has search

but it will never work not all the time it works reliably. If we ask chat GPT questions related to search, it will answer sometimes. But, it does not have the accuracy and reliability of perplexity.

How do you differentiate that? You think the advantage of having material perplexity from

many open sources. Correct. So, the main advantage is in the complexity is the advantage Correct The main advantage is real time knowledge AI model already has the knowledge But that knowledge is not real time There is a cut off date for that The amazing aspect of the human world is that new things are constantly happening

Something happens We all have questions about it In fact, most of the daily questions we get are about what is happening in the world We don't ask about what has already happened

We only ask about what is happening now

We only ask about what is happening now based on that We don't even need a reflexory to know about history we are asking about the history of the world AI models don't know about history even reflexology doesn't need it but we need to know the things that happened in the past accurately

AI models will make mistakes sometimes because the memorization of existing knowledge is not perfect it can never be perfect accuracy is needed and references are also important. Sometimes when we read a media article, we question if what they say is correct. We have a doubt.

Especially in a world where we can project fake content using AI, sources are still important. Otherwise we can't tell which is real and which is fake. So, sources, real time facts and accuracy. These three are timeless. Timeless and important, especially in a time where AI content is flooding the internet and social media.

It is so hard to know what is real and what is and fake it's so important to focus on these three things and some company needs to explicitly exist to focus on them the problem with the other chatbots is there's a thing in AI called hallucination hallucination means AI making up its own things. Yes, it will think of it and say something. It will say it as if it is true. Correct.

But, I am asking, isn't this hallucination issue in perplexity?

In perplexity, hallucination may be one out of hundred times, two out of hundred times, it will be there. But, by design, it is meant to fight it, not encourage it. By always pulling information from authoritative relevant sources with human knowledge and only using that to answer your question. It's meant to fight the hallucination problem whereas in other products like chai chitti

sometimes hallucination is good there like okay can you generate me ideas for a baby's name, can you write a poem for my dad, can you talk to me like a friend, I'm feeling really lonely. Only if you have imagination and hallucination, the product's feature will be good. So if the product has both hallucination and bug feature, it cannot do both that well. So what we thought was, hallucination being a bug on the use case on the search use case here like quite big

Like number of questions from one person can ask where accuracy is important is already a lot

So for hallucination is only bug

Yes Other use case would a friend perplexity is not meant to be a friend companion writing your poems essays

Can you define who is perplexity? Is it my friend or my teacher or companion or a soul mate?

It's meant to be a teacher perplexity fundamentally is meant to just be your knowledge guide. Knowledge guide? Yes. Okay. When you are trying to go to an expert to get help on something, you expect them to be truthful.

You expect them to have integrity in what they say. If they lack that and they just say things that you want to hear, then you won't trust them again. For us, like perplexity is only meant to be the most accurate. It'll still be friendly. My goal is not to build a product that feels very robotic.

So the answer we give should be correct. That's the most important thing. If we say it warmly, it's even better. But if we say it wrong, just because it pleases you, which we should not say incorrect things.

Yeah, that's what I am asking. Because, hallucination as a part, it gives a creative space to build on one side. On the other side, if I want a perfect answer, I will cut the hallucination. But again, if you say A, answer engine, now Google is putting A, A answer in the front page.

And Google is trying to do the same thing as us. What does perplexity do exactly? Google is copying it. How do I differentiate? Assume Google does, this is not an exact number, but 10 to 15 billion searches a day worldwide. A lot.

10 to 15 billion. 10 to 15 billion searches a day. Oru naalaki. Oru naalaki. Out of which 30 to 40% of those assume like 3 to 4 billion. No revenue, like just not even a single link is needed

like you can just literally give a answer knowledge card like for example if you ask Rajinikanth age it will just give you that number whatever 70s right there's no need to click on a link go and update so other layer of the world three to four billion queries will go there. And AI is expanding that 3 to 4 billion number to more where there is no need to click on anything and then there is lot of questions, there is lot of searches where there is no

revenue made. The way Google makes revenue is there are some searches with commercial intent, shopping, travel, legal like you are hiring a lawyer for something, local services, let's say you are hiring a plumber to come fix your thing, movers, in the money or a jewelry, in the Jewellery In these categories we when we search there is a business link

if we click on that link Google will charge revenue and that is called Adwords that product is called Adwords so the way they are able to monetize it is they say

everyone comes to Google for everything, all the questions. And as a result, we own all the search traffic. So even if they learn about a brand on Instagram, they come back to Google to actually go to the link. And as a result, we make money on every click. So even if 5% or 10% of the searches of 10 billion like 1 billion searches a day

they can monetize they are good so the key question then comes to like do they have an incentive to give you the right answer on those commercial queries no they have an incentive to sell you to sell your query to the person who bids the most on that so they just connect them correct there and they make money by giving you giving your query to the person who bids the most on that word so for example if you're trying to buy jewelry whoever pays the most for that ad word will be recommended and then you go there there's no if you ask a question on like

what is the best jacket to buy? What are the best shoes for me? I'm planning a vacation to Los Angeles. Which hotel should I stay in? Can you recommend me some local experiences to do around there? These kinds of queries don't have any incentive to give a good answer.

