View From The Top with Aravind Srinivas, Cofounder and CEO of Perplexity

Introduction and Background

• Aravind Srinivas is welcomed to Stanford, and he mentions he is from Berkeley, representing with blue, and is happy to be at Stanford 15s.
• Many in the audience are active Perplexity users, especially with free Perplexity Pro for all Stanford students, and are excited to have Aravind Srinivas there 27s.
• To craft questions, Perplexity was used, and it suggested discussing Aravind's personal backstory, the early days of Perplexity, the company today, and various leadership lessons 1m12s.
• Perplexity also suggested asking about something the audience may not know about Aravind, the funniest thing about him, and questions to inject humor into the conversation 1m46s.
• Rapid-fire questions were also suggested to be asked at the end of the interview 2m0s.
• Perplexity provided interesting insights, including Aravind's love for cricket, teaching himself to program after missing getting into computer science at IIT Madras by 0.01 points, and his connection to Sundar Pichai, the CEO of Google, who grew up in the same hometown in Chennai, India 2m26s.
• Aravind attributes the success of many tech entrepreneurs from Chennai to the culture of trying to excel and do their best, with a strong emphasis on education and valuing being scholarly and well-read 3m36s.
• He also mentions that people from Chennai are known for being obsessed with cricket statistics 4m23s.
• Growing up in India, cricket was a significant part of life, and knowledge about the game was valued over wealth, with an emphasis on understanding statistics and consistency in performance 4m36s.
• The decision to pursue a PhD in computer science at UC Berkeley was made after not getting accepted into Stanford, and this academic background has shaped the approach to building Perplexity 5m3s.

The Influence of Academia on Perplexity

• The experience at UC Berkeley, where citations were a crucial aspect of academic work, influenced the development of Perplexity, with a focus on providing sources for every answer to establish trust and authority 5m28s.
• The concept of citations and academic currency in the community was learned through working on papers and understanding the importance of simplicity and citability in ideas 6m32s.
• The trade-off between writing complicated, creative ideas and simple, citable ones was a key takeaway from the academic experience, and this has been applied to Perplexity's approach to providing answers 6m54s.
• The inspiration for Google's search engine, which used academic citation graphs and web hyperlinks, also influenced the development of Perplexity's approach to providing trustworthy answers 7m25s.
• Perplexity's unique product experience is built on the idea of providing answers that are backed by sources with domain authority or trust, similar to academic papers, and this is a direct result of the academic roots and experience 8m17s.

Perplexity's Mission and Vision

• Democratizing access to knowledge is crucial as it is a moral duty for individuals to seek wisdom and become perpetual learning machines, and having access to the right tools is essential for making progress in understanding the world better 9m7s.
• The goal of Perplexity is to improve access to information and create the world's first answer engine, making knowledge widely accessible for all, allowing people to ask any question and get an instant answer 8m43s.
• The tagline of Perplexity is "where knowledge begins" because there is no end to knowledge, and the company aims to help people keep getting better by providing them with the tools to do so 9m35s.
• The increasing efficiency and affordability of AI models will make it possible to create a widely accessible version of the answer engine, helping people ask any question and get an instant answer 10m11s.

Building the Perplexity Team

• When building the initial founding team of Perplexity, the qualities looked for were people with complementary skills, who were better in their areas of expertise, and did not overlap with the founder's skills 11m21s.
• The core founding team of Perplexity consisted of Aravind Srinivas, Dennis, and Johnny, who brought together a combination of skills in AI, software engineering, problem-solving, and competitive programming 11m37s.
• Johnny, one of the co-founders, was a world-class competitive programmer who represented the United States at the II and had a strong background in problem-solving and AI 11m47s.
• Dennis, another co-founder, had a strong background in AI and software engineering, which complemented the skills of the other founders 12m20s.
• The academic background of the founders helped them meet like-minded individuals who were motivated and deep thinkers, which was beneficial in building the team 10m49s.
• The founding team's diverse skills allowed them to take bold risks and set up a mission to build a completely new search experience, which would have been impossible to achieve alone 12m24s.
• Over time, the team hired more people with new skills, such as front-end programming and writing Kura kernels, which had an incremental and multiplicative effect on the company's growth 12m39s.
• The team's design was improved by hiring someone specifically for that role, who brought a new perspective and skills, creating a multiplicative effect on the company's product 13m31s.

