Former OpenAI artificial intelligence research scientist Aravind Srinivas cofounded Perplexity on 3 August 2022 together with former Meta Platforms AI researcher Denis Yarats and computer scientist and Databricks cofounder Andy Konwinski. Their goal was to address the problem of sifting through a myriad of information on the internet and offer a more engaging, reliable, and intelligent way for people to search and discover information.
Perplexity Ask was launched on 7 December 2022. It was rebranded to Perplexity on 8 February 2023. This service is both a search engine product and a chatbot or an answer engine that uses large language models to provide answers to complex questions. It is also a prime example of generative artificial intelligence that uses retrieval-augmented generation to summarize search results with citations instead of listing links to online references.
Pros: Advantages of Perplexity and Notable Applications
This service is one of the notable products that emerged during the initial wave of the artificial intelligence revolution that began in 2022 due to the demonstration of the applications and advantages or importance of large language models and the expansion of practical generative artificial intelligence applications. It competes with ChatGPT from OpenAI, Bing Chat and Copilot from Microsoft, Gemini from Google, and HuggingChat from Hugging Face.
Below are the notable advantages and applications of Perplexity:
1. Searched-Focused Generative AI Chatbot
The main selling point and advantage of Perplexity over search engines such as Google Search and generative AI chatbots such as ChatGPT and Google Gemini is that it is an AI chatbot designed for web search. It combines the usefulness of search engines and the generative capabilities of AI chatbots. This is similar to the selling point of Bing Chat or Microsoft Copilot and somewhat to the extended web search plugin or extension available on ChatGPT Plus.
2. Runs on LLMs and RAG Framework
Powering this service are various open-source and closed-source LLMs such as GPT, Mistral, and LLaMA. These are based on the transformer architecture and autoregressive modeling. Their large training datasets and parameters enable advanced natural language processing applications. A retrieval-augmented generation framework using the proprietary PPLX online models is built on top of the LLMs for better information retrieval and reduced hallucination.
3. Interactive and Pedagogical Web Search
The combination of LLMs and RAG or the searched-focused generative AI chatbot selling point defines the more specific advantages of Perplexity. It provides its users with an interactive and conversational web search experience. This makes it ideal for research purposes or for collecting reference materials for electronic and paper trailing. It also has specific and various modes for writing, solving math problems, helping with coding, and video search.
4. Straightforward and Intuitive User Interface
Perplexity has the same clean and simple user interface as the more popular generative chatbots like ChatGPT, Google Gemini, and Copilot. What sets it apart from the competition is that its minimalist design principle still strikes a balance between simplicity, power, and functionality. The interactive elements such as buttons and icons demonstrate full functionalities or features and capabilities of Perplexity without being too cluttered and overwhelming.
Cons: Disadvantages of Perplexity and Key Limitations
The strengths or advantages of Perplexity are demonstrated through search-related applications that require real-time data. However, when it comes to long-form content generation and processing extensive bodies of text, other AI chatbots or conversational AI such as ChatGPT have better capabilities. Take note that both ChatGPT and Google Gemini also excel in providing faster responses and are equipped with better multimodal language models.
Below are the key disadvantages and limitations of Perplexity:
1. Language and Conversational Limitations
One of the notable disadvantages of Perplexity is that it lacks a natural language flow. Responses are accurate most of the time but can sometimes feel robotic because of the absence of fluid and natural-sounding human conversation. It is also dependent on facts or information. This means that is less adept at handling subjective questions or generating casual conversations. It is also prone to some of the limitations of LLMs such as limited context understanding.
2. Search and Response Shortcomings
It tends to fall short in explaining or answering questions within the realms of more complicated science. This generative AI chatbot also lacks problem-solving capabilities and is not suitable for planning. Furthermore, similar to the pressing problems with conversation artificial intelligence, responses can either be too short or inconsistent. The breadth of search is limited. Search results are not guaranteed to reference the best materials. It also mixes up information.
3. Not Suited For Analytical Applications
Another disadvantage of Perplexity is that it is not intended for applications that require analysis. It still has intermediate analytical capabilities. It can analyze and summarize a text, interpret an image or graphics, and process content from various modalities such as video. However, when compared with the GPT-4 model in ChatGPT and the Gemini model of Google Gemini, it does not produce more insightful analyses. This chatbot is built for search first and foremost.
4. Capabilities of the Free and Paid Versions
This generative and conversation AI chatbot is free. A user can sign up and log in using his or her Google account. There is also a paid or subscription-based or Pro version that is priced at USD 20.00 per month or USD 200.00 when paid in full for a full-year subscription. The Pro version gives users unlimited file upload and access to additional AI models like the advanced GPT-4. The responses are better and more insightful. There is also dedicated customer support.
Rundown: Advantages and Disadvantages of Perplexity
The aforementioned pros and cons of Perplexity make it one of the best generative artificial intelligence applications out there. It has similarities to other generative and conversational AI chatbots but what sets it apart from the rest is that is a search-focused generative application. This makes it a useful tool for research and trailing. It is a viable alternative to traditional search engines. However, for deeper and extensive trailing, search engines such as Google Search or Microsoft Bing remain unparalleled for the simplest reason that these tools have a deeper and wider breadth of search. Perplexity is ideal as a starting point for deeper research undertakings and it also excels in answering or responding to less complex inquiries or popular topics.