Those are all influenced answers.

Correct. Where does perplexity differentiate? Perplexity is not influenced by ads. Perplexity will just give you the right answer for you based on your past, whatever the AI understands about you. Perplexity will just use that, look up the information from the web and give you the relevant answer.

So, Perplexity won't give me an ad influenced answer. This is the USP. But, in the future, won't you go to ads?

We won't. We make money through subscriptions. We have some other ideas on how to monetize free users without having to influence the accuracy of the answer. Now, fundamentally, the problem with Google is that in the same interface

there is link clicking, answers, ads you won't get a consistent experience and as traffic grows, the Google's distribution advantage will decrease now maybe they have 10 billion a day, we have 70 million a day, there is a difference like that But as we grow to that billion range scale, the volume difference will decrease People's trust factor will matter in the long term

Which product do you really trust? Do you trust the product where revenues made through making you click or do you trust a product where revenues may be made based on how accurate the answer is?

The second one.

Second. So long term there is no misalignment for us. Whereas long term there is a misalignment for us. Whereas long term there is a misalignment for them. Now Google makes revenue in many different ways. YouTube is there, Google Cloud is there, so many different divisions, it's a giant. Like giant is going other than giant.

But over the revenue we make lower concept like margins, profit margins. So if you have three businesses, each of them are earning $100, but the first business, the profit margins are 90% that is you you only spend ten dollars for that hundred so you pocket back ninety and profits the other two businesses you

only profit twenty dollars thirty dollars so what matters more the first one. So search on the link based advertising almost hundred percent. So in that model you you are taking another call completely perplexed. In that model, you have to fight.

That's what you are fighting with the ad influence information versus accurate information.

Correct.

This is where you differentiate yourself.

Correct. And in the business, there is a concept called your margin is my opportunity. That is what Jeff Bezos said. Jeff Bezos is an Amazon accountant. margins my opportunity or a concept in enough in order to build a Profitable business, but with decent margins you can compete against someone in that same field with infinite margins almost hundred percent margins and Squeeze them bring them down. So that way they are forced to come down anyway.

So for you going from 0 to 20% margins is great. You are a money losing business. So making some profit is already great for you. But for them coming down from 90 to like 30 is very bad. So that's where you have an asymmetric advantage. If a small example now cameras Kodak used to be a company produce real cameras but then got disrupted because you can put it on your phone for someone to not charge anything for

putting a camera on the phone is no-brainer because they going for that margins that Kodak is losing. Earlier they would sell encyclopedias and real CDs then you could in actual book cracks like shelf of books. Wikipedia just gave away all that margins for free. So for you charging a little bit margins on something that someone else is doing for infinite margins is a no brainer.

Right now perplexity is not even profitable so for us to go and do these commercial queries correctly and even charging a little bit for like fulfilling the transaction or subscription fee our margins will be always lower than Google search advertising margins but because of that we can be very aggressive in growing okay of that we can be very aggressive in growing. Okay. We can we can be very aggressive in building this new experience they cannot because for them if they copy us on

commercial queries, their margins will be less. So they can only keep inconsistent experience. That is a business limitation. So, there is a phrase, victim of your own success. If it is not there, you cannot compete in this field. You cannot compete with them. So, this is a structural disadvantage.

They know it. So, they cannot move that fast. And they know that this is the future too. Like 5 years from now, you're just gonna tell an agent to go buy things for you. You're not actually even gonna shop the traditional way.

They know that too. But we can move faster there to get there than them. And when we get there, they'll also be relevant in that space. They're very, you know, it's a technology company at the end, but it won't

have the 90% monopoly share it has in search now. They have 90% of the search traffic. So let's say on an average every day, total number of searches, outside Google also. So 90% of the search traffic is from Google.

Yes.

