Perplexity CEO Stanford MBA Entrepreneurship Dialogue & Insights: The Future of AI Search is Not "Google," but Completing Tasks for Users

Recently, Aravind Srinivas, co-founder and CEO of Perplexity, had an in-depth conversation with MBA student Aislin Roth at Stanford Business School.

Srinivas talked about his journey from growing up in India to founding Perplexity, from his academic pursuit in India to pursuing a PhD in computer science in the United States, and then building an AI search engine valued at three times its original worth.

• Entrepreneurial Insights: As a founder starting from scratch, Srinivas shared how he led the team to grow into a company valued at $9 billion.

• Ethics and Future of AI: He deeply analyzed Perplexity's stance on sensitive topics such as advertising introduction and copyright disputes, as well as his vision for the development of ethical AI.

• Disruptive Reshaping of Search: How does Perplexity help users not just find information, but complete tasks and save time, unlike traditional search engines?

Here is the complete content of the conversation:

Aislin Roth

We are thrilled to have you here today. Many of our audience members are active users of Perplexity, especially thanks to the free Perplexity Pro service provided to all Stanford students. So, we are incredibly excited about your presence.

To start, I want to use Perplexity to help me design my questions. Let's see what it has to say. I entered a prompt: I am preparing for a one-hour live interview with Stanford Business School students, interviewing Arvin. What questions should I ask him?

Aravind Srinivas

I hope these questions aren't too difficult. However, you might have missed some things in your prompt.

Aislin Roth

That might be your responsibility! So, Perplexity did provide a very detailed response. In summary, it suggested we discuss the following aspects: first, your personal background; second, the early stages of Perplexity; next, the current state of the company; and finally, various leadership lessons learned.

How about that outline? I think it sounds like a good framework, but I want us to make it a bit more personal.

So I tested my follow-up questioning ability and asked Perplexity: What might the audience not know about Arvin? What is the most interesting thing about him? What questions can I ask him to add humor to this conversation? Lastly, I also asked what questions would be suitable for a quick-fire round at the end of the interview.

So, stay tuned, everyone. Finally, I asked one more question: Does Perplexity make mistakes?

Aravind Srinivas

More than you think.

Aislin Roth These questions are indeed more interesting than my initial prompt. For example, I learned that you love cricket; secondly, I found out that you actually started self-learning programming because you missed the admission score for the computer science major by just 0.001 points; finally, I learned that you have a connection with Google CEO Sundar Pichai, as both of you come from the same Indian city, Chennai.

Perplexity also suggested that I watch the video of you and Sundar in Chennai. So, I think starting from here is a good choice. I won't include the video, but Arvind, do you think there's something special about the water in Chennai? How come so many successful tech entrepreneurs have emerged from there?

Aravind Srinivas

This is indeed interesting. I think it's not just Chennai; many cities in India have also produced a lot of outstanding talent. I want to mention a very common cultural phenomenon, which is the pursuit of excellence, striving to do the best, and placing a high value on education. This is common in Chennai and many other cities.

At least in my circle, respect for knowledge and being well-educated is even greater than the pursuit of wealth. If you are knowledgeable, you will earn a lot of respect from people. I think this directly influences everyone's attitude towards learning, not just limited to studying for exams, but delving deeply into subjects.

Chennai also has a phenomenon where there are a lot of cricket fans. They pay special attention to all the statistics. Even before Google or other quick query tools appeared, we could recite all the players' data and were fascinated by data analysis such as scoring rates and averages.

You could say that before formally studying statistics, we had already grasped many basic concepts through cricket. I think this spirit of pursuing knowledge is a common point for many people from Chennai who have succeeded in Silicon Valley or elsewhere.

Aislin Roth

So to sum it up, knowledge is more important than wealth, and also to watch more cricket matches, right? It's a pity that you left India to pursue a Ph.D. in computer science at Berkeley instead of coming to Stanford.

Aravind Srinivas

Actually, I wasn't admitted to Stanford; that's true.

Aislin Roth

But we are fortunate to have you with us today, which is really lucky. So, how has your academic background influenced your creation of Perplexity?

