Perplexity CEO: AI Is About to Change the Internet Forever | Aravind Srinivas

0:00This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales, using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify. This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks.

0:45Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses set up required, compatibility and availability varies 18+. so if you were 18 today in the us and had to start all over again would you still do a bachelor's phd realistically it's gone this is erevin ceo and founder of perplexity a 20 billion dollar company challenging giants like open ai google and meta they just released comet the first ai browser that thanks and act for you it also understands you your needs your budget

1:19Aravind Srinivas:everything. It has memory. I tried it myself and wow, it can book meetings, reply to emails, research products, even shop online, all without me. Amazon has AI to do the recommendation ranking. Google has AI to do the search ranking. So they can influence things and manipulate you to buying stuff. But for the first time, you have an AI in your hands. It actually protects you. It actually gives you power against the big tech. But while AI gives you unprecedented power, it's also reshaping the world of work faster than anyone expected. And it looks like those entry-level jobs are disappearing, right?

1:52What do you think is going to happen?

1:53Aravind Srinivas:I certainly think people need to push further. So is there still opportunity or is it time to rethink everything? Hello everyone, welcome to Silicon Valley Girl. It is the first time that I have a repeated guest on my podcast and your progress has been tremendous. We did a podcast with Aravind, the founder of Purple Lake City, in November 2023. The company's valuation back then was$500 million. Now it's around... Actually, we were just closing our C++ around then. So it was more like$150 million and we're just about to close around at$520 million. Okay. And now it's$20 billion? Yeah, roughly. Okay.

2:34Can you describe this period of time in one word?

2:36Aravind Srinivas:Exponential. Exponential? Yeah. What do you think has played a tremendous role in this exponential growth? I think our constant urge for shipping fast, yet maintaining the quality, our relentlessness. I love the word relentless. In fact, I wanted to buy the domain relentless.com, except I figured it was owned by Jeff Bezos. Really? So if you actually type relentless.com on your browser, it will redirect to amazon.com. He's an investor in our company. That's like the spirit of the company. We don't ever give up. We are very curious. We constantly keep asking questions on how to improve the product.

3:20Aravind Srinivas:The first thing I do when I wake up is read everything people, all the users saying on different social platforms. I make sure most of the bugs are immediately attended to. And you may think like, oh, what really comes out of this process of waking up every day and just bug fixing and triaging and trying to identify places for improvement. But we really believe in the mantra of like 1.01 to the power of 365 is 37.78. Can you explain? Okay, so if you do a 1 % improvement every day, how much do you improve at the end of the year? You improve 3 ,700%. You don't improve like 1 % multiplied by 365.

4:01Aravind Srinivas:That's the concept of the exponential. It compounds. And so it may look like small steps every day. But as the changes propagate to a larger scale of users, they bring in more users through word of mouth. And then those users are experiencing all the improvements you're making. People feel great using a product that you used last week and then use it a couple of weeks later. And it feels much more different, much faster, much cleaner, much more accurate. So then people start trusting. We earn the user's trust through our accuracy, our speed of improvement. And your marketing. Formula One, the Wong.

4:42Have you seen the spikes after Formula One and when the episode went live?

4:46Aravind Srinivas:It's hard to attribute to these kind of brand partnerships. It's very easy to attribute. You run some ads on Instagram or YouTube, and there's a clear call-to-action button in terms of going to the App Store or Play Store and getting an install, and then looking at like how many queries that translated to or whether that user ended up subscribing to the app. It's very easy to track this and optimize the funnels. It's much harder to optimize, okay, like if Lewis Hamilton has perplexity on his helmet, how many people are actually going and installing the app after a Formula One race? Yeah. It's pretty difficult to track that.

5:20It is.

5:21Aravind Srinivas:Best way to track that is like brand awareness. Like, oh, like are people in the UK or people who typically are in the Formula One crowd, are they more aware of perplexity now? and I think that should be the case. We have run some basic studies but the core idea there is honestly to associate our brand with some of the greatest people of all time. You know like Steve Jobs ran this famous Think Different campaign, right? He got the inspiration from Nike because Nike celebrated great athletes so that's why in the ad you see Einstein, you see Gandhi, you see all these like you know the Beatles, you see all the people who really inspired Steve and so that's one thing we've been pretty intentional about like who we want to associate with in terms of the brand like at least in sports Lewis Hamilton is one of the greatest of all time it's considered the greatest of all time in Formula One and so that's why we did that brand partnership and we're also going to be doing more like particularly you know an F1 it's a pretty cerebral sport like everyone's attracted by all the hospitality and like the race cars and like it's a very high adrenaline thing, but fundamentally the sport is largely engineering-oriented.