Funding and Investment

• When raising their series A financing round, the team found out that Microsoft was launching a search competitor, Bing, but their investors, Nea, expressed confidence in the team and the company, which gave them the confidence to continue 14m11s.
• The investors, Nea, called to reassure the team that they believed in them and would not back out of the deal, despite the news of Microsoft's launch, which was a crucial moment for the company 15m37s.
• The team's response to the news was not to worry, but to focus on finding a way to move forward and continue growing, thanks to the confidence boost from their investors 15m52s.
• Perplexity's investors include Jeff Bezos, Yan Lecun, and Nvidia, and the company's success in attracting these investors can be attributed to creative fundraising strategies 16m11s.
• Yan Lecun, also known as the "Godfather of AI," was initially difficult to reach, but the founders managed to meet with him by camping out in front of his office at NYU, and they impressed him with a Twitter demo that allowed him to search his own tweets and see who was replying to him 16m36s.
• The founders also impressed other investors, such as Karpathy, by sending them a link to try the product directly, rather than sending a deck, which was a strategy that worked well for them 17m41s.
• The key takeaway from these experiences is that it's essential to play to your strengths and not try to do something you're not good at, such as making decks, and instead focus on creating a working product that can be tried instantly 17m58s.
• Having a working product that can be tried instantly is more effective than having a deck, as it communicates more and shows that the product works, which is essential for investors 18m28s.
• The founders have not used decks much, even for their series A and B funding rounds, and instead prefer to write memos or Notion documents, which is a strategy that has worked well for them 18m52s.
• Successful decks from the past, such as those from Airbnb, LinkedIn, and Facebook, can be confusing and make it difficult to know how to create an original deck, which is why the founders have avoided trying to make decks 19m11s.

Perplexity's Approach to AI Models

• Perplexity is building an answer engine without owning the content or models, instead relying on using APIs and post-training them for specific tasks like summarization and referencing. 20m8s
• The decision to use other people's models was made due to the conviction that models would become increasingly commoditized, requiring an enormous amount of funding to develop and maintain. 20m51s
• The company chose not to compete in building their own models, as it would require losing billions of dollars per year, and instead focused on shaping existing models for a better end-consumer experience. 21m6s
• This approach has proven to be correct, as many companies that attempted to build their own models are no longer in existence, highlighting the need for significant funding or a different approach. 21m33s
• The cost of using APIs for models is decreasing rapidly, with a 2x reduction every four months, making it a good time to be an application layer company using these models. 22m14s
• Open-source models are keeping a check on closed-source models, bringing the price down and increasing the level of intelligence and reasoning, making it a perfect time to build on top of these models. 22m30s
• Perplexity aims to create a successful business by building on top of existing models and technologies, similar to how other successful companies have done in the past, such as Coca-Cola leveraging refrigeration technology. 23m2s
• The company's goal is to create something that provides immense value to consumers by using the right packaging and foundational technology, making it worth building and leveraging existing technologies. 23m31s

Advertising and Monetization

• Perplexity has introduced advertising for the first time, with a strategy that differs from Google's overreliance on advertising, aiming to avoid influencing the accuracy and truthfulness of answers 23m50s.
• Perplexity's ad unit is designed to suggest follow-up questions after providing unbiased and truthful answers, allowing brands to get users' attention without manipulating the original answer 24m56s.
• The ad unit is still in its early days, with a few brands experimenting with it, and the major concern is the lack of control over the answer, which requires courage for brands to try out this new style 26m2s.
• Perplexity is clear on not trying to influence the accuracy and truthfulness of answers, to avoid ending up like Google, where people are frustrated with the answers 26m31s.

