- Perplexity describes itself as a versatile tool for exploring information and satisfying curiosity, combining the capabilities of a chatbot with those of a search engine. It could be likened to a blend of ChatGPT and Google Search, though it doesn’t replace either directly. In essence, it embodies the direction Google aims to take with Gemini, but with a more streamlined approach.
- The functionality of Perplexity is akin to a chatbot: users pose questions, and it provides answers. However, it also seamlessly retrieves information from recent articles. It scans the web daily, allowing users to inquire about current news, sports scores, and other common search topics.
- Although Perplexity cannot fully replace traditional search engines yet, it is surprisingly functional and effective within its scope.
Perplexity AI has quickly become a leading name in the realm of AI-driven search engines, carving out a significant role within the rapidly evolving sectors of search technology and artificial intelligence. With its innovative approach and formidable capabilities, Perplexity AI is not just transforming generative search experiences but also capturing the attention of people globally.
In this article, we delve into key facts that shed light on the development, reach, and potential of Perplexity AI, getting a deeper insight into detailed understanding about Perplexity AI.
What is Perplexity AI?
Perplexity represents a fusion of advanced language models geared towards enhancing web searches beyond the capabilities of conventional search engines. At its core lies a proprietary layer, harmonizing various foundational models including OpenAI’s GPT-3.5 and GPT-4, Anthropic’s Claude 2, and Google DeepMind’s Gemini (formerly known as Bard).
Its standout feature lies in its adeptness at swiftly scouring the web and amalgamating diverse sources of information to furnish comprehensive results. Moreover, it boasts a suite of convenient features for refining source selection, coupled with multimodal capabilities, enabling analysis beyond mere text. Additionally, users benefit from a copilot option, facilitating interactive engagement with the Perplexity LLM layer to refine queries and enhance result accuracy.
Also read: Perplexity’s ongoing fundraising spree
Also read: How to create a large language model (LLM)?
Key features of Perplexity AI: Focus and Related
When conducting online searches, many of us append “Reddit” to our queries, such as “best backup software reddit,” to steer clear of low-quality sites rife with affiliate links. Similarly, I found solace in Perplexity’s Focus feature.
This functionality empowers users to designate specific sites for search, including Reddit, which Perplexity then consolidates into its responses. Selections span the entire web, Wolfram Alpha, Academic Writing, YouTube, and Reddit. Opting out of searching is as simple as selecting “Writing Mode,” allowing the LLM to generate responses independently. Currently, the available sites for selection are somewhat restricted, but I anticipate expansion in the future.
Another commendable aspect of Perplexity is its “Related” feature. Following the provision of its primary output, Perplexity presents a series of prospective follow-up queries under this category.
In moments of brainstorming or ideation, determining the conversational trajectory can prove challenging. Here, Perplexity steps in with follow-up inquiries, occasionally proposing astute questions that might have eluded consideration. Furthermore, I appreciate that responses are intrinsically linked to the search process by default.
However, the efficacy of the “Related” questions varied considerably. While some prompted intriguing avenues of exploration, others seemed at odds with my intentions, offering follow-up queries tailored for the company rather than pertinent search terms for Perplexity itself. An example arose when soliciting suggestions for questions to pose to a women’s health startup, where Perplexity’s suggested follow-up queries veered off course.
How does Perplexity AI work?
Perplexity harnesses the prowess of several large language models (LLMs) to fuel its natural language processing capabilities. Among its roster are GPT-4, Claude 3, Mistral Large, and Perplexity’s bespoke models. These LLMs serve the dual purpose of comprehending user queries and distilling pertinent answers.
In tandem, Perplexity boasts an in-built search engine tasked with sourcing and indexing relevant information. The company asserts Perplexity’s daily scouring of the internet, yet its ability to provide real-time updates—such as live soccer scores—suggests instantaneous indexing for select content.
Perplexity offers two search modalities: Quick Search, geared towards swift, fundamental responses, and Pro Search, adept at discerning nuanced inquiries and tailoring bespoke responses. The latter even engages users with follow-up questions to refine its outputs.
Irrespective of the search type employed, Perplexity follows a consistent workflow: parsing user queries, deciphering intent, scouring the web for pertinent sources, and presenting concise summaries. For instance, querying about Zone 2 training benefits prompts Perplexity to identify the subject as moderate aerobic exercise benefits, scour authoritative health and fitness resources, and furnish a succinct synopsis.
Though Quick and Pro searches yield akin information, the latter delves deeper and proffers tailored recommendations based on specified exercise preferences.
Furthermore, Perplexity operates akin to a conversational agent, preserving the conversational thread (referred to as Threads) to consider prior interactions.
Notably, Perplexity furnishes users with a reference list, accompanied by footnotes delineating the sources for each key datum. This feature distinguishes it from conventional search engines, affording users the opportunity for deeper exploration beyond the AI-generated summary.
What can you do with Perplexity AI?
If you’ve ever thought about using Perplexity for your queries instead of Google, Bing, or another search engine, you’re onto something. It’s often more reliable than a chatbot and quicker than a typical search engine when it comes to delivering useful answers.
Consider this: while chatbots are often suggested for planning trips and finding recipes, most tests reveal their mediocrity in both areas. Perplexity, on the other hand, taps into multiple web sources, not just a single LLM, which tends to yield better results. Plus, you can delve deeper into the articles it summarizes.
For instance, when I asked Perplexity for a spaghetti bolognese recipe for eight people, it didn’t disappoint. It not only provided a workable recipe but also adapted it for four people. What’s more, it got the little details right, like listing ingredients in the order they’re used. And after checking the sources, I found no evidence of direct copying from any single recipe. While I can’t guarantee every recipe will be perfect, Perplexity is likely to guide you in the right direction and offer links to reliable recipes.