Aravind Srinivas

This is actually very core. Perplexity has emphasized citing sources from the very beginning. This is true; I'm not making up a story. When I first arrived at Berkeley, I thought the Ph.D. life was magical, like in the movie "The Theory of Everything," where you just knock on your advisor's door and say, "This is my thesis idea." But reality is not like that. You need to accumulate step by step and come up with your original ideas.

Institutions like Berkeley or Stanford are great because they provide a framework to help newcomers find direction, rather than completely leaving them to their own devices. When I entered the lab, I was asked to assist a senior student in writing a paper based on their ideas, learning skills through this process. Later, I learned about the concept of citation The influence of academia depends not only on whether papers are accepted but also on whether they are cited and used as a basis for further research. I found that writing complex ideas can increase the chances of paper acceptance, but if it is too complex, few people will cite it. Therefore, it is essential to find a balance that allows the paper to be accepted while also making it easy for others to cite.

Additionally, I learned that the core inspiration for the Google search engine comes from academic citation graphs, combining the concept of web hyperlinks. This academic background directly influenced Perplexity. Our core question is: what if AI's answers were like academic papers, requiring citations for every sentence?

These sources need to come from resources on the internet that have a certain level of authority and credibility. We have embedded this concept into the system, creating a very unique product experience. Thus, my academic foundation has indeed had a profound impact on the birth of Perplexity.

Aislin Roth

Now I know that Perplexity's academic focus feature is one of the keys to its popularity at Stanford. It allows users to cite only academic journals in their research, which is very helpful for us students. At Perplexity, you are very focused on improving information access channels and are building the world's first "answer engine." Why is democratizing knowledge access so important to you?

Aravind Srinivas

Because I personally enjoy using it very much. As I mentioned earlier, I come from a culture that values knowledge highly. As Charlie Munger said, "The best thing you can do for others is to help them learn and acquire more knowledge."

For each of us, the pursuit of wisdom and becoming a continuous learning machine is almost a moral responsibility. There is nothing that helps us improve ourselves more than this.

If you focus on wealth or social networks, you might consider them as metrics for measuring progress in life. But to a certain extent, these things lose their appeal and no longer serve as your motivation.

On the other hand, knowledge has no end. This is also why Perplexity's slogan is "The Starting Point of Knowledge," because knowledge is indeed infinite. You can only keep getting better.

If we had to choose a metric to measure our progress, it would be whether our understanding of the world has deepened. If this is at the core of human nature, then ensuring that everyone can conveniently access tools that help them achieve this goal becomes crucial.

We are doing our utmost to achieve this. Of course, some advanced services are still limited by paywalls, but as these AI models become cheaper, smarter, and more efficient, and are refined into smaller versions, it becomes possible to create a widely accessible version that everyone can use to get instant answers to any questions.

Aislin Roth

"The Starting Point of Knowledge," yes, that's right, I love this concept. Now I want to return to the origins of Perplexity. To tackle such a massive problem, you need an excellent team. So how did you choose your team members, and how did you form your initial founding team? Aravind Srinivas

I was fortunate to meet one of my co-founders, Dennis, during my PhD studies. This is also the benefit of an academic background—you meet very motivated and insightful people.

At that time, we both wrote papers on the same idea almost on the same day, which is how we got to know each other. He had spent some time in my lab as a visiting student, and we often brainstormed ideas for potential collaboration, although we didn't achieve much at that time.

For founding team members, I believe an important quality is complementarity. You don't need to be good at what they are good at; ideally, they should be better than you in those areas. At the same time, you also don't want to interfere with their work in those areas.

In our team, Dennis and Johnny are my core founding team members. Johnny is a top-tier competitive programming expert who represented the United States at the International Olympiad in Informatics (IOI). Those familiar with competitive programming might know a top coder named Tourist, and the only time someone has ever beaten Tourist at the IOI was Johnny.

Geniuses like Johnny can not only write stunning code but also solve complex problems quickly. He possesses exceptional programming and problem-solving skills.

Dennis, on the other hand, has a strong background in AI and software engineering. The combination of these two gives me the confidence to take risks and establish the ambitious goal of creating a brand new search experience, which would otherwise be nearly impossible to achieve.

Over time, we gradually recruited more people with new skills. For example, none of the three of us had front-end programming skills, so we hired someone proficient in full-stack development.

We also hired someone very skilled at writing CUDA kernels. The team's capabilities increased not only in skills with the addition of new members but also achieved a multiplicative effect in overall efficiency. This enhancement is essential.