So Google 15 billion, 90% remaining 10% is with Google Now, how much is with other companies? Is it still in the desktop? So, Google 15 billion, 90% means remaining 10% is 1.5 So, total is like 17 billion searches 15 billion will go to Google

So, AI queries let's say 5 years from now, that will also be 20 billion, 30 billion that 90% will not go to Google. There will be more players.

There will be a different business model. There will be a lower margin business model. And so they cannot play this monopoly game there. And we want to capture a good chunk of that market. And there is lot of use cases of AI. There will be companionship, friendship, writing, essays,

coding, so many models, many products, and so over a part of the experience, Silape will be the best, there'll be different kinds of monetization, subscription, usage-based monetization, agents so on the world on the Google will be one of the players, not the only player. And that itself is a massive change from how the search market played out.

Is perplexity trying to break the monopoly of Google?

Yes.

Okay. I think your mom loves it. She always wants you to work for Google and she will be very happy that you are contesting

them or challenging them. I was an intern there.

I know.

I know.

I know.

I know. I know. I know. Google has like, the way I think about it is like, there's Alphabet, that's the parent company, it has Google search and so many other things. Everything. I'm not going after the other things. Like for example, Android, YouTube, Google cloud, Waymo. So their company is a real miracle.

That's an emperor.

I am going after the Google search market. Basically, search is becoming AI. And assistance. And so, they are also going to compete there through a product called Gemini. Perplexity and Claude. Grock, there are many other players.

There are many players. Each one of them targets a specific area. If Claude targets you, Grok seems to concentrate more on your reasoning. Whereas, Chad Gpt, Chad Gpt 5, Gemini. Who has the biggest advantage?

Temporarily, the advantage is mainly Chad Gpt. He is the first mover big brand everyone associates AI with chat GPT is an association and I think 65% market share

so I think the current market share is 60% so there are like daily active users only measure it or how many queries a day so perplexity

like around 60 to 70 million daily queries chanchipiti is around one and a half to two billion so billion yeah okay so Gemini is somewhere in middle between the two I think these are the three main in terms of widespread adoption but growth of chanchipiti is slowing down obviously they got there first and they grew really fast. Others are still growing, Gemini and Perplexity

are still growing. And what kind of use cases go to different, AIs are also different. Like Gemini grew a lot because of the nano banana thing, you know, image editing.

Yeah, image editing.

And I think Chantjeepti also had that Studio Ghibli moment and people use Chats GP for writing code. There's a lot of different use cases. Our goal is when they're searching for information they come to us. Anything related to accuracy and search and information and knowledge, research, discovery, they come to complexity. If you go and do image editing, writing, code generation in other AIs, completely reasonable. We don't want everything.

You don't have such ideas. I don't plan to come to all the segments.

Product is designed in a way where it can be helpful for everything.

Okay.

But we don't need to win at everything.

So what's your core intention then? But we don't need to win at everything.

So what's your core intention then?

Core intention is to support research and knowledge and discovery. And I don't think it's a small market. No, it's a big one. Knowledge discovery is a big one. It's as big or even bigger than Google search in our opinion. And we want to just go for that and build agents so agents that can so when you hire when you when you hire a person in your team, you would expect them to know how to

use a web browser you would expect them to know how to use mobile apps internet search engine you will just take it for granted AI's don't have that skill entirely today. And there needs to be one company that focuses and nails that. So, that's why we built the Comet browser.

Yeah, what's the purpose of, what's the idea behind the Comet browser? Because, there's already Chrome, Firefox, there are some other also, is there? Yeah.

So, why did you to have a comet browser? So comet browser is not meant to be yet another browser. There will be an agent capability. Now you can go and tell comet, search all the books that have been recommended by so and so and put it on my Amazon cart and place it for me Oh, it will work like an agent Correct and why are we able to build this, Perplexity able to build this is because Perplexity as a product was designed to use the internet all the time to pull the sources a natural extension of it is to use the internet to not just pull sources

but to take actions also like okay, from the webpage you read, click on the buttons, fill in the input entry fields, go from a sequence of thoughts to a sequence of actions that is the main difference if we build this, we can build an agent If you have an assistant, what tasks will you give them?

Maintaining your calendar, maintaining your email, ordering groceries, shopping, booking hotels, reservation for restaurants If an AI has to do all these, it needs to have the ability to use the browser

But an agentic AI already does this

An agentic AI needs to have a browser

So this is like an agentic AI browser for everyone

Correct Without the browser functionality, you cannot build an agentic AI You have to own the browser You have to be available for the user wherever they are so if you are on youtube

and you are watching a video and you want to modify the transcript and email it to your team you need to have an AI already embedded in the browser similarly

if you are writing a document and you are creating an event and you have a list of attendees You have to send an email to each one of them

and give them invites and manage it These actions are done by the government browser itself For example, there is an interview with Arvind Srinivas Now, in an IT office, if you send the employees to send the transcriptions they will take it from the YouTube video

translate it in English transcribe it and send it to everyone This much of processing can be done by a web browser

Wow! How can perplexity be correct 80% of the time 3 years ago, and 20% of the time there were mistakes, the same is the case with Comet now.

