- 16 hours ago
Category
🗞
NewsTranscript
00:00You started Perplexity with the ambitious mission of taking on Google using AI.
00:06And what most people don't know is that you were actually an intern at Google DeepMind in 2019.
00:13It was your first job after studying for your PhD at UC Berkeley.
00:18Go Bears!
00:20Did you know when you were interning at Google that you would, just a few years later,
00:24start a multi-billion dollar startup going against it?
00:28Definitely not.
00:30But I did everything an intern would do, just sleep in the office.
00:35They had all these pods which you can use to sleep in the office.
00:39And they had a library.
00:41They had a lot of food.
00:43A lot of junk food, too, which, you know, definitely I put on a lot of weight while I was at London.
00:50And so it was one of the best times I had in my life.
00:53I think there's a lot of things.
00:57I would say that DeepMind was almost like a different company.
01:02I even remember how, like, Demis had all these crazy visions of how, like, AI will just be a scientist by itself,
01:11and it'll solve all theorems and proofs.
01:14And, like, they were on a completely different trajectory.
01:16So now, like, they got folded back into Google to, you know, get them back into this AI race on the product stuff.
01:24But to me, like, DeepMind meant a lot more in terms of just going for that core artificial general intelligence dream.
01:32And it was a fantastic time.
01:34So you went on to work at OpenAI in its early days and then eventually decided to break out and start Perplexity.
01:41And for those who don't know, can you tell us a little bit about what Perplexity is?
01:45So Perplexity is an answer machine.
01:48You get answers that you can really trust.
01:50We were the first to reimagine search in the form of conversations and answers with trusted sources.
01:57And we did it in a way that essentially became the blueprint for every other chatbot out there.
02:03Yes, everybody else has integrated search, too.
02:05But when we started, people made fun of us saying, in AI, hallucination is meant to be a feature, not a bug.
02:12And so I even had a former Googler who was one of my early seed investors, and he said, this looks good.
02:20But the reason ChatGPT is going viral now is because people want to laugh at AI's mistakes.
02:26And your product is trying to, like, design it in a way where you do not make mistakes because you're always pulling sources and summarizing what humans are saying.
02:34You're not telling what the AI thinks.
02:36So I don't think this will work.
02:38That was what was told to me.
02:39Like, he literally said you should go and build something like Character.AI, like, where people actually enjoy it, like, AI saying arbitrary random stuff.
02:49And, you know, I'm glad I didn't listen to that.
02:52And I believe that, like, this is fundamentally a completely different tool.
03:00We didn't even build it as, like, a Google replacement, actually.
03:04We just built it because it was really useful to us.
03:06I was a very newbie founder, CEO.
03:10I didn't take any health insurance because I was, like, I'm all in on this company.
03:14I don't care.
03:15You know, if the company doesn't work, like, you know, that's my health insurance, actually.
03:18And then my founders, the co-founders were married, so they had health insurance to their wives.
03:23So that was when my first employee we hired, he asked for insurance.
03:28I had no idea, like, which HR tool to use, what are all these different plans, how to even, like, figure out which plan to go for.
03:36And, of course, insurance is that one ad word category that Google makes, like, billions of dollars a year.
03:40So they have literally zero incentive to tell you, like, which plan to pick or which provider to go for.
03:46And that's when we started building these tools that could just give you answers directly in a chat interface but plug into the web.
03:54And so we kind of essentially built it as a tool we wanted and our friends wanted it, our investors wanted it, and then we put it out into the world.
04:02I never thought about it as, like, oh, like, links, 10 blue links to answers.
04:08But once we put it out, that's when we realized, like, based on what people are saying, hey, this is actually, like, a much better way to, like, consume the Internet.
04:15Like, we're all living on the Internet.
04:17And so I would rather just feel the Internet through answers instead of links.
04:21Got it.
04:23So, yeah, you guys were really one of the first AI companies to do real-time AI search focused on that and linking back.
04:32Now everyone's doing that, right, including Google, including OpenAI.
04:35So what are you doing to differentiate?
04:38And do you want to talk a little bit about Comet, your new browser, which I took a sneak peek at and is really interesting?
04:43Sure.
04:44Look, with respect to Google, I mean, I don't mind saying this.
04:47Like, every year it's the same feature announced in the I.O.
04:50Like, I think in 2023 they called it search generative experience.