For instance, our design team is also very well-known, but no one in the founding team is skilled in design. So we specifically recruited a designer who is passionate about the product but lacks an AI background.

He previously worked at Quora, focusing on how humans answer questions. When we provided an AI platform for him to design, his imagination played a huge role, creating results that had a complete multiplicative effect. This is my overall approach when expanding the team.

Aislin Roth

Find those partners who can exponentially amplify your abilities.

Aravind Srinivas

That's right.

Aislin Roth

Less than a year later, while you were in the process of Series A funding, you learned that one of your key competitors, OpenAI, had just launched a competing search product. How did you respond when you heard this news? What gave you the confidence that Perplexity still had a place? Aravind Srinivas

Are you referring to the situation during the Series A funding?

Aislin Roth

Yes, including the search competition product launched by OpenAI.

Aravind Srinivas

Actually, I need to clarify that it wasn't OpenAI that launched the search product at that time, but Microsoft was about to release a new version of Bing. It sounds like a story from a Silicon Valley TV show. We were finalizing some terms in the NEA office, and later Dennis and I went to a nearby Blue Bottle coffee shop to relax, feeling like we had finally completed it.

However, The Verge suddenly published a report revealing that Bing would launch on Monday, even leaking screenshots of the A/B testing. There is a process in venture capital called "due diligence," which generally takes 30 days.

At that time, another venture capital firm extended their due diligence period from 30 days to 45 days after seeing this news. I thought to myself, this seems a bit off; maybe they want to back out.

But an NEA partner called me on Saturday morning and told me not to worry about Microsoft's news; they believed in us and would not back out of the deal. This gave us a lot of confidence.

This was crucial for us because I had heard many stories like this: founders get the investment term sheet but ultimately do not receive the funding. However, NEA showed us real trust. This trust made us even more convinced that despite facing competition from Microsoft, we still had a chance to carve out a place for ourselves.

Aislin Roth

Fortunately, your investors chose to support you. As a challenger, you must have needed a lot of creativity during the fundraising process.

You successfully attracted investments from Amazon's Jeff Bezos, "father of AI" Yann LeCun, and Nvidia. So, how did you assemble such an excellent team of investors? Are there any experiences or stories worth sharing?

Aravind Srinivas

Here's an interesting story. Dennis was at New York University (NYU) at the time, so he knew Yann. But Yann is a celebrity, and it wasn't easy to reach him. At that time, Yann was on vacation in France and had just returned to the NYU campus.

Since we happened to be in New York, we decided to wait outside his office for several hours, even having lunch there. When he came back and saw us, he said, "You’re still waiting? Well, let's chat."

We finally secured half an hour with him. To impress him, we prepared a Twitter-based search demo that allowed him to search for his tweets, see who replied to him, and how many followers those responders had. This feature was very close to everyone's daily curiosity. After using it for ten minutes, he said, "Okay, I want to invest." Similar things have happened to other investors, such as Andrej Karpathy. He is a celebrity who asked us for an investment deck. I simply gave him the product link and let him try it himself. Jeff Dean was the same. All these investors were deeply attracted after actually using the product.

The biggest lesson I learned from this is that if something is not your strong suit, like making an investment deck, then don't force yourself to do it.

For me, demonstrating the product's effectiveness is more effective than any deck. So, I make sure the product link is directly usable and runs well. If the user experience is great, the product itself will be more persuasive than any document.

First of all, most investors are busy, and they are likely to use their phones more than sitting in front of a computer reading the deck word for word. Secondly, if you are not good at making decks, then don't try. By the way, I have hardly ever made a deck.

Even in our Series A funding, the materials I provided were very concise; by the time of Series B, C, and even D funding, I almost completely replaced traditional investment decks with memos or Notion documents.

When I see the decks of some successful companies, like Airbnb, LinkedIn, and Facebook, I find that they are very different in style, making it even more confusing to know which way to imitate or how to be original. So I simply gave up making complex decks and focused on showcasing our product's value.

Aislin Roth

So, focus on your strengths rather than blindly imitating others. This is a great lesson. Among today's audience are many future entrepreneurs, and I think this reminds us that persistence, determination, and hard work can pay off, right?