6:33Aravind Srinivas:Who has the best engine? Who's licensing whose engines? I think it's all about constructors competing for who has the best cars. And skills still matters. Conditions matters. Tires matters. Everything matters. So we do see a lot of people asking these kind of questions. There was this huge spike in queries in June compared to May. Can you attribute that to something? What happened? I don't know if it happened in June or July, but definitely we grew a lot in the last couple of months. Actually, the main event that happened in July that I would say is fundamentally led to growth is like our release of our agentic browser, Comet.

7:11Aravind Srinivas:For the first time, you could have a browser that can actually think with you and take actions on your behalf and like actually go and do stuff. I have Comet right here. Yeah. Can you show one killer feature that you think everyone should start using in common? I love this thing because it's a very simple thing where it kind of always works. You know, if you want to go to, there's some video I remember I watched before, but I don't know exactly what it is. And I can search for it with just simple natural language and ask it to just open and play from that exact moment. So for example, there's a video where Jensen Huang says, I would rather torture people to greatness than fire them.

7:57Aravind Srinivas:Can you find it and play it for me from that exact moment? So the point here is not that it can do a video search or anything like that. It's how it does multi-step reasoning, finds the video, pulls the transcript, and actually opens tab and starts playing. wow that was the phrase i was hoping and then you can chat with this video something that i did yesterday when i was watching one of your interviews there are a lot of other videos here i could go here this is like one hour long i could literally say like what are some uh um non-trivial things jensen said in this interview something like that right like i don't want to watch everything i mean maybe he's saying a lot of things he's repetitively saying that AI is going to explode, blah, blah, blah, but I only want to know new things.

8:49Aravind Srinivas:And so it's reading all the transcripts. It's particularly talking about GPUs as time machines. I've never heard him say that. GPUs as time machines. And of course, scale and flexibility, science to application, omniverse, viscosity. So I can actually say, can you just share this as an email? I don't know if you're connected to your team. I have, yeah, yeah. Can you send this as an email to Erevan Srinivas, the founder and CEO of Reflexity? You can actually even edit it. You can even add, say, sent from comment or something like that. And you can even refine the email if you want. And you can just send it.

9:32Try it from there.

9:33Aravind Srinivas:You don't have to copy paste this. You don't have to open this tab. You're going to have to compose email. This is just one example. You could also schedule meetings from here. You could, say, order the leather jacket Jensen's wearing for me on Amazon. It's going to find it. Of course, it's going to be pretty expensive, so it's not going to place the order directly. So it's going to do a lot of these multi-step things. What's going to happen to businesses who survive on paid ads? Is it going to skip paid ads, or how is it going to research information about what product to buy? So you can instruct your comment assistant to say, like, I want you to just skip all the ads.

10:10Aravind Srinivas:When you're tasking Comet to go watch YouTube for you, you can task Comet to say, ignore all the ads, just get me the core sense of what the content is about. You could ask Comet to say, like, go look at like favorite Instagram influencers are wearing and go read the reviews of all of them. And then tell me what's best for me based on what you already know about me. and go buy it for me. So how can businesses optimize for that experience? I would say that what businesses should really do is really focus on just building a great product and get a lot of people to be talking about it. Like reviews, right?

10:46Because Perplexity reads reviews, watches videos. So get as much content about your product as possible, right?

10:51Aravind Srinivas:The solution shouldn't be to paying people to write great things about you. But honestly, building a great product and getting the word out and then people talk about it, rave about it, and then comment reads all that and is able to suggest that to you. That would be the way to go in my opinion. How fast do you think we're going to switch to asking our AI assistants to buy things for us? I think you can do that already. I mean, there's still friction in society. And I remember you were talking about how we're overpaying real estate people, financial advisors. But still, when I think about a person who's in their 60s, who's done it for a while, they're not going to just switch to AI to do things for them?

11:33Aravind Srinivas:I'll tell you an anecdote which kind of convinced me why it's worth building these things. I was once having a haircut somewhere in Noye Valley when I used to live there. And another person was also having a haircut in the same place. And he was somewhat older. And he was ranting about how he spent his entire day until that haircut, probably three, four hours, just looking for a new washing machine. Because it's a pretty expensive purchase. so you're going to like spend a lot of time deciding what to buy and you've faced it your entire day just reading reviews and watching videos and like scrolling through reviews and and you end up getting even more confused than you were at the beginning right and this is the kind of thing we want comment to help you like make really good decisions really quickly uh and delegate all the mundane uh cognitive labor to something like comment like why do you want to go read all the reviews like it's boring like anyway like people are paid to write good things so how do you like get the signal from the noise it's like you have to consume the reviews across different platforms not just one platform i could pay someone to write good reviews on google maps but um i couldn't necessarily pay someone to do it on every single platform i wouldn't be that thoughtful so definitely if your product is pretty bad uh overhyped then someone's going to say something negative on YouTube or Redis.