Yes, it will evolve in the next stage.

What we do is, at the end of the decade, if there is a personal agent, there should be a browser like Comet. Whether we use that browser or not, whether we use the browser itself or are we just giving voice task? That is irrelevant but you need to have the infrastructure of a browsing agent to go use all these different tools, websites, fill up forms, upload documents, ingest documents, control your email, your calendar, personalize to you You need a personal AI

So, this is what you are trying to fill in

Yes, and that is the next step There is a chatbot game, different people, different market share, all that is ok But, when an agent can do anything you want, chat feels like legacy. AI is moving so fast, chat was invented 3 years ago. But I am telling you that 3 years from now, chat will feel old. And agent, background agents that can, you have a cloud, you have a personal cloud, you

have a bunch of agents, they will be doing your work and life. You will have a personal cloud or agents or agent force. You can deploy that.

So, when Comet is positioned correctly, down the line, in 3-4 years, this will be the pioneer?

Yes. And by the way, Google will also do it with Chrome.

Yeah.

This is again where the business model problem. I can go to Comet and say, book a hotel that I like. the business model problem now in the comet kit up way you know pretty sure hotel book one of the insolent along and I learned the advertisers long ignore penny you're looking to hold a suitable of San Francisco the book funny that I am budget I've been so long pono on a Google London the pun now advertisement travel category for advertisement revenue last shrink that's why the number 2 advertiser on google is booking.com tens of billions of dollars spent a year

almost 80-90% margins

so again authenticity and accuracy

this is the perplexity authenticity, accuracy and personalized AI and how to monetize that if an agent saves 100 dollars, would you be willing to give 10 dollars back to the agent?

Okay, that's it.

Build a different economy. Don't try to go... Same format of ad based. Yes.

So, in a subscription based?

Subscription usage based. Okay. I have booked a flight, but I am going to sleep, so I cannot decide on a cheaper flight than that. So I give that task to the agent and sleep. When I sleep, if the flight price is cheaper than the destination, I can book it at the right time, change my seat, check in and order a taxi, how convenient it would be in the morning. Simple. I will pay only service tax. Our philosophy is in the concept of search like a billionaire, browse like a billionaire. You can use the same iPhone as Donald Trump, but Donald Trump's life, his, his team is there to set up his life.

They are not there.

So, can we build that now?

You can build it with these products. If you build it and give an affordable price to everyone, their life will be so much easier. Everyone says, if I get one more hour every day, I can spend extra time with my family,

I can spend extra time with my family, I can spend extra time with my kids, I have work, I can pay mortgage, I can pay their kids' tuition. But how can we make it easier? We can make it easier with tools like Comet.

And your mentor, even today, Sathya Nadella, in fact, she is your co-pilot. So, on one hand he is your competitor, but still he is your mentor and you talk about him often. How does that combination work?

He has given me a lot of advice. I have listened to him for many things. I seek his advice proactively myself. He also calls me now and then and he's a learner. He's a learning machine I would say. Even at this age how much success he's had in the last 10 years since becoming the Microsoft CEO. He's still at it like as if he wants to win and having that day zero mentality and so even now he's very technical like I ask questions about

okay what do you think about this model, what do you think about that model, what do you think about this eval score, this model guard, is it actually real? How is it working in your product? He exchanges notes a lot. And calls are on weekends, so there's no time to zone out for him. He just keeps going. And so that's what I really admire.

I think the mentality to still keep going and feel like a learning machine at this age and how much of a success he has had

So he is a learning machine in your perspective Yes Very inspiring actually In Indian entrepreneurs or in the technical industry, who do you pay attention to?

So I met Mukesh Ambani once when I came to the Bay Area He gave me a quote that was very impactful. He said, content is king but distribution is God. Content is king, distribution is God. In the sense, if you have a great product, it's really good, you're a king. But the one who knows how to distribute the product and actually get all the users, they control you still.

So essentially the takeaways, distribution is extremely important. And that was a time I think we had, we grew a lot since that time, a number of users, everything multiplied by two to three X since then.