Aravind Srinivas

Yes, you don't need to be good at everything. For example, if you are a founder or CEO but can't make a deck, that's perfectly fine.

Aislin Roth

Many consultants present here specialize in making presentations, so we might have exactly the opposite skill sets. Maybe we can collaborate to become a good team.

Now, Perplexity is building an answer engine, but you neither own the content nor the models. So, what is your technological barrier? Why is Perplexity's approach superior to direct vertical integration?

Aravind Srinivas

You mean to politely ask if we are just a "shell," right?

Aislin Roth

That's what you said, not what I said.

Aravind Srinivas

Yes, but it is indeed a question worth discussing. About a year ago, there was a significant divide in the community regarding what type of startups should invest in: should these companies train their own models or use APIs? Our belief at the time was that, first, models would become increasingly commoditized. If you want to be one of those companies building models, you need to have massive financial backing and be able to operate as a company that can lose billions of dollars each year yet still continue to function.

We are neither in that position nor do we want to become such a company. So we chose to use models from other companies and adapt them into products suitable for end-to-end consumer search experiences.

It turns out that this decision was correct. Many companies that tried to build their own models no longer exist. This clearly indicates that you either raise $10 billion or you shouldn't attempt to build foundational models yourself.

What we are more focused on is providing high-quality answers. The question is: does providing accurate answers for everyone require building your own foundational model? If the answer is an absolute "yes," then we shouldn't pursue this without raising $10 billion.

But I believe that with the advancement of open-source technology and the reduction in model costs, the costs of these APIs will halve every four months. If this trend continues for a year or two, we will at least enjoy a cost reduction of 10 to 100 times, while the intelligence and reasoning capabilities of these models continue to improve.

Open-source models constrain the pricing of closed-source models. This is the best time to build applications. We focus on how to leverage these models for post-optimization, ensuring they perform well in summarization, citation, formatting, and creating customized user interfaces for multiple verticals such as finance, sports, reasoning, etc. There is so much more to do beyond the models, making it entirely worthwhile to build a differentiated business.

Ultimately, most successful businesses are essentially some form of "shell." For example, Coca-Cola might not have succeeded without the support of refrigeration technology, but it is now an extremely valuable direct-to-consumer product.

So, you can create a "magic formula," find the right packaging, leverage existing foundational technologies, and provide tremendous value to consumers, which is entirely worth building. That is exactly what we aim to achieve.

Aislin Roth

What should be built in-house, and when can existing technologies be leveraged? That's a great strategy. You have publicly criticized Google for being overly reliant on its advertising model. However, just last week, Perplexity announced its first introduction of ads. What is your commercialization strategy? What role will advertising play in the future?

Aravind Srinivas

Perplexity's advertising model is completely different from Google's. One of Google's issues is that its ad units overlap with its answer units.

In other words, Google provides answers through a series of links, and the ranking of those links can be influenced by ad bidding. When users search for relevant information, it can be frustrating if the ranking of links is manipulated by ads.

We want to avoid such pitfalls and choose a lower-profit but fairer advertising model for users. We believe that answers must be objective and truthful, unaffected by ads. After providing answers, we suggest some follow-up questions to users, which may be influenced by ads For example, if a user searches for "I'm looking for running shoes," we will provide an answer and may suggest the user ask, "What makes Adidas tennis shoes better than Nike?" Such questions. Users can choose to ignore these questions.

This form of advertising is currently in the experimental stage. We are collaborating with some brands willing to try this model. Currently, the biggest concern for brands is that they cannot control the content of the answers. It takes a certain amount of courage to accept such an experiment. At the same time, since this form of advertising may not necessarily bring a large amount of traffic to brands, the return on investment (ROI) is still unclear.

Nevertheless, one thing we are very clear about is that it will never affect the accuracy and authenticity of the answers. Once we follow Google's old path of increasing advertising revenue at the expense of user experience, user trust will decline, which is not what we want.

Aislin Roth

So, when I input the initial prompt for this interview, I will never see ads answering, right? Well, it's reassuring to hear that. However, Perplexity is clearly innovating rapidly, and this speed of innovation has also sparked some controversy.

For example, News Corp, the parent company of The Wall Street Journal, has sued you, accusing you of copyright infringement. The New York Times has also issued a cease and desist notice, accusing you of improperly using its content. How are you responding to these challenges? What is your vision for the development of ethical AI?

