- Perplexity AI provides users with streamlined, traceable, high-quality answers by integrating knowledge from multiple sources, innovative algorithms, and human intervention to achieve powerful knowledge integration.
- Users can ask questions in natural language, and Perplexity AI understands the context and provides humanised answers, rather than simply listing links to search results.
- Perplexity AI demonstrates rapid product iteration through frequent version updates, rapid response to new technologies, timely optimisations, and startup agility.
Perplexity AI provides a new AI-powered search and Q&A experience by integrating knowledge from multiple sources, natural language interaction, and rapid iteration.
Powerful knowledge integration
Perplexity AI can collect and integrate relevant information from various sources on the internet (including academic papers, news reports, and social media) to provide users with comprehensive and accurate responses.
Unlike traditional search engines, Perplexity AI focuses on providing answers with source citations and clearly labelled citation locations. Free-form dialogue is avoided, making the search experience more direct and precise.
Perplexity AI ranks based more on semantic information and the quality of the content itself, independent of factors such as SEO (search engine optimisation) and advertising.
Perplexity AI uses RAG (Retrieval Augmented Generation) technology, which first uses the search engine to “recall” the most relevant links from external indexes, and then synthesises and submits user questions and retrieves answer to a large model, which completes the task of integrating and generating answers.
The algorithms are innovative in both recall and sorting to ensure the usefulness of the content and the accuracy of the references.
The Source Edit function allows users to edit the reference sources and research, effectively reducing the interference of irrelevant sources on the results. The search results include not only links but also ChatGPT-style Q&A, combining the breadth of knowledge of traditional search engines with the reasoning and generating power of large models.
Also read: Perplexity’s ongoing fundraising spree
Natural language interaction
Perplexity AI employs technologies that mainly include large-scale language models and natural language processing, which enable it to understand the semantics of a user query and provide the most relevant answer based on context, an interaction that is more intuitive and efficient.
What makes Perplexity AI innovative compared to existing search engines is that it is the world’s first search engine to incorporate dialogue and links, which allows it to recognise and respond to more ambiguous or abstract language to mimic the verbal queries of most people.
Perplexity AI is particularly good at searching and integrating knowledge in the academic domain, supporting researchers by quickly retrieving and summarising the latest research findings.
Also read: xAI open-sources base model of Grok without any training code
Rapid iteration
Perplexity AI’s product iteration is fast, allowing it to incorporate the latest AI advances promptly. In less than a year, Perplexity AI has made several key iterations. In a period of frequent updates, new features are introduced almost weekly, and these updates focus on the functionality level of the product.
The Perplexity team has a keen product insight into search. Half a month after the launch of GPT-3.5, they went live with a SQL-based (Structured Query Language) Twitter search analytics engine, and Perplexity introduced innovative features such as Source Edit, Focus Search, and more to address the shortcomings of the answer engine, such as outdated information and irrelevant results.