But I took, I already was aware that distribution is important. I completely get it. But the way he said it made me realize, okay, I need to just live and breathe distribution. And I think that was very useful to me.

So, and other than that, I like different entrepreneurs text me on Twitter, like Indian entrepreneurs. I meet them sometimes. I met the Paytm CEO Vijay Shrekhshan. We did some partnership together. He's also very, very nice.

He's very supportive of me. So almost all of them have said great things even Sridhar Vembu texted me saying he likes perplexity and good yeah okay yeah so I think I have generally like you know you're in the Indian entrepreneurs or a touchy touchy like a silly like another local lawyer three other but I met and then we have a big partnership with Airtel.

Yeah, that's what I was going to ask. Gopal Mittal, we had a talk and then I talked to his son also.

What is the combination of Airtel and Perplexity? Airtel is a company that provides perplexity to 38 crore households for free. What is that?

content is king distribution is god so why did Jio even get market share they came and gave away the technology for free got a lot of people provided amazing service and then charged

so that's what we want to do too we need someone for that we don't have a residence in India already. So we partnered with Airtel. So they take our Perplexity Pro, which is $20 a month, so 17,000 rupees a month, and give it away for free.

And you'll be like, oh, we're losing so much revenue. It's okay. First, give people an amazing experience. Get them habituated to this new habit. Currently, the habit is going to Google and typing keywords.

Let them get habituated to asking questions here and give it away for a year. It's on you now to provide good service. And after a year, you can come up with a new pricing model for India that will work for India based on how the usage is how much you spend per query we have one year to gather in the free one the other one year go

right yeah free yeah so the the pro interview so India specific pricing okay could be five dollars a month four dollars4 a month, I don't know. And we'll still keep the service, there will be a version of perplexity, it will be free, but we'll upsell some of those people to convert and then we'll share some of that revenue with Airtel, that is the idea. And it's a very successful, probably the most successful partnership we have done so far in terms of

how many people availed the offer and continuing to use it. Pushed us to the top in the App Store and the Play Store. Like today we are the number one app in both Indian App Store and Play Store.

Do they use the word, A, correctly in India? How do you see it? Because you said that instead of typing keywords, we are going to make them into question marks. So we move from here to there. But do Indian customers or Indians

capitalise on artificial intelligence?

Yes.

How?

We don't have a tool to ask questions. But Indians are deeply curious They are already very curious That's why we use social media We use a lot of tools without knowing how to channelize YouTube, Instagram, Twitter

But we can't ask questions through any of those tools We can only consume For the first time, you have a tool like this You can ask any question Accurate answers will keep coming That is the main change

Do you think this is a blend for Indian nature?

100% Indian usage is very high in terms of number of daily queries, number of users and the growth is on since Airtel partnership was announced not only Airtel customers are using it

but many other people are starting to adopt Perplexity through that marketing they did and many people started using Perplexity and Word of Mouth growth started So, mobile app usage is very high in India So, that is the main difference to America In America, we have mobile app usage

and laptop usage But in India, mobile usage is predominantly highest So, Android, mobile. And, how important is this is that Android is Google controlled. So, Gemini is always there. You don't even need to install the app.

It will come automatically. So, when you have a lock-in system, a distribution advantage like that, we compete with Google with such a partnership. Otherwise, they will always keep on...

They have an upper hand on it.

Correct.

And so, that's why the phrase, content is getting distributional, that partnership...

Oh, okay.

Actually, how does your brain capture it immediately? I see a lot of this. From what Einstein said, to what Ambani said, to what the Prime Minister said Because to grab this thing and execute it immediately

Is it that your mind is set in a conscious way over the flow? Or is it your default wiring?

After practicing a few things, the job of a CEO is One day you will get information from different sides of the company So your job is to focus on the signal to noise What is the most important thing? You have to execute it You have to ignore everything else

There is no chance that everything in a company is perfect A lot of things be suboptimal Ignoring these suboptimalities and focusing on the most important task is a non-trivial way of dealing with it If you ask a person what they are working on they will tell you 5 things.

They will not tell you even 1 thing. If you ask them to pick one out of those 5 things, most of the people will fail at it. That's why they need a manager to tell them that one most important thing. Those who know that first most important thing don't even need to be managed. So, when you are a CEO you have no manager so you have to at least be good at that, picking that one thing.

So, you don't have to do everything right. The right thing and the important thing, you have to do it right.

You have to be decisive. Actually, this is one thing. I am saying this because you mentioned it in Sathyanand. This is one thing he is very good at. Everyone in Microsoft would tell you that he is very decisive. So why can't most people be decisive because there is a cost to mistakes. Especially when you are at a higher and higher level.

