Perplexity emerges as an innovative and unique tool in today’s digital world, filled with various LLMs chatbots. At first glance, its tone and language might resemble other platforms, like ChatGPT, but Perplexity isn’t just another tool for writing or creative brainstorming. Its true purpose lies in providing well-informed, accurate answers to your questions. What sets Perplexity apart as the pioneering answer engine is its ability to offer responses that are not only insightful but also backed by reliable internet sources.
Think of Google and its web crawlers that scour the internet, adding pages to an index and presenting them based on a sophisticated algorithm every time you search. Perplexity differentiates itself by utilizing its sophisticated AI technology for searching and synthesizing information, crafting answers in a detailed, article-like format tailored to your queries. After using Perplexity for several days, I’ve discovered that it could be a handy tool for gathering relevant sources, saving the effort of browsing through the dozens of websites that Google typically displays. I’m excited to share an in-depth analysis of Perplexity and explore its real-world utility in this article.
A Unique Approach to Search
Perplexity AI is an AI answer search engine that uses existing LLMs to generate a response based on your input prompt, but with its own AI system, it compiles internet links that answer your question. Their pro plan allows switching between a couple of well-known L.L.M.s such as Google’s Gemini, Mistral 7bi, Anthropic Claude 2.1, and OpenAI GPT-4. In simple words, it is like Google but simplified, in the sense that it does not give back hundreds of links so you can choose the one you think would give you the answers; instead, it summarizes the links that match your search the closest and gives you a well-structured paragraph that answers your question.
This response addresses your query and includes relevant links, allowing you to explore the topic further if the sources align with your needs. Following this, the platform’s AI enhances user engagement by offering summaries replete with source citations, primarily websites and articles. This feature encourages users to delve deeper, posing follow-up questions to understand the subject better.
On January 4th, Perplexity AI announced that it raised $73.6 million in a funding round led by IVP—additional investments from Databricks Ventures, NEA, Nvidia, and Jeff Bezos. According to TechCrunch, the round values Perplexity at 520 million, and even though it might seem like a lot for many startups, this is chump change when it comes to gen AI startups, as they said in their article.
Hands-On With Perplexity
I don’t want to regurgitate the same things that TechCrunch, Forbes, or other articles have said about Perplexity and its usefulness. Rather than taking others’ word for it, the best approach is to personally try Perplexity AI for an hour or two, focusing on your specific needs to gauge its performance. This hands-on experience is crucial, much like what happened with ChatGPT and other large language models.
These tools transitioned from mere technological novelties to essential assets because users found practical applications beyond the initial exploration and fun. Their value emerged as they began assisting in daily work and creative endeavors. Far from being mere playthings, these large language models proved their worth in real-world tasks, integrating seamlessly into various aspects of work and creativity. Their success can be attributed to their user-friendly interfaces and sustained usefulness, a trait uncommon in new tech products.
I incorporated Perplexity AI into my everyday search and internet experience for the last few weeks. I used Perplexity as intended: to improve and, in the best-case scenario, replace conventional search engines, most commonly known as Google, since no one uses Bing. Perplexity’s CEO, Aravind Srinivas, stated that “their product exists to find answers to any question, and that is why they refer to Perplexity as an answer engine.”
As I was testing Perplexity to search for the best musical instrument to go with my setup, the engine replied with a series of options, each with its respective citation. Perplexity may provide a list of options as an answer, but at times, it offers several paragraphs, each accompanied by a footnote citing its source. Srinivas said that his academic background shaped the design of Perplexity, noting that every statement in a paper requires a citation for credibility. The output from Perplexity is akin to a well-researched paper. At the same time, it might not always offer the precise answer desired, but it is valuable for further exploration into its sourced information.
Optimizing Your Use Of Perplexity AI
The best way to use Perplexity is by combining ChatGPT and Perplexity. You can craft the perfect prompt with ChatGPT and then feed it to Perplexity to create a good draft on the thing you need, and then you can check if Perplexity can back the output. Even though generative AI tools are helpful, these tools are just tools and no more than that.
Thousands of software and apps have changed every industry, but I don’t think we are close to observing a significant change in society. It just feels like the new technology that, sooner or later, would appear in the tech world, such as when smartphones came to the public. Smartphones revolutionized our interaction with the internet, society, and the world, marking a monumental technological shift. However, there’s a prevailing notion that the changes brought by generative AI will surpass even this, leading to transformations unlike anything we’ve seen before. I don’t think this is the case yet, but I would like to see how these companies and tools can make me change my mind.
How to solve the shortcomings of Perplexity
Some of the challenges with tools like Perplexity differ from the platform itself but stem from the nuances of how LLMs function. Given that Perplexity leverages multiple LLM models, it’s crucial to fully understand their use to harness their capabilities. A significant hurdle with LLMs is that many users become discouraged after a while, mainly because these tools yield different results. This issue arises because efficiently utilizing an LLM requires crafting a prompt tailored to your task.
As I mentioned, the key to unlocking Perplexity’s potential lies in developing the perfect prompt for your desired outcome. More than merely inputting simple, vague prompts won’t cut it; such an approach leads to generic results that fail to meet specific needs. One improvement I’d suggest for Perplexity (and many other LLM chatbots) is including detailed documentation on crafting effective prompts. While this may initially create some friction for users, it’s a strategy that will pay off in the long run.
Developers, understanding the intricacies of their AI models, can guide users in forming prompts that elicit the most accurate responses. This process may be intuitive, as users often seek specific answers to their unique queries. However, by guiding users towards particular types of prompts for specific tasks, developers can provide a starting point from which users can refine their queries, thus streamlining their journey to achieving their goals. Using the tool itself to do this is feasible; for instance, while hundreds of prompt libraries are available for AI model usage, a library curated by developers would be far more beneficial, as they can create guides on refining your prompt to achieve the desired result iteratively.
Another pitfall with Perplexity is its occasional inaccuracy. When attempting to correct it, the tool sometimes rephrases the same incorrect information. I assume it conducts a new search within its AI model and either needs help finding helpful information or misinterprets the query as a repetition. A potential solution could be that if a user inputs three or four prompts without getting the desired answers, Perplexity could start asking clarifying questions about what the user is seeking. Through these inquiries, it could either develop a more specific prompt tailored to the user’s needs or, eventually, provide the sought-after answers directly. A collaborative effort is essential to harness Perplexity’s capabilities fully: users must refine their query skills while the AI improves response accuracy.