Addressing Legal and Ethical Challenges

• The company is handling recent challenges, including a lawsuit from News Corp for copyright infringement and a cease and assist order from The New York Times for inappropriate content use, with a vision for ethical AI development 26m57s.
• Perplexity believes in its approach to AI development, as stated on its blog post, and is committed to handling these challenges while continuing to innovate 27m18s.
• No one has copyright or ownership over truth or facts, and this applies to journalism as well, where referencing existing information from other sources is allowed, as long as the specific expression of truth is not copied verbatim 27m24s.
• The specific way something is written can have a copyright angle, which is relevant to the Open AI and New York Times scenario, but Perplexity is referencing truth that already exists in outlets and summarizing and synthesizing it for the user in a search experience 28m4s.
• There is a difference between AI that trains on proprietary content and AI that uses sources to give answers without training, and Perplexity falls into the latter category 28m23s.
• Perplexity relies on an open and thriving ecosystem of journalism to survive and improve, and the company needs real-time information created every day, which is why it wants to support publishers financially 28m44s.
• To address this, Perplexity will share ad revenue with publishers through a program that is not a short-term licensing model, but rather a long-term model where revenue is shared on a query level basis as the company scales 29m2s.
• The program is inspired by how Spotify shares revenue, and several publishers, including Fortune, Time, and WordPress, have signed up to be part of it, with more partners to be announced in the coming weeks 29m31s.
• Perplexity has also made grants to Northwestern University to research how tools like theirs can help journalists write better and do fact checks more efficiently 29m50s.
• The company is confident that its program will resonate with the journalism community and help both parties flourish together in the future 29m45s.
• Perplexity handles allegations of plagiarism by always attributing sources, which makes it hard to claim plagiarism, and the company is trying its best to synthesize information rather than reproduce it verbatim 30m39s.
• The goal is to summarize and synthesize information from a diverse set of sources, giving credit to the original sources, and controlling AI systems to the best possible extent 31m8s.
• A revenue-sharing model is being implemented, where ad revenue is shared with the original source, unlike Google, which only provides traffic but keeps the ad revenue 31m31s.
• This approach aims to create a sustainable system where users can access information without being bombarded with ads, and content creators can monetize their work more effectively 32m7s.
• APIs are being offered for free to journalists and websites, allowing them to build AI-native products and chatbots, and to create a system that is economically lucrative for them 32m18s.

Perplexity's Long-Term Goals and Vision

• The company's vision is to become a reliable answer machine, helping people get answers to their questions and accomplish tasks, and making transactions more efficient 33m24s.
• The goal is to become a category-creating company, like Uber, Facebook, and Airbnb, and to be a history-defining company in the future 32m52s.
• The company's leadership journey has involved rapid growth, from a scrappy founder to a CEO of a $9 billion AI company, and the leader is still learning and upgrading their skills 34m27s.
• A bias for action is encouraged in the company to maintain speed, even with a growing team of around 100 people, with the goal of solving the problem of scaling while staying fast 34m53s.
• The company avoids hiring experts who have already had major successes, instead opting for talented individuals who have not yet had their first major hit, as they tend to be more motivated and driven 35m42s.
• This approach is based on the idea that people who have already achieved success may struggle to push themselves for further success, and may not be as motivated to put in the effort required for a new challenge 36m47s.
• The company's approach to hiring and talent development is focused on giving people opportunities to try new things and take on new challenges, rather than relying on established experts 35m33s.
• The CEO uses the product extensively, averaging at least 10 queries a day, to stay close to the source of truth and make informed decisions 37m17s.
• This approach helps to identify customer frustrations and user pain points, and allows the CEO to provide feedback to engineers and product managers to drive improvements 38m8s.
• The company's focus on staying scrappy and agile at scale is driven by a desire to maintain a fast and innovative culture, even as the team grows 38m33s.
• The goal is for Perplexity to be known for helping the world become smarter, with users feeling they have learned something new after using the platform, making them wiser and more knowledgeable 38m52s.
• Most consumer products tend to waste people's time, but Perplexity aims to be different, with features like the Discover feed allowing users to learn something new while scrolling through it 39m43s.
• The ultimate goal is to make assistance and personal help accessible to everyone, similar to how billionaires have access to executive assistance, by leveraging AI to help people with mundane tasks and planning 39m52s.
• This vision is inspired by a fireside chat between Sam Altman and Bill Gates, where Gates described a future with AGI as being able to live a life similar to his, with access to information and assistance at all times 40m6s.
• The idea is to make life easier for people by providing them with tools that can truly understand and help them, making their lives more manageable and freeing up time for more important things 41m10s.
• The founder's goal is to leave a mark in the world by creating a tool that helps people live a better life, and if Perplexity can provide a life similar to Bill Gates', he would be very happy 41m37s.