Aravind Srinivas

Our position is this, which we have also mentioned in our blog post: No one owns the copyright or ownership of facts or truths. This applies equally in the field of journalism.

For example, in our interview, you mentioned that The New York Times sued Perplexity; this fact has been reported by others, and now you mention it in our interview. Can anyone claim copyright over this statement and prohibit you from mentioning it? Obviously not, right? The truth should be widely disseminated.

Of course, the specific way the truth is expressed—how a particular article is written—may involve copyright issues. This is actually the core of the dispute between OpenAI and The New York Times. But what we do is cite the truths that already exist in these channels and summarize and synthesize them for users in the context of the search experience.

Therefore, people need to distinguish between two situations: one is training AI to use proprietary content; the other is AI using this content only as a source and providing answers without actual training occurring.

We have made this clear in our response. We also emphasize that only if the journalism ecosystem remains open and prosperous can our products continue to develop and improve. Because we do need to generate real-time information every day. If there is a lack of proper economic incentives, it will be detrimental.

Therefore, our solution is to profit from advertising and share a portion of that advertising revenue with publishers. We have created a publisher program similar to Spotify's revenue-sharing model, rather than simply paying for the licensing of content for a certain period On the contrary, we propose a long-term model: as our usage scale and revenue grow, we will share revenue with publishers based on query revenue sharing.

Publications like Fortune, Time Magazine, Dersh Pico, and WordPress have already joined this program. In the coming weeks, we will announce more partners. We are very confident that this program will ultimately gain widespread recognition in the news industry.

Additionally, we have provided funding to Northwestern University to support research on how tools like ours can help journalists write better.

All journalists need to fact-check, and we are a great fact-checking tool. I believe the current turbulent situation will gradually calm down. In a year or two, we will have a system that allows different groups to thrive together.

Aislin Roth

Then ask a related question. Earlier today, we talked about the academic roots of Perplexity and the importance of citations. So, how do you respond to the accusations of plagiarism against you by journalists?

Aravind Srinivas

That's right, this involves a deeper definition of "plagiarism." If you do not attribute sources, it can easily be seen as plagiarism. However, we always attribute sources, which makes it difficult for plagiarism accusations to hold.

Moreover, we are not copying content verbatim. Although AI can sometimes be unreliable and may produce instances of more than three or four words being repeated, which can raise debates about "paraphrasing" and "synthesis."

But our goal is to summarize and synthesize diverse sources while ensuring that all original sources receive proper attribution. We do our best to control these AIs to ensure clarity in citations and attribution.

Aislin Roth

I really like the Spotify analogy you mentioned. It is indeed a rapidly changing way.

Aravind Srinivas

Indeed. If we profit from advertising revenue and you are the content source, we will share that revenue with the content source. Google has never shared advertising revenue with publishers because they provide traffic, but the advertising revenue is generated on their platform.

The only way for publishers to profit from traffic is through another Google product, AdSense, by placing pop-up ads on their own websites. This is also why many users dislike directly visiting news websites, as these sites are filled with ads, and users have to close many pop-up ads.

This system of relying solely on traffic for profit is unsustainable. We need to create something that users truly want. Additionally, we provide publishers with free APIs to help them build AI-native products and chatbots on their own websites.

If users only want to access the articles they have written, they can ask questions directly on their websites. We also provide our tools for free to the employees of these news organizations. This model can create economic sustainability for them while enhancing the overall experience Aislin Roth

The future is indeed full of imagination. If we take a step back and consider the greatest tech companies in history, almost every one of them is a category creator, like Uber, Facebook, Airbnb, and Salesforce. So, ten years from now, when people look back at this moment, what kind of historically defining company do you think you are building?

Aravind Srinivas

I would say that if we can help people answer all their questions and complete all their tasks, we will definitely be among that group. We are gradually getting closer to becoming a reliable answer machine. I know you previously asked, "Will AI make mistakes?"

My answer is: Yes, we still make many mistakes every day. But if you look at it from a broader perspective, as the models continue to improve and we optimize our network coverage, the frequency of these mistakes will decrease from 1/10 or 1/100 to 1/1000 or 1/10000. We will achieve an order of magnitude improvement.