There is a cost to making mistakes. You will be judged, you will be criticized. So you need to have the courage to make incorrect decisions also. Okay. Damages were a lot, but that shouldn't be a reason to not take a decision and just keep oh I do Okay, so under and the courage you know correct up on a low. I'm ready to take a corner So on the courage of your money, you know the very of try bono, okay? It is almost important thing so they are just like go do it

Okay, I was almost okay number mistake one of the course correct one of the quickly. Like the one the other dog. It is a decision. I mean, I mean, date data driven on the number Martin.

And Elon Musk is very good at that. He takes a lot of decisions. Some people think he's erratic and all that. It doesn't matter. When he is wrong, he quickly course corrects. Okay. He corrects and comes back. So, you should not be like that without taking a decision.

Yes.

Okay.

You have learnt a lot in your childhood. Even today you visit Phil's Cafe. Phil's Cafe.

Where you actually learnt AI.

I go there from time to time. Oh, you still go to Phil's Cafe? Philz Cafe? Where you actually learn TA? I go there from time to time

Oh, you still go to Philz Cafe?

Starbucks, Philz Coffee, Beats There are a lot of places

You were studying a lot while studying at Philz Cafe?

Yes, yes Because it was near the house where we lived in Berkeley And I didn't have the lab space When I came here, I didn't have a dedicated advisor Every student would seek an advisor That advisor didn't have a slot

He said, you prove yourself and then you get a slot So I just went to the cafe and worked day and night

You worked there?

I would go there at 5.30 in the morning. My first lecture was at 10 o'clock. I worked from 5.30 to 10. I attended lectures from 10 to 1. I attended lectures from 10 to 1. I went to the campus library from 1.30 to 5.00.

I went to the campus library from 1.30 to 5.00. I went to the campus library from 5.30 to 8 and then go home and sleep I did that for the entire semester I did that, took the top grades in the courses and published a paper That's how I got a job in the lab The senior past student in the lab was one of the OpenAI co-founders He gave me an internship offer at OpenAI.

And that's what gave me the exposure to... This was in 2018, before it was a non-profit research lab. Yeah, it started like that in the beginning. Actually, in America, students face some difficulties in America. How did you cope up with such difficulties? Managing finances, getting a guide...

My memory is, I first opened an account in Bank of America. We get a salary debit only if we open an account, we get a salary debit from the university. So I asked for a credit card. They didn't have a credit history. Social security and a credit history is what they give you. I didn't have a credit history, so they didn't give me a credit card. But what happened was The rent was

1500 dollars That is what I paid Together, I was with a roommate The total rent was 3000 dollars a month So I shared 1500 dollars The salary that Berkel gave me

was 2500 dollars So If I lost the rent, I only had 1000 dollars left 1000 dollars per month is 30 dollars a day So, if I lose my rent, I only have $1000 left. $1000 per month is $30 a day. So $30 a day.

The first time I went to a restaurant and had lunch, the bill was $25. So, if the lunch itself is $25, I have to pay for dinner and lunch. So, what I did at that time was I had saved my stipends for my internship in my undergraduate So, I transferred that to my American account and used that to get my credit card

After getting that credit card So, credit card is what we can pay everything in a one month delay So, I used that credit card was the only way to pay everything in a one month delay So, I used the credit card to pay the deficit and the credit card bill for the next month That's what I did Then, in the OpenAI internship, I got $10,000 as my first month's salary So, the day I made my debut, I went to the Union Street Apple store and bought my first iPhone.

You didn't have an iPhone until then? No, I didn't. I bought my first iPhone, the iPhone X. I got stability only after buying that.

Everyone in your family has from a finance background right?

My father is a chartered accountant My mother works at the Providence Fund

So you are the only one in your family who is from this side

Engineering, Technology, Computer Science STEM side I was studying technology, computer science, STEM

What did your family say when you started your company?

My mother took care of the risk factors I started the company after working for a year So, even if I don't do well for a year or two You have a reserve. Yes, you have a reserve. So you can manage with that.

But to take that risk, you need a mindset. So I always… A PhD is a risk. After you finish your undergraduate, you can go get a job. And you can work at Google and keep earning money. In fact, my friends would tell me back then,

oh, by the time you finish PhD, we will have got three levels of promotions here. You would come join us an entry level employee after PhD. The Google leveling system, undergraduate company, you can join at L3 and then you get three promos in a period of five years.

You'll be at L6. But after PhD, if you come and work at Google, you will join at L4. So they would make fun saying, oh, you would join at L4. By the time you join, we'll be at L6. You'll be earning a lot more money.

