Perplexity CEO: Disrupting Google Search with AI

The Logan Bartlett Show・2 minutes read

Arvin Shabos, CEO of Perplexity AI, discusses artificial intelligence and competing with Google by providing accessible answers to all users, aiming for a utopian world where time is a luxury. The balance between traditional search engines like Google and answer engines like Perplexity AI highlights the importance of evolving towards personalized and accurate responses for users.

Insights

  • Google's business model can hinder user experience, creating a demand for alternatives like Perplexity AI that focus on accessible answers for all users, regardless of resources.
  • Transitioning from traditional search engines to answer engines poses challenges for algorithms like Google, emphasizing the importance of balancing traditional search with answer engines like Perplexity AI.
  • Perplexity AI aims to provide personalized answers from the entire internet, drawing inspiration from Wikipedia and aspiring to play a significant role in determining truth across complex topics, while also emphasizing the importance of disclaimers for sensitive information.

Get key ideas from YouTube videos. It’s free

Recent questions

  • How does Perplexity AI differentiate from Google?

    Perplexity AI aims to provide accessible answers to user queries, focusing on user experience and personalized responses. The CEO, Arvin Shabos, discusses the importance of transitioning from traditional search engines to answer engines, drawing inspiration from Wikipedia to offer comprehensive information from the entire internet. Perplexity AI's vision aligns with creating a utopian world where time is valued as a luxury, allowing individuals to pursue their passions. The company emphasizes the balance between traditional search and answer engines, contrasting with Google's search-focused business model. By prioritizing user experience and personalized responses, Perplexity AI aims to offer a unique alternative to Google's approach.

  • What challenges does Google face in transitioning its business model?

    Google faces challenges in transitioning from a search-focused business model to a more answer-oriented one, similar to the evolution of Google Cloud. The company's traditional search engine approach conflicts with providing accessible answers to user queries, leading to the need for alternatives like Perplexity AI. The evolution from link-based search engines to answer engines poses difficulties for algorithms like Google, which are accustomed to providing various unrelated information in search results. The shift towards answering specific questions requires a significant adjustment in Google's business model, highlighting the importance of adapting to changing user preferences and technological advancements.

  • How does Perplexity AI aim to enhance user engagement?

    Perplexity AI aims to engage users by providing personalized daily questions and information about famous people, interesting facts, and global events. The app focuses on offering a unique user experience, akin to the difference between radio and Spotify, to keep users engaged even without specific queries. By introducing suggested follow-ups, the app doubled its average session time, emphasizing the importance of user interaction and engagement. The success metric for the app is the number of daily queries, correlating with increased usage and product improvement. Perplexity AI prioritizes accuracy, reliability, latency, UX, and iterative improvement to enhance user engagement and overall experience.

  • What is the significance of citations in academic and AI contexts?

    Citations play a crucial role in academic and AI contexts, serving as a currency in academia and a measure of credibility and reliability. Good papers with many citations are valued similarly to the importance of citations in PageRank algorithms. In AI, citations are essential for establishing the validity and accuracy of information, especially on sensitive topics like suicide, where disclaimers are necessary. Building scholarly tools like Perplexity AI relies on citations to increase knowledge and prevent misuse. The emphasis on citations reflects a commitment to academic rigor and truth-seeking in both academic and AI research.

  • How does Perplexity AI approach hiring and candidate selection?

    Perplexity AI seeks fast learners with diverse project experience and success in different areas, focusing on mentality and culture fit during the interview process. The company values candidates who demonstrate interest in getting things done and align with the company's mission. Coding assessments are essential in evaluating candidates, with at least six or seven interviews conducted to make informed hiring decisions. Perplexity AI emphasizes making generous offers to exceptional candidates, even if it involves internal debate. By prioritizing diverse skill sets, culture fit, and alignment with the company's mission, Perplexity AI aims to build a strong team capable of driving innovation and success in the AI industry.

Related videos

Summary

00:00

"Perplexity AI: Competing with Google for Answers"

  • Arvin Shabos, CEO of Perplexity AI, discusses artificial intelligence and his decision to compete with Google.
  • Google's business model conflicts with user experience, leading to the need for alternatives like Perplexity AI.
  • The importance of providing accessible answers to questions for all users, regardless of resources.
  • The vision of a utopian world where time is a luxury and everyone can pursue their passions.
  • Cultural differences between Deep Mind and Open AI, where Arvin previously worked.
  • The drawbacks of regulating AI too strictly and the considerations in providing answers to potentially harmful questions.
  • The evolution from traditional search engines to answer engines like Perplexity AI.
  • The balance between traditional search and answer engines, with Google leaning towards traditional search.
  • Perplexity AI draws inspiration from Wikipedia and aims to provide personalized answers from the entire internet.
  • The challenges Google faces in transitioning from a search-focused business model to a more answer-oriented one, similar to the evolution of Google Cloud.

16:19

Challenges in Algorithmic Question Answering Systems

  • Jerry West was 63 and a half in 1962.
  • Google search results provided various unrelated information, not answering the question accurately.
  • Chat GPT gave the shortest player in NBA history as an answer, which was directionally right but different.
  • Transitioning from link-based to answering specific questions is challenging for algorithms like Google.
  • Chat GPT's purpose is more about interaction with the model rather than web search.
  • Citations are crucial due to their academic background, seen as a currency in academia.
  • Good papers have many citations, similar to the importance of citations in PageRank.
  • Perplexity aims to play a significant role in determining truth across complex topics.
  • Providing information on sensitive topics like suicide should be accompanied by disclaimers.
  • Building a scholarly tool like Perplexity helps in increasing knowledge and avoiding misuse.