04:55In 2024 they called it AI overview.
04:58This year they called it AI mode.
05:00Next year they'll call it another name.
05:02But at the end of the day, the feature never gets shipped to the user.
05:05Right?
05:05Like, literally, this is not hard.
05:07Like, they have all the models.
05:09They have the best index.
05:10They have all the best infrastructure.
05:12They have their own hardware.
05:13Data centers, everything.
05:15Three billion, four billion users.
05:17So why not just change it?
05:18Change Google.com.
05:20Like, make it, like, perplexity.
05:21If that, anyway, AI mode is literally, like, even the font is, like, basically perplexity now.
05:27That's the extent to which they've studied our product.
05:29Great.
05:30So why don't you go ahead and ship it?
05:31You cannot because you lose all the ad revenue.
05:36Like, if you can go and ask, like, what are the best sneakers for you to buy for, or, like, what are the best jackets for you to buy?
05:42Or best airlines to pick for this something?
05:45Or best hotels to stay at a place?
05:47How are you going to charge all these people for money?
05:49Oh, if you literally, in fact, like, leave the AI answers.
05:53These days, if you just go and ask for the NBA score, before even triggering the score widget, they put the Ticketmaster ads right at the top.
06:03Because they need the money.
06:05And if that, that's the money that's paying for all the R&D spend on Gemini.
06:10Right.
06:10But what Google does have is reach, right?
06:12So they have, you know, billions of users.
06:14You use it every day.
06:15But they cannot ship it.
06:17So that's, that's, and that's been consistent over the last three years.
06:21And let's talk about, though, I mean, they do ship a lot.
06:23But let's talk about what you are going to do to try to get that same kind of reach and that same kind of integration that users have with Google in their everyday life.
06:32Yeah.
06:32Which, as I understand it, the browser is a big part of that strategy.
06:35So can you tell us that?
06:36Why do you think browsers are the next frontier for successful AI companies?
06:40Yeah.
06:40Often what's said, like, why did Sundar become the CEO of Google?
06:45It's because, like, he focused on the Chrome project, right?
06:48He was the one running the Chrome project, and that ended up being one of the biggest weapons for them, this battle against Microsoft.
06:56And it took them many years, like probably 10 years, to actually become the leading browser.
07:02The browser is the front end of your internet, like experiencing the internet.
07:08And the Omnibox, the search box, is where almost all the Google queries are going.
07:13That's why if you go to Google Trends and see the topmost queries by volume, it's always weather, or, like, Amazon, or Reddit, or Instagram, Twitter, like these one-word queries.
07:24So most people are just using Google as a navigational tool.
07:28But if you can blend navigation, information, and activity, like transactions, and, like, doing actual browsing sessions, all in, like, one clean interface, using the browsing infrastructure,
07:41you can actually, like, go for it all in, like, one single tool.
07:46So that's, and if you really want to transition AI from answers to actions, to doing stuff for you, answers are essentially, like, four or five searches in one.
07:59Actions are, like, an entire browsing session in, like, one prompt.
08:02You really need to actually have a browser and hybridize the compute on the client and the server side in the seamless way possible.
08:12And that's, that calls for rethinking the whole browser.
08:16We're not actually, like, thinking about Comet as a, yet another browser.
08:20It'll be a cognitive operating system.
08:22It'll have, it'll be there for you every time, any time for work or life, as an assistant on the side or, like, just going and doing browsing sessions for you.
08:32And I think that'll fundamentally make us rethink, like, like, how we even think about the Internet.
08:37Like, earlier we would browse the Internet, but now people are increasingly living on the Internet.
08:42Like, a lot of our life actually exists there.
08:44And if you want to build a proactive, personalized AI, it needs to live together with you.
08:50And that's why we need to rethink the browser entirely.
08:53I want to talk a little bit about growth.
08:56You all have tripled your valuation, I believe, twice in the past couple years.
09:02Let's talk about where you're at in terms of how many people are using Perplexity.
09:06Can you give us the latest on how many people are actually searching on Perplexity using it every day?
09:10Yeah, I believe in May we did about, like, 780 million queries.
09:14Per month, is that right?
09:18Yeah, per month.
09:19And I think that's growing, like, at more than 20% month over month.
09:23So give it a year, we'll be doing, like, we'll be doing, like, billion queries a week if we can sustain this growth rate.