If we can become a reliable answer machine that everyone can widely access, not only providing answers but also helping users complete tasks, such as making transactions, purchasing goods, booking flights, and finding the best prices, making users' lives more efficient while saving time. I believe we will become a product and company that defines the industry.

Aislin Roth

That's really exciting. I hope that in a few years, we can see you achieve these goals. We have much more to discuss about Perplexity, but I want to take a few minutes to talk about you personally and your leadership style. In just two years, you have grown from a "lean" founder to the CEO of a $9 billion AI company. What has your leadership journey looked like at these different stages?

Aravind Srinivas

I try to keep upgrading myself, but to be honest, I am not yet an experienced and mature CEO. However, I would say that I place a strong emphasis on action and do my best to encourage everyone in the company to have that attitude.

I believe it is this action-oriented culture that has helped us maintain efficiency even as our team has grown to 100 people. A founder I greatly admire once told me that when your team exceeds 100 people, you are destined to become slow. At that time, I was very determined to prove him wrong.

So far, we have done quite well. Of course, as the team continues to expand, we will encounter more issues related to scale, such as how to maintain speed. I am determined to address these issues and hope my solutions will be helpful to others.

Additionally, I give opportunities to those who may not yet be experts in a particular field to try things they have not yet proven their abilities in.

I believe that there is no need to hire the former head of growth from Instagram to be the head of growth or product at Perplexity. Many companies fall into this "trap" of thinking they need to poach talent from other top consumer companies, but I have not done that I prefer to hire those with potential who have not yet achieved their "first major accomplishment."

I really like a saying, "Rowing against the current, if you do not advance, you will retreat." Although this saying is not my original, it is very applicable.

I hope more people can try this approach, letting newcomers learn to swim in the water instead of always choosing the most well-known experts. One main reason is that most people find it difficult to regain the drive to invest a lot of energy and time after achieving significant success. So I would choose those who have not yet achieved great success but are very motivated.

At the same time, I adhere to an action-oriented philosophy, personally trying and understanding the product. I use Perplexity an average of at least 10 times a day, and there are even users who use it more than I do, which makes me very happy.

I believe this helps me make more informed decisions. If you stop using the products developed by your own company, it is easy to become disconnected from reality and make decisions based solely on others' feedback. It is crucial to stay as close to the source of facts as possible.

When people complain on social platforms about "this feature is not usable" or "that feature has problems," I am happy to personally engage in customer support. I am not saying we do not need a customer support team, but when you experience the user's frustration yourself, you gain a deeper understanding of the issues.

I also complain to engineers and product managers about my own products, such as "this part should be faster," so that you are not always limited to whiteboard and strategic discussions, but can use your own product for an extended period, leading to better decision-making.

Aislin Roth

How to maintain a lean style during expansion is a great topic. Before we enter the audience Q&A session, I have one last question. This year's theme for our "Top Perspectives" event is "Leave Your Mark." So Arvind, how do you hope to be remembered?

Aravind Srinivas

I hope Perplexity, or rather I and Perplexity, will be seen as a company that helps the world become smarter. If people using Perplexity feel they have become smarter, learned new knowledge, and can go to sleep with more wisdom than when they woke up in the morning, I would be very happy because that is not easy.

Most consumer products ultimately waste people's time. I myself am addicted to some products, spending a lot of time on them, but ultimately feel bad, as if I wasted several hours.

But Perplexity is not such a product. I do not want it to be a product that makes people waste time. Even with our "Discover" feature, users have told me that they learn new things while browsing, and I hope this experience can continue.

Additionally, I hope we can help people accomplish more. Not everyone can afford an assistant, whether it is an administrative assistant or a personal assistant.

I remember during my internship at OpenAI in 2018, Sam Altman had a fireside chat with Bill Gates. Sam asked Bill, what will the world be like after AGI emerges? Bill's response was interesting; he said, "It would be like living my life."

What he meant was that with AGI, everyone could live like a billionaire. If I want to learn about a topic, I don't need to read myself; instead, I can have others read for me and prepare reports and presentations.

If I need to go somewhere, I would have a private jet and a team to handle travel, dining, and fitness arrangements. This kind of life feels like using a "cheat code."

I believe that if AI can truly understand you and help you plan, book, and complete those tedious tasks online, your life could become as easy as a billionaire's. If we can become one of those tools, I would feel like I left a good mark in this world.

