WHEN: Today, Tuesday, April 23, 2024
WHERE: CNBC’s “Squawk Box”
Following is the unofficial transcript of a CNBC exclusive interview with Perplexity Founder & CEO Aravind Srinivas on CNBC’s “Squawk Box” (M-F, 6AM-9AM ET) today, Tuesday, April 23. Following is a link to video on CNBC.com: https://www.cnbc.com/video/2024/04/23/perplexity-ceo-aravind-srinivas-on-ai-tech-race-competition-with-google-and-enterprise-launch.html.
All references must be sourced to CNBC.
ANDREW ROSS SORKIN: Welcome back to “Squawk Box.” New this morning, artificial intelligence company Perplexity raising more funding with notable investors, including Stan Druckenmiller, Jeff Bezos has been in the company as well. It’s launching its new enterprise tool for paying customers. And joining us right now this morning, this is an exclusive interview with Perplexity Founder and CEO, Aravind Srinivas. It’s good to see you this morning, congratulations. I talk about your company a lot because it’s really come on in a remarkable way, and it’s pretty amazing to watch what you’ve built. Tell us where you think you stand in the stack, if you could, of how we think about your company versus the other companies in the AI space. We just spoke to the CEO of Anthropic, obviously OpenAI. We can go down the list, Google, and now Amazon trying to build with Olympus and other things.
ARAVIND SRINIVAS: Thank you for having me here, Andrew. So firstly, we are an application layer company. We try to build a product that directly reaches every single user, every single person on the planet, giving them accurate, instant, reliable answers to any question that they can ask. The difference from other AI companies is we are not a foundation model builder. We’d rather take a model and shape it and package it into a great end-user facing application.
SORKIN: What do you — who do you look at as the biggest competition for what you’re doing right now?
SRINIVAS: The incumbent is always our biggest competition, Google, because we are entering the search category, except the difference has been unlike previous competitors to Google that tried to take market share in the 10 blue link search engine market share. Instead, we’re trying to create a new category called the answer engine, where you just get a direct answer to any question you ask, and it disrupts the whole 10 blue link user interface, and it introduces the innovator dialogue for them where going very aggressive on this new user interface will kind of take away their existing cash cow. That’s what this new opportunity for us. And that’s why we are going very aggressive on this new thing.
SORKIN: So let me ask you about this because there is an entire ecosystem and economy, as you know, around what might be described as the blue link economy, right, the search economy, which allows folks both to search for things, but then to send people back to sites across, across the country and the world, and that’s how the search advertising business has worked, but even just the organic native links that appear in search results. There is a real question mark about what happens to that economy, not the search advertising part of the economy, but the actual economy of those who are creating the underlying content that effectively is being scraped. If it all goes back into a system like yours, and you produce the perfected result, and nobody ultimately is going back to the original source or website, what that does to the system. What do you think about that?
SRINIVAS: So Perplexity has been revolutionary in the fact that ever since the first day we launched, we always cite our sources. This is very much unlike other chatbots that started with directly just giving you the tokens that, you know, the model predicted, and more than 10 percent of our queries have people clicking on the links, and a lot of people praise us and, like, even are noticing the referral traffic from Perplexity. So unlike other AI chatbots that take all your data from these different links and try to train models on it, we do not do that. Our goal is to synthesize information from several different sources, aggregate it into a good summary, and give the user more guidance in which link they want to go click on.
SORKIN: But fair to say, even if you’re delivering some referral business, fair to say that the referral business long-term, if the answers are as good as I think you want them to be, that there should be less referral business ultimately.
SRINIVAS: Yeah, so the quality of the referral is much higher here. The intent of the user is much higher here, right? They’re only coming and reading a link after, despite reading all the summarized answer, they’re specifically knowing which link to go click on. They want to read more. I think the analogy I want to give you is a movie trailer with an actual movie. The summary that we provide you is like a movie trailer, or a teaser even you can say, but people will still go and check out an actual movie. And that’s what happens here. Like, you still need good content, you still want to go and read the original content and Perplexity lets you do that very easily.
SORKIN: Long-term, your business model looks like what? It’s an enterprise business where people are paying $40 a month per seat. There’s an advertising component to it. That price comes down at the consumer level. What does that look like to you?
SRINIVAS: Yeah, so, so far we’ve been doing direct consumer subscriptions and that’s gone really well. Except now we are ready for the next phase where every person, every billionaire like Jensen Huang or Jeff Bezos, Tobi Lutke, they’re all using Perplexity because the only thing they don’t have in plenty is their time. They’re all in their pursuit of accurate source of information and knowledge. And every knowledge worker has the same problem too. They’re all maxed out on time. So Perplexity Enterprise Pro will basically be the exact replica of the Consumer Pro, except it will come with enterprise-grade security, compliance, single sign-on, user management, all the features you need for companies to feel comfortable letting their employees use it at work.
SORKIN: OK, so the big question, though, is if I put my prompt or my question into Perplexity, who owns the prompt and what’s the security around it? Because that’s obviously the big question. Everyone’s sitting around saying, can I paste this part of my document into this program without worrying that what I just pasted is now going to be used to either train or teach this program to — to provide search results in the future?
SRINIVAS: Yeah, so the Enterprise Pro really solves all these problems. We are not going to take your prompts. You’re not going to train on any of this. And even in our Consumer product, if you turn off the AI data usage, we’re not actually training on any of your data.
SORKIN: When you look at this, the distinction for those viewers who just haven’t played with it yet, and it really is a remarkable program, the distinction between what you think you do and Google does today, given what they’re trying to do with Gemini and where you think this goes?
SRINIVAS: Yeah, so first of all, like I said, Google has been the most wide — the Google search engine has been the most widely used enterprise technology. You might not even see it as an enterprise tool, but it is — it has been the most widely used enterprise technology. Everybody uses it multiple times a day for day-to-day work, except you are paying for it through your data, your lost attention, your time. And that’s what Perplexity Enterprise Pro solves. Now, you can ask, why would Google not do something similar with Gemini? The reason is that, like, the more and more people start using Gemini, at least people in the United States, the knowledge workers, people who are actually capable of paying a lot, people who advertisers really want to target, if they stop using the 10 billion user interface and start going to Gemini, it just really lowers their margins on the advertising business. And this business, the subscription business or the enterprise business, while it can be amazing for a newcomer like us, it may be actually a poor idea for Google to go aggressive on because it might just be a worse business for them. And they have such a big cash cow to protect. So that’s why, like, we are feeling comfortable about competing with them in this category.
SORKIN: I want to congratulate you. It really is, as I mentioned, a pretty cool product. And it’s great to follow your progress. Hope you come on back and we can talk more. Thank you.
SRINIVAS: Thank you.
SORKIN: You bet.