12:54Aravind Srinivas:And then Plexi will give you a pretty balanced perspective. Okay, here are the five different things. Imagine reading 30, 40 tabs in a matter of a few seconds or minutes and then telling you what to do. And it also understands you, your needs, your budget, everything. It has memory. The comet can answer questions about yourself. It can give you tips on how to do better. So I think that's the kind of way in which we feel AI is less about whether someone's old or young, whether someone's more used to traditional tools or quick to change and adopt new things. It's truly about the utility. But I'll give you another example.

13:33Aravind Srinivas:I was doing a stage talk for a partnership with Motorola. I asked Comet to go read all the talks given by Jensen, Steve Jobs, and everything and also read the current draft that was written by my comms person and then i want you to edit it a little bit to feel a little more grandiose but i still want to preserve how i speak so you watch my interviews too and you watch these guys interviews and you try to blend a little where you know i want to say something that slightly feel more inspiring but i don't want to appear inauthentic and so um and it did a great job actually you did such a good job that i was like damn i would have paid someone like thousand dollars for that let's transition to 2030 what a shopping like everybody has an ai agent yeah they talk to yeah so amazon app is dead walmart i wouldn't say it's dead uh fundamentally the the modes for companies like amazon and walmart is physical real estate the warehouses the stores like walmart stores um and like the packaging and the supply chain like like but the app itself the shopping experience is going to be the core shopping experience might might not matter this is essentially what we've been talking about ever since the beginning it doesn't matter if it's amazon or google or walmart i think it's fundamentally like what's good for the user right like actually the founder of amazon and so say this like you always want a long-term alignment between the customer and the shareholder and as you said as a customer you just said that you don't care about ads right like you actually don't want yeah you want you want what is good for you so just show me that and i'll pay you more money and that's that's the kind of relationship you're going to have with your comment assistant maybe the advertisers are still going to exist it's not like a complete end of ads advertisers might be trying to talk to the agent, not to you.

15:32Aravind Srinivas:But the agent would protect... But then it can go to ignore the ads. Exactly. And that's a handshake agreement you have with your agent that the advertiser cannot influence or cannot manipulate. And that's what protects you. But I do think you might be open to some ads. For example, let's say you have an assistant that works in your team. You're running a podcast company. And so maybe there is a new microphone that's really good. but not many people know about it. They go and tell the person working in your team, hey, can you please tell Silicon Valley Go to go use this R microphone? Actually, it's pretty good.

16:08Aravind Srinivas:That's kind of like an ad to you, but it's being told to your team and you trust that person on your team so you know that that person's not going to unnecessarily... So you're going to evaluate, not just take it as... Yeah, exactly. Is that how you're going to monetize this experience? We haven't thought deeply about it. I'm just imagining a future where ads were done with the level of agents and also between apps. So for example, you might be tasking Comet to do something, like book me a ride. And then the three different apps, like Uber, Lyft, and something else might be like, hey, I think I can deliver the best ride for her.

16:48Aravind Srinivas:And then the agent knows your preference, that you prefer Uber, let's say. And then it could be like, you know what? I really want to win her business, so I'm actually going to offer this right for free. Just tell her that the experience will be better than Uber. The competitor could say that. The agent could charge some revenue for the apps trying to get your attention and share it back with you, which Google never does. So Google takes your queries, does a bidding process among different companies trying to bid for that query, and takes all the money. Doesn't share it with anyone. That's why they have profit margins It's a high answer.

17:27Aravind Srinivas:But that's bad for the user because you at the end are getting influenced. On the other hand, I'm saying there are two levels of protection. One is I don't let you directly see any ad. And even if your agent sees the ad, the company that runs the agent gets some of it, but the user also gets some of it. And that's the best way to continue to win the user's trust by saying, okay, I'm actually open to being advertised to through my agent. and in return for that, I actually want some money back. You don't get to take all of it. When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs.

18:06It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications, and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a$75 sponsored job credit at indeed.com slash podcast. That's Indeed.com slash podcast. Terms and conditions apply. Need a hiring hero? This is a job for Indeed Sponsored Jobs. This episode is brought to you by State Farm. Listening to this podcast instead of doom scrolling? Smart move. Another smart move? Getting help from one of State Farm's 19 ,000 local agents when you choose to bundle home and auto.

18:43Bundling. Just another way to save with the personal price plan. Prices are based on rating plans that vary by state. Coverage options are selected by the customer. Availability, amount of discounts and savings, and eligibility vary by state. I love that model. That way our margins are lower, but our trust is higher.