Q&A with Aravind Srinivas

• The founder is open to audience questions and is willing to share his experiences and insights with the audience 41m46s.
• When asked about his journey from PhD student to CEO, the founder is asked to share a habit he had to let go of and a new one he learned that helped him become a leader 42m2s.
• Waking up early has been a helpful habit, with the goal of getting more hours in the day, and going to bed early to achieve this, with the last time waking up later than 8:00 a.m. being at least two to four years ago 42m32s.
• Regular workouts have become a priority, aiming for at least three days a week, which is a change from not working out much before 43m8s.
• The biggest risk or challenge facing the company is the difficulty in scaling while maintaining quality, as it gets harder to execute well without a drop in quality when the number of people increases 43m48s.
• This challenge is referred to as "edification," where the product gets worse in quality for initial loyalists as the business scales 44m32s.
• As a leader, seeking opinions from others, such as the chief business officer Dimitri, is crucial in addressing ethical issues, and trusting others' instincts when they are better suited to solve a problem 45m23s.
• Engaging with people on the other side of an issue and educating them on the company's goals can be an effective approach, as seen in the instance with Forbes, where a critic changed their stance after being explained the company's intentions 46m9s.
• The goal is to help people find the answers they're looking for, and introducing ads in suggestions doesn't affect the unbiased nature of the information presented to users, as the answers to those suggested questions are still unbiased and uninfluenced by ads 47m15s.
• The key to making ads work without interfering with the core value of the product is relevance, and evidence shows that people find relevant ads, such as those on Instagram, useful and often make purchases as a result 48m16s.
• To bring a high-quality product to scale, the company needs to figure out intelligent sweet spots of monetization, but the core value of the product remains that any question asked will have an answer that is uninfluenced by ads and always truthful 48m40s.
• As the co-founder of Perplexity, the question that is most perplexing is not explicitly stated, but rather leads to a round of Rapid Fire questions 49m11s.
• If not the CEO of Perplexity, the alternative career path would likely be AI research, which was the previous field of work 49m54s.
• The all-time favorite Cricket moment is when India won the World Cup in 2011 50m6s.
• The preferred Tech Visionary to have dinner with, dead or alive, is Larry Page 50m18s.
• The strangest search query seen on Perplexity is someone buying a face mask with an opening only for the eyes, possibly for skiing or another specific context 50m36s.
• A user's search query was analyzed, which involved biking in cold weather, and the intent behind the query was to find a product that could cover the face, keep the user warm, and allow them to breathe while only having an opening for the eyes 51m7s.
• The user's search query was ambiguous, as it could also be interpreted as planning a heist, but the actual intent was related to biking in cold weather 51m1s.
• The search results provided to the user were effective, as they ended up buying a product directly from the search results 51m25s.
• The conversation with Aravind Srinivas, Cofounder and CEO of Perplexity, came to a close, with appreciation expressed for his participation 51m34s.