09:31And that's pretty impressive because we were, like, the first day in, like, 2022, we did 3,000 queries.
09:38Just this one single day.
09:39So from there to, like, doing 30 million a day now, it's been phenomenal growth.
09:44And we still think that same trajectory is possible even now, especially with all the kind of distribution partnerships we're, like, trying to seek and, like, all the browser that we're working on.
09:57And, like, if people are in the browser, it's infinite retention.
09:59So everything on the search box, everything on the new tab page, everything you're doing on the sidecar, any of the page you're in, these are all going to be extra queries for active user, as well as seeking new users who just are tired of, like, legacy browsers like Chrome.
10:14I think that's going to be the way to grow this over the coming year.
10:18Yes, and I want to talk about those partnerships in a second.
10:21But before we do, just on the fundraising route, as I mentioned, you tripled your valuation last year, tripled again a few months later.
10:28We've reported that you're now in talks to raise another $500 million at a $14 billion valuation.
10:33Taking a step back, is it challenging to raise these mega rounds that get bigger and bigger?
10:40It is expensive to run an AI company going against the big guys.
10:44What is it like, fundraising, in this market as a startup?
10:47Fundraising is super interesting to you, but it's actually, I would prefer not to, like, do it at all, you know?
10:54I think it's, look, I think we're attracting the funding because the product is really good.
11:01I think our vision for the future is always, like, unique and, like, original, and we're not, like, chasing others.
11:07We're actually, like, setting the road map for other people, including, like, taking our font and, like, UI and everything, sure.
11:14But literally, like, the, you know, you said search, but it's not just that.
11:19Even the agentic, the tool calls, how iteratively we call the tools, our research agents, like, the recent feature we put out is labs.
11:29These are all setting the next steps for how people will use these tools, and I think the browser will be very unique in its own way, too.
11:37So people, investors, are excited about the possibility of, like, you know, completely changing the front end in which we experience the whole web.
11:45And the web is so valuable.
11:47And the other thing that's most important to consider about perplexity is we're the only company, like, razor focused on all the accuracy aspects of AI.
11:56Like, we really want us to be the accuracy layer for AI, and that's going to influence both human and AI decision-making.
12:05And every day, like, trillions of dollars worth of decisions are made across retail, finance markets, exchanges, everything.
12:13And if we can influence a big chunk of that, then that will automatically means, like, we can be worth trillions of dollars in market cap one day.
12:20You mentioned the major partnership distribution deals that you're doing with hardware makers.
12:28Perplexity is going to be pre-installed on Motorola phones.
12:30We've also reported that you are nearing a deal to get perplexity pre-installed on Samsung phones, potentially replacing Google's Gemini assistant.
12:39Can you tell us a little bit more about these hardware partnerships and why they're important to your strategy?
12:45Yeah, so towards the end of last year, we started thinking about how can we go beyond just an app on the phone to being a native assistant.
12:55The Google assistant is a terrible experience.
12:58And so, well, I think even they know it.
13:01That's why they're deprecating it themselves and trying to replace it with Gemini.
13:04So there's literally no assistant that you can have with the invoke with the action button or gesture at the bottom of your screen and just quickly talk to it and voice and ask questions and ask it to do stuff for you, like simply playing media or setting alarms, reminders, sending your emails, you know, setting smarter reminders.
13:26These things never even worked.
13:27And so we started to think about perplexity more as an AI that's just helpful and right there for you all the time.
13:33Like I use the word cognitive operating system.
13:36You can fundamentally rethink the OS on all your devices if you think AI first.
13:42And so we started building the Android assistant.
13:44We put it out in January this year.
13:45It was super exciting.
13:47And a lot of OEMs saw that and they were like, damn, this is the coolest thing we've seen.
13:51And, like, we just can completely, like, make Android, like, really much better.
13:57And so they started talking to us.
13:59So that's how the Motorola partnership came out.
14:01And that's how every other partnership that we're in talks with are, like, merging.
14:05And can you confirm that Samsung is also going to be a partner?
14:07Well, we're talking to a lot of people.
14:09So that's all I can say.
14:10Is it difficult negotiating with phone makers who may not want to compromise their relationship with Google?
14:17I thought so.
14:18And definitely Google has given us an extremely hard time.
14:20Like, every time we're very close to signing a deal, like, there's always, like, some calls from Mountain View that are being made.