19:02Aravind Srinivas:So the lifetime value, the LTV, is higher because your trust more. And as long as the agent is not getting manipulated, and therefore you're not getting manipulated, this model could work. So all roads lead to a future where the ad margins are going to go lower. That's for sure. because the user has an AI in their hands for the first time. Until now, AI existed in the hands of the advertising companies. Like Amazon has AI do the recommendation ranking. Google has AI to do the search ranking. So they can influence things and manipulate you to buying stuff because that's what the advertisers are.

19:38Aravind Srinivas:Those are the tools being provided to the advertisers to optimize their campaigns. But for the first time, you have an AI in your hands, the common system. system that works for me that works for you it doesn't work for the advertiser it actually protects you it actually gives you power against them you have intelligence on your demand to summon and do your bidding based on the prompt and your memory and your preferences and so that's a contract between you and agent and that's giving you power against the big tech uh and and therefore uh some of that money can flow back to you for the first time so that i love that i love the way you think about that i hope everybody who's creating ai agents thinks that you do yeah What about other jobs?

20:19So I asked Perplexity. I asked it to analyze my portfolio. And I asked it whether I should find my financial advisor. And it was like, yes, you should do this immediately. That's the first time I saw something that definitive from AI because it said, this is strategy he's utilizing, he's buying very expensive mutual funds. What's going to happen to these professions? Financial advisors, real estate brokers? Because Perplexity does it now and it created the whole strategy for me to rebalance and everything.

20:45Aravind Srinivas:Yeah, I've thought about this, actually. The way I think about it is like real estate agents getting the listings on like Redfin or Zillow with like a comment like browser, let's say you just have a recurring task that's running. It's not yet there. So people who watch this and go and try it, it doesn't work yet, but it's going to work pretty soon, which is you could just have a task that says, you know, every day morning, just look at all the new listings on Redfin. You know, these are the preferences I have. I know Redfin kind of already has some of these features, but it's very hard-coded and there's a lot of selector drops and filters don't really work properly.

21:19Aravind Srinivas:But you could have something like every time there's a house below a certain budget and that's within this neighborhood, make sure to send me a push notification and submit an application on my behalf and schedule a meeting for a visit. You have access to my calendar, everything. So if your agent only shows you listings on Redfin and Zillow, I think they're useless. But if your agent is doing off-market deals for you, they're still pretty useful because those things do not exist on the web. I feel like having the power of intelligence and computer at your hands, it always comes back to that, means you can push the existing industry to work harder for you.

21:59Aravind Srinivas:Same thing with private wealth management or financial advisors. If the job of the financial advisor is to tell you what stocks to buy, then I don't think it's useful. they're not going to be smarter than perplexity which you can go and summon and ask it to go read all the uh analyst reports in every stock and like all the morning day news and they come back to you and tell you like the impact in your portfolio uh what would warren buffett do if he had exactly exactly your portfolio like like it can answer questions like that that no financial advisor can but if your financial advisor does more stuff for you like getting you some access to certain funds like private equity or hedge funds or access to like funds of Citadel or something like that, that you cannot, or like some startups, venture capital, which you cannot get through just the public markets, then maybe they're useful, right?

22:49Aravind Srinivas:It's all about like, you have this new power in your hands. So go and ask the people who are taking your money to do more work for you or the contract between you and them. And it looks like those entry-level jobs are disappearing, right? because they are an entry-level financial advisor. You can only do managing stocks. What do you think is going to happen? Yeah, I certainly think people need to push further. If you're just literally managing portfolio, you're not doing much. I have some private wealth managers too. And I think the way I used to, like helping me move money around, those kind of things, because I don't have time to manage all this myself.

23:26Aravind Srinivas:There, they help me through their time and they also give me some advice on like you know what to do stuff how to save taxes a little bit yeah those kind of things so your value add should expand further than just managing a portfolio if we apply this to any market how can you do this as a recent graduate and do you believe people still need to do their phds like you did or yeah that era is over the fundamental thing you learn a phd is uh learning to learn like i i did a lot of research in my phd which doesn't matter today but i got the ability to like go learn a new topic dig deep understand everything gather my information consult like whoever's the best uh expert which by the way is not needed anymore because you have tools like perplexity but let's say you know you you have the ability to ask the right questions and then you acquire a certain level of subject expertise to go make decisions and like you know branch out that is what you acquire that skill is it worth four years that's the thing I would say like I decided to do a PhD for many reasons.

24:33Aravind Srinivas:And one of them obviously was my own interest in the topic. But the other was my, that was my only way to get to America. A lot of Indians come here for masters. But you got to realize that in order to come here for masters, you need to be able to fund your own tuition fee. And I didn't have that ability. So PhD, the university pays for you. Masters, you pay the university. So that was one of my reasons, not the only reason. So obviously, just because someone gives you a free entry ticket to the United States doesn't mean you should do a PhD because it's a pretty grueling journey. So I had a deep passion for the topic.