You'll just wasted five years. That's how they would talk about PhD. But if I hadn't done a PhD, I wouldn't have had the exposure, the technical depth in AI, the understanding of AI, the access and exposure I got by doing internship at OpenAI DeepMind. My whole trajectory is defined by all the decisions and the risks I took back then. So you want to take the right risk at the right time?

Yes, but you will never know the timing like I worked in AI since 2014 Even back from back in the days that I idea started doing research in 2015 Not because I knew in 2022 AI boom will take off Because it was just interesting so far the right the right advice is follow your heart.

That's the most important. Do what you like. It can be open at any time. I read somewhere that you said that Indians should create many organizations. We should not leave even after managing the organizations.

What do you want to say in that? mainly the mentality if we keep telling our kids our next generation will get a job in this company

we will be constantly stuck in that mentality so we won't have the entrepreneurial mindset we won't have the entrepreneurial mindset, we won't have the creativity to go try new things. I told you in the case, imagine there's a world where comet can just do things,

you're gonna you can just summon delegate tasks, so that is you yourself should be like an entrepreneur. And we all have that like we are you know there so many, Indian population is so big, a lot of families with lot of kids and they don't have the finances that people in the western world have. Still they manage. So we are very good at taking limited resources, building families, managing households. That is our natural strength.

Yes, it is already there. So we have to exhibit that. We have to exhibit that in actual value creation. We should not do it only in money saving. And similarly, we should not go out of the conservative mindset

and take some risk and borrow some capital and create something new and create value and recover the money spent. That venture capital mindset should come into India

and more companies should be created. And you should not fantasize just going and climbing the ladder in an existing company. You should dream about creating new companies, new products, global products, not just Indian products. And that's how you put India on the map.

You don't think there with a foundation model?

China is doing big, US.

Only two countries though.

China and US.

Largely.

Largely, these two countries are coming up with foundation models. There are many Indians behind that.

There are.

Not only in China, but in India as well. Why don't we have a foundation model for India's power and strength?

Investment is not there. Someone needs to invest. OpenAI was bankrolled by Elon Musk in the beginning, and then Microsoft, and then now a lot of venture capital. There are Indian labs, but the scale of investment is completely different.

So, tech business. Yes, compute, energy, data center, chips, models, research talent, the salaries they get paid, another scale. Now China created DeepSeek with much lower resources but not minuscule either Like they definitely have 50,000 GPUs that costs a lot of money. Most of the model labs in China are spending billions of dollars a year on running the labs.

So if you want to be a serious player, you should also be willing to spend that much capital. So someone has to come and finance it, either like Reliance has to finance it, or like Dhani Industries has to finance it, the government has to finance it, local venture capital should finance it. And so that is the most important thing.

Arvind Srinivas, you bid for Chrono for 35 billion dollars.

34.5 billion.

What is the purpose of that? You would have been planning to commit yourself. 35.5 34.5 What is the purpose of that? You would have been planning to commit You have an answer engine Why did you buy Chrome from Google?

You know at that time there was a case going on between Google and the Department of Justice and there was an expectation that the Department of Justice would

force Google to divest that is sell Chrome

Yes, to break the monopoly

So we said one of the defense arguments Google used to prevent that was to say that is no reliance, there is no buyer nobody can actually buy this and run it So Google said that they can't buy it,, they said there is no one to buy it. There is no credible non monopoly buyer, Microsoft can buy it, Meta can buy it, but these are all monopolies and they don't know. So there is no non monopoly buyers argument they used.

So we said no that's not true, there is a non monopoly buyer that is us. We will have the financing and we will keep, we're not gonna own Chrome and change it to Comet and become monopoly ourselves. We'll keep it as is. We'll run it in a neutral way. We'll keep the Google search engine

and then we'll improve the browser. So this was a way for us to tell Google and DOJ that no, your argument is not correct. If DOJ rules that Chrome should still remain with Google, let it be. But if they rule that Chrome should be divested, then we are a credible buyer. So it was an opportunistic attempt to encourage more competition. Otherwise they're just going to keep ruling the place.

Yeah. So in a turn, a $20 billion company, a $35 billion company.

That's happened in the past, by the way. People have done leveraged buyouts. You can monetize. You can easily monetize Chrome by just putting Google Search as the default search, having their ads run on on it and then have them share revenue with you Just like how they share so the way Apple monetizes Safari browser is Google pays

Apple 20 billion dollars a year. Oh Because all the ads Revenue that's made on Safari Apple gets a share of it. How good. If this situation arises again in the future, Arvind Srinivas will try to buy Chrome.