14:26I don't know it's from where.
14:28But, so, you know, they definitely don't want us to succeed.
14:33So, but we're not going to, like, I mean, if I thought, like, it's hard to succeed, we wouldn't be doing this company.
14:41So, and I think it's good for the world that, like, finally someone is trying to, like, actually fight and compete against this distribution lock-ins that they have.
14:51And the DOJ case is going on.
14:54And I hope something good comes out of it.
14:56We've made our statements that how Android as an OS needs to be a lot more open.
15:00And perplexity needs to be, not just perplexity, by the way, I'm happy for ChatGPT or Claude or, like, Microsoft Copilot.
15:07Every contender in the space should be an option for the default AI.
15:11And Google cannot pay their way in to be the default, especially when they don't even have the best assistant.
15:21And that's, you know, one thing that's interesting about perplexity is you offer all, or many different AI models, right, on your platform.
15:30I want to also talk about, you know, you mentioned how important it is to be accurate.
15:35And right now, it is a tough market economically to produce accurate information.
15:41I know that firsthand as a journalist, right?
15:43You've had some issues with media companies in the past who have accused perplexity of scrubbing content.
15:49You've since struck revenue share deals with other major outlets, like the LA Times.
15:54What do you think...
15:55But we're very interested in doing it with Bloomberg, too, so...
15:58That's not my department.
16:02But what do you think is a sustainable business model here for accurate, truthful information to be produced at the same time that AI chatbots are bringing more of this information into their tools and not to publish your sites?
16:16Yeah, so fundamentally, like, sources and citations has been the forefront of our product.
16:23I think we've used LLMs as engines that are doing the reasoning and synthesis and summarization.
16:30But the actual authentic sources of content are always mentioned as citations in our UI.
16:37And I'm happy others are adopting that, too.
16:39And we hope to drive traffic through those citations.
16:42We're never...
16:43Like, I'm never going to, like, lie and say stuff like, oh, that we're still sending a lot of traffic and all this stuff that the bigger companies are trying to say.
16:50All that's, like, definitely wrong.
16:52We've been very transparent that, yes, there will be less traffic referrals from these UIs.
16:57And we're being open about it.
16:58And on the other hand, what we're saying is, like, okay, like, imagine we make some revenue per query through some experimental work that we're doing on advertising.
17:09Or even, like, say we have sell somebody or, like, someone converts to becoming a pro user through a query that cited, like, a very valuable piece of content.
17:18We could potentially, like, just share the revenue with them.
17:20That was the model we came up with.
17:22But, again, this is grounded in Jeff Bezos' thought of, like, your margins, my opportunity.
17:28Who has the least incentive to share any revenue with the publishers?
17:31It's Google.
17:32Like, they drive traffic to others, but they take all the ad revenue.
17:36Right?
17:37We don't have to have the margins they have.
17:39So we are happy sharing revenue with the publishers.
17:42So that was the origins of our publisher program.
17:45And we're happy that it's slowly expanding.
17:47It's not growing at a crazy pace.
17:49Because everybody wants to, like, everybody's skeptical, like, how much revenue can be generated in these AI platforms at a query level.
17:57No one's really managed that.
17:59People are still doing the subscription-based monetization.
18:02But in the future, you could imagine research agents, like, these deep research agents can plug into, like, different data providers.
18:08And then the user can even, like, pay for accessing some things, just like they, you know, pay for these paywalls.
18:15And then the agent can, like, you know, see through the paywalls because the user actually paid for it.
18:21And that revenue can be shared between the publisher and the AI.
18:25There's a lot of ways to do this.
18:26But the number one thing to understand here is, like, the intention to share revenue and the intention to be open about, like, okay, there's definitely not going to be the old-school style of, like, just traffic referrals from these new UIs.
18:39It's just how these tools work.
18:42And it's important to, like, also remember, like, the more you get cited in an AI tool, the branding for the source also, like, increases.
18:50It's, like, I would argue that maybe years from now, it's probably better to be an authoritative source consistently in an AI product than ranking number three on the 10 rule in QI.
Recommended
9:42
|
Up next
1:58
1:57
14:52
6:50
6:23
4:37
5:51
19:15
17:17
19:31
0:28
2:18
3:32
15:17
10:35
3:35
3:58
5:43
7:01
19:46
2:21
9:56
Be the first to comment