31:16

"Interviewing Fast Learners for Success"

  • Fast learners are ideal candidates, but hard to interview for; look for diverse project experience and success in different areas.
  • Interview for mentality and culture fit, focusing on interest in getting things done and alignment with the company's mission.
  • Coding assessments are essential; conduct at least six or seven interviews to make a decision.
  • Making generous offers to exceptional candidates is crucial, even if it involves debate within the company.
  • Subscription pricing models in the industry are often arbitrary; copying successful models like ChatGPT can lead to widespread adoption.
  • Usage-driven business models tend to be more profitable in the long run compared to subscription models.
  • Free versions of products should continue to improve to drive adoption, while paid versions offer additional value.
  • Going vertical in AI can limit the model's generality and magic, impacting its ability to handle diverse tasks effectively.
  • Access to real-time data is crucial; licensing deals and API access are necessary to integrate data sources for consumer applications.
  • Retrieval augmented generation (RAG) involves pulling relevant documents to enhance AI responses; the approach differs for consumer and B2B applications.

46:23

"App boosts engagement with personalized daily questions"

  • The app provides personalized daily questions and information about famous people, interesting facts, and global events.
  • It aims to engage users even without specific queries, akin to the difference between radio and Spotify.
  • Spotify's challenge is the "cold start problem," where users struggle to decide what to play.
  • Introducing suggested follow-ups doubled the app's average session time from four to eight minutes.
  • The success metric for the app is the number of daily queries, correlating with increased usage and product improvement.
  • Five key dimensions for the business include accuracy, reliability, latency, UX, and iterative improvement.
  • The app's Northstar metric is the number of daily queries, crucial for enhancing accuracy and user engagement.
  • The text discusses the equilibrium among model providers like OpenAI, Anthropic, Meta, and Google.
  • The app is experimenting with a more concise and neutral model, fine-tuned with user data and human annotations.
  • Despite rising GPU costs, the Next Generation chip offers increased compute power in a smaller volume, making it cost-effective.

01:01:58

"Time, AI, Startups: Future Innovations Unveiled"

  • People value time over money as time is a limited resource, and giving people back their time is seen as a luxury.
  • In a utopian world, individuals would prioritize time over money, allowing everyone to pursue activities they love without working excessively.
  • The next big advancements in artificial intelligence are expected to come from better utilization of existing hardware, leading to more efficient and powerful models.
  • Breakthroughs in AI have historically been about optimizing compute power, like the Transformer model improving GPU utilization.
  • The future of AI breakthroughs will likely involve models that can think independently, conduct experiments, draw conclusions, and iterate towards truth.
  • The training process for AI models should involve interaction with the world and generating their own data, rather than just post-training and API delivery.
  • The valuation of startups over $10 billion is significant for founders, employees, and investors, even if it may not have the same impact on larger tech companies.
  • Larry Page's approach to product development, focusing on simplicity and intuition, has inspired the idea of making products that anticipate user needs.
  • Prioritizing accuracy, speed, and readability in product development reflects a commitment to caring for users and continuously improving.
  • Encouraging diverse viewpoints and avoiding complacency in beliefs is crucial for personal and professional growth, as well as fostering a culture of truth-seeking.

01:18:00

"AI, CEO Skills, and Perplexity's Journey"

  • AI is seen as a beneficial tool for humans to improve their lives by working on areas of improvement and seeking truth.
  • The speaker acknowledges their shortcomings in certain CEO-related skills and aims for continuous improvement.
  • Contrasting the cultures of Deep Mind and Open AI, Deep Mind focuses on elegant solutions through deep thinking, while Open AI prioritizes quick problem-solving iterations.
  • Both organizations share a commitment to high-quality outputs and pushing boundaries in their work.
  • The speaker's passion for search led to the creation of a company, Perplexity, with a focus on smart individuals working on challenging problems.
  • Perplexity's journey began with a focus on search-related products, aiming for continual product improvement through user interaction.
  • The speaker believes that the best ideas are often those that seem dumb initially but have a contrarian nature, like competing with Google in the search domain.
  • A pivotal moment for Perplexity was the unexpected increase in user engagement post-launch, leading to further product enhancements and growth.
  • The speaker emphasizes the importance of launching ideas promptly, even if they seem insignificant, as it can lead to valuable learnings and progress.
  • Qualifications and experience from Open AI and Deep Mind helped the speaker secure funding for Perplexity, cutting through the noise of numerous startup pitches.

01:32:29

"Strategies for Success and User Retention"

  • In multiple-choice exams, when unsure of an answer, the strategy is to rule out options first and then convert the problem into an X versus Y scenario, increasing the chances of getting it right to 50%.
  • Success often comes from iterating and giving oneself more opportunities rather than relying solely on genius, as exemplified by Steve Jobs' process of connecting the dots and iterating to create groundbreaking products like the iPhone.
  • Perplexity aims to be a Swiss army knife for knowledge, providing users with comprehensive information and research assistance for everyday decisions, positioning itself as a valuable tool without adopting a negative branding approach like some competitors.
  • The future goal for Perplexity is to achieve high user retention by becoming a daily essential like a toothbrush, addressing challenges such as user awareness, habit changes, and competition with established platforms through continuous iteration and value addition.
Channel avatarChannel avatarChannel avatarChannel avatarChannel avatar

Try it yourself — It’s free.