It was amazing to organize an Icons dinner with Aravind Srinivas, the founder and CEO of Perplexity, and hear your vision for reinventing search with LLMs. Thanks so much for sharing your insights with us, Aravind! Thanks also to Silkroad Innovation Hub for hosting and Effective Soft for sponsoring the dinner. Key takeaways below:
Perplexity’s Product Vision:
Perplexity has a great opportunity to leverage its leading LLM-powered search engine to build the best general-purpose, personalized assistant for consumers by expanding its data sources. Aravind imagines layering in voice-enabled AI agents for shopping, appointment booking, and much more.
The future of AI is multi-model: Perplexity leverages a combination of small and large models to answer user questions depending on complexity, maximizing the tradeoffs behind accuracy, latency, and cost. Choosing how to stitch models together and route the right question to the right model will be critical for enterprises seeking to leverage AI going forward.
Imagine Perplexity as the expert network of the future. In response to a user’s question, a team of AI agents could email, conduct voice calls with, and compensate several human experts, and then analyze the calls to deliver a response to a user for a fixed price.
Advice for Founders:
Start a company in a space where you have an asymmetric advantage.
Prioritize building a great product, not focusing on competition.
Be patient – recognize it may take several iterations of your original idea to find product-market fit as it did for Perplexity.
Don’t be afraid to go against conventional wisdom. When Aravind founded Perplexity, it was contrarian to believe you could create durable value as an AI company without training your own foundation model.