Friendly Insights on How to Use Perplexity

Thank you for joining us on our journey exploring the world of artificial intelligence and its resources. We continue our adventure with Perplexity, the AI-powered knowledge assistant, taking a deep dive into how the tool works and its underlying technology. 

In starting our adventure, we take a look at the front page of today’s Perplexity front page:

At first glance, we see a clean, minimalist design with intuitive navigation. Right in the middle of the web screen is its search bar, inviting users to “Ask anything…” Below the question space, we see a setting for language preference and a paper clip icon for uploading documents and images for analysis. In addition, there is an “Auto” setting, providing options for users to search their inquiries through “Auto” (best for daily search), “Pro Search” (3x more sources), “Deep Research” (in depth reports), and “Reasoning with R1” (DeepSeek’s new model only on US servers). Important to note that the integration of DeepSeek R1’s removes Chinese censorship constraints and utilizes DeepSeek’s advanced research capability, as it displays real-time “thinking chains” during responses, showing immediate steps and citations in its logical process. As I asked Perplexity a question on how DeepSeek R1 compared to other AI models, I got to witness it’s reasoning process, step by step, shown below.  

Continuing our discussion on how Perplexity works, we return to the first image and see a clickable banner for Deep Research. This advanced feature refines searches with comprehensive results, typically advertised to academics and professionals. Scrolling further, Perplexity presents a news feed of trending topics, weather, and market updates. Again, Perplexity presents itself as more than just a search assistant, but a high-powered informational hub. Lastly, the navigation bar at the bottom of the interface ensures a seamless user experience with “Home” (default landing page), “Discover” (space for trending topics), “Spaces” (community-driven space for idea creation and collaboration) and “Library” (place to save past searches, responses, and bookmarked items).  

Lifting the hood up on the Perplexity platform and its magic, we find Perplexity is a hybrid between a search engine and an AI assistant. Perplexity utilizes large language models (LLMs) and real time web access to generate responses. In particular, Perplexity makers have confirmed using OpenAI’s GPT model, Anthropic’s Claude, and DeepSeek within its Perplexity model (Perplexity Team, 2024). It's an amalgamation of the AI giants! In maintaining real time web access, Perplexity fetches and cites external sources through RAG (retrieval-augmented generation). It web scrapes and indexes live sources to verify facts instead of relying purely on model memory (which is more reminiscent of ChatGPT). Finally, in its platform performance, Perplexity, similar to Google, uses keyword-based searches to provide results, displaying them with natural language understanding and follow-up questions for further research investigation. Again, we definitely value the hard work Perplexity does to help us sift through all the buzz to find well-sourced results to our everyday questions. Tune in next week for our follow up post on how Perplexity is used in real-world scenarios!           

References

Perplexity Team. (2024, October 1). Getting started with Perplexity. Perplexity. https://www.perplexity.ai/hub/blog/getting-started-with-perplexity