100%.

100%? Yes.

Okay. You are a better eye than Google.

No, competition is important.

Yes, simple.

Yes.

Competition is important.

I love this attitude. Okay. Because I am here and you are transparent on that. Their value is different and mine is different. But still I am trying to be...

Yeah, why not?

Why not?

Nobody else can do it because everyone else including OpenAI they will consider it as a monopoly because

OpenAI is part almost 30-50% owned by Microsoft right now. Okay. And basically, only on Chromium, Perplexity is being constructed. Google's engine on Chromium. So, many of us are being built on those open sources. Open sources have democratized, so that I can build a product, a tool. But why is Perplexity not an open source?

Open source projects are done only by developing an ecosystem We should not just do open source So for us, Chromium is an open source project But Chrome and Chromium are not the same Chrome has a lot of proprietary features Comet is also a Chromium engine product

But there are proprietary features Microsoft Edge browser is also a Chromium engine product But proprietary features So we will contribute to the Chromium project Just like Google does And that is the important part

we will not necessarily have to make Comet open source neither is Chrome open source yeah Chrome is not open source exactly so that we are open sourcing a lot of things actually in them perplexity inference engine or an area kernels within our game video the joint we have open do a lot of things with the inference engine and kernels

Jeff Bezos has done Datab, Databricks has done it, NVIDIA has done it. So you will have GPS, no problem. And so this is a good combination. Age wise, a company with a slightly smaller age, how do such big players come in and invest?

Because the product is good. I think they are seeing the attraction, the product. So all the CEOs of all these, like Jensen, Bezos, Databricks, Aligot, see they are all users of perplexity themselves so they tend to like come and see like most of the investments you got is because people use the product, they like it, they are automatically becoming users of it and as a result they are backing this. Now Google has said that they will give only 10 results per page, because they are controlling

the data, probably they need maximum data for Gemini, or many SEOs are using maximum data. In what way will this affect Perplexity? We are not affected by that, because we are building our own index. Perplexity is building its own index. We announced that we have our own search API,plexity is building its own index, we announced that

we have our own search API and developers can actually use our search API too and so our index is already like growing very fast to 100 billion. So you don't depend Google? Not overly, we look at ranking signals of different search engines to compare our thing and keep improving but we don't heavily rely on the Google search index.

These are some personal questions that only if you cricket scoring love by grandma interest on this that I know about it. Would I favorite cricketer here?

Sachin he actually messaged me once and I got I got introduced him message me once and he he yeah I was very surprised.

Oh, okay. Sachin's favourite. Do you watch Tamil cinema?

Occasionally, yeah.

Occasionally. Who's your favourite actor?

Rajinikanth.

Rajinikanth, okay. Why Rajinikanth?

Style.

Style? Okay. You asked about him the moment you came. Because you referred to what he said about me. That's why I asked you. What is your most favourite spot in Chennai? Where do you visit when you come here?

My home.

Home only?

Where do you like to go when you go out? Marina beach, IIT campus.

So you have been to IIT campus now and then?

Yeah, every time I go, I check it out. Was your mother there or father? Mother. Yeah, every time I go check it out once. Okay, so what's the ultimate aim of Aravind Srinivas? To democratize knowledge. Democratize knowledge. In business perspective? Create a company, already created that, make the company give the life that very few people have access to, that is billionaires, and give it to everybody. One piece of advice to the students, especially those who study in India, especially those

who study in the tech industry, one piece of advice.

Stay hungry, stay curious.

How do you make the people who are working midway relevant? Because they have a small confusion that the technology is changing. So if they are working as an engineer, a software engineer, what would you say to them?

Adapt fast. Definitely, the world is changing. So, adapt really quickly.

Superb, man!

It's awesome! Of course, chasing is really worth it. It's really worth it to continue studying Arvind Srinivas. Thank you, Arvind. Thank you Aravind. Many questions, many answers.

You started by saying that questions are more important than answers. You gave many answers and raised many questions. This interview will be useful for many people. Thanks for your time. We have been talking for a long time and now we have time. Best wishes. Good luck.

Thank you. Thank you. I am Aravind Srinivas. Title sponsors Appasamy Real Estates Dr. Sri Varma's Wellness Supported by NexCura powered by

Gen AI Healthcare Vishar Gurukulam powered by Pissupilladha Sunland Danish Ahmed College of Engineering Danish Ahmed College of Engineering

Stratos Alteza on OMR by Appaswamy Real Estates the Thank you.