Aislin Roth

If Perplexity could let me live like Bill Gates, I would be very happy too. Now, we enter the audience Q&A session. If you have questions, please raise your hand. Selected audience members can stand up and state their questions and grade.

Audience

Arvind, first thank you. You are a young leader admired by many of us. I am a fan of "Atomic Habits." In your journey from a PhD student to CEO of Perplexity, what is one habit you have given up? And what is a new habit you have learned that has helped you become the leader you are today?

Aravind Srinivas

One habit I have given up—though I'm not sure if it was "necessary" to give up—but I did give it up, is sleeping in. I feel this has helped me a lot, giving me more time. It also means I need to go to bed earlier.

I haven't woken up after 8 AM for the past two or three years, no matter which city I'm in. This habit of waking up early makes me feel like I have more time every day. Another habit is trying to work out three times a week. I didn't exercise often before, but now I make an effort to do so.

Audience

Thank you for being here today, Arvind. My name is Raben, and I am a second-year MBA student. I completed a dual degree at the Indian Institute of Technology Bombay, majoring in Analytics Engineering. I found your previous answer about technological barriers very insightful, but I would like to hear what you think is the biggest risk or challenge companies currently face?

Aravind Srinivas

Today's challenge? I think it's the same as what I mentioned earlier. Almost all startups trying to scale inevitably become slower when their team size reaches hundreds or even thousands.

You would expect that with an increase in team size, project progress would at least be linear, but executing multiple projects with high quality becomes very difficult at the same time. When quality declines, users notice and feel that the product has regressed and become worse.

This is what is called "anxiedification": when you try to expand your business and user base, the quality of the product declines, especially for those early loyal users who liked you because of the product quality. Therefore, I think this is the biggest challenge we face Audience

This is an extension of Aislin's previous question regarding the ethical issues faced by Perplexity. Now that you have personally experienced quite a bit of criticism in these areas, as a leader, you must have spent a lot of time thinking about these issues.

I am curious about how you handle these ethical issues. Whose opinions do you refer to when addressing these problems? Is there an example that illustrates how your stance on certain important issues has changed?

Aravind Srinivas

Thank you very much for your question. I think there are some great people within the company, for example, our publisher program is the brainchild of our Chief Business Officer, Dimitri.

So, one thing I've learned is that just because I'm the CEO doesn't mean I have to solve every problem. If someone is better than me at something, I should trust them to get it done.

One thing I used to not believe much in is that many problems can be resolved by communicating with the other party and explaining what you are doing.

For example, Forbes expressed dissatisfaction with some of our practices, such as page displays. But when I actually met with someone who criticized me on Twitter and explained what we were doing, he at least shook my hand and said, "Oh, I never understood that this is what you were doing."

So, I think we have more work to do. I haven't met with all the community members who have opinions about us, but I feel that when controversies arise, the first step should be to reach out and communicate directly, which is something I intend to continue doing.

Audience

Hello, I am Varuna, a first-year MBA student, and a computer science graduate from the Indian Institute of Technology Kanpur. You are a great source of inspiration for me. I strongly resonate with your point about helping people find the answers they are looking for.

But when you mention suggesting questions through advertisements, such as "Why are Adidas sneakers better than Nike," does this affect the way you present information to users? For instance, this might not be the answer that users are truly looking for.

Aravind Srinivas

Yes, but that is a question, not an answer. That is just a suggested question. If you choose to explore that question further, the answer will be objective and not influenced by Adidas or any brand.

Moreover, our goal is to understand each user more deeply in the future, making these suggested questions more personalized, rather than simply sponsored questions chosen by brands. There is evidence that, for example, many people find ads on Instagram very relevant and are therefore willing to purchase products through it.

So, I think "relevance" is the key to making ads work, while ensuring that the ads do not interfere with the core value of the product. If users ask any questions, the answers should always be unaffected by ads and trustworthy, then the product fulfills its value However, to provide such high-quality products that almost never make mistakes while allowing the company to maintain profitability during scale expansion, we indeed need to find an intelligent commercialization balance.

With Newin, original title: "In-depth | Perplexity CEO Stanford MBA Entrepreneurship Dialogue & Insights: The Future of AI Search is Not 'Google', but Completing Tasks for Users"

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk