Perplexity AI: online search now accesses reliable information

 For Jeff Bezos, Perplexity AI will disrupt the way we search for information. The founder of Amazon is so convinced of this that he invested in it by participating in a consortium of investors which injected 73.6 million dollars into the startup which, for the moment, does not have a head office clean and brings together less than forty people. collaborators in a coworking space in San Francisco.

To understand what Perplexity AI is, how it works and because it arouses so much interest, especially from people very experienced in tech, it is appropriate to start from the hypotheses and objectives of the Californian startup whose origins go back moreover in a handful of months.



Perplexity AI: the new frontier of online research with artificial intelligence

Perplexity AI is a chatbot and conversational search engine that leverages OpenAI's GPT-4 and combines it with a large language model (LLM) that uses natural language (NLP) and machine learning to understand how context in which a question is asked is more precise.

In simple terms: Perplexity AI is a search engine that presents itself as a chatbot and offers the possibility of being queried sequentially in order to better contextualize the query (typically a question) formulated by the user.

The linguistic models on which Perplexity AI is based rely on the web to offer answers and rely on open source technology from Mistral and Meta (Llama2).

Brief History of Perplexity AI

The startup was founded in August 2022 by Andy Konwinski, Aravind Srinivas, Denis Yarats and Johnny Ho, a quartet with a history of developing large language models. which aims to broaden access to knowledge . An expression that requires further study, since it refers to search engines that, in fact, already participate in this extension.

Being able to combine natural language processing, large language models and the Web coincides with greater contextualization of user questions and, therefore, with the possibility of providing rapid answers even in the face of complex and detailed questions. What currently sets Perplexity AI apart from its many competitors is precisely its ability to achieve a high degree of depth. thanks to the use of linguistic models and the possibility of exploiting data extracted from the Web thanks to a search system based on Bing, Microsoft's search engine.

All of this prepares for the speed with which Perplexity AI is able to return results and the reliability of the results.

Perplexity in the context of natural language analysis

As mentioned, Perplexity AI searches the web and leverages natural language and machine learning to return accurate and complete answers.

To analyze the web, it uses a large language model to identify content consistent with the question asked by the user also assessing the reliability of the sources queried with the aim of reducing hallucinations, i.e. The use of facts and data that have no basis in reality.

Users can ask natural language questions that Perplexity AI responds to in natural language, adapting the content of the answers to the level of knowledge with which the questions were formulated. The follow-up process allows you to delve deeper into a topic through more detailed and in-depth questions.

How Perplexity AI works

When a user asks a question, Perplexity AI searches the web, selects the most relevant pages, then large linguistic models take over and extrapolate the most accurate information to the question.

This sequence of steps aims to represent a clever proportion between the typical functions of a search engine and those of a language model: on the one hand the need to consult several links (typical of search engines) and on the other hand the speed of execution of a language model which can however be superficial .

Video

To increase the reliability of answers, Perplexity AI was trained to highlight citations – that is, the same mechanism that rewards academic articles – also providing a set of relevant questions aimed at delving deeper into the topic requested by the reader. 'user.

Users are an active part of it, so much so that they can delete – among those returned by Perplexity – information that turns out to be incorrect, unfounded or insufficiently corroborated by the data.

Practical strategies for integrating Perplexity AI into business processes

There are several ways to integrate Perplexity AI into business processes but, for the avoidance of doubt, to achieve satisfactory results the organization must have a solid understanding of the benefits conferred by the use of AI in general . Thinking that Perplexity AI can adapt to internal business needs without the right culture is risky and misleading.

One of the areas where Perplexity AI can help is the research phase: scanning different sources while being able to count on reliable results gives a boost to market analysis and monitoring activities, i.e. say competitive analysis. More generally, this can be invaluable for extracting value from data, leveraging speed and efficiency. .

Entering the field of marketing, Perplexity AI can help content creators. However, there are employment strategies that delve deeper into business processes:

  • research and development, thanks to the possibility of identifying ideas in relation to market needs;
  • process automation;
  • customer support, providing quick and precise answers to questions asked;
  • internal support, facilitating the transfer of knowledge and information within the company, answering specific questions asked by employees.

Whatever you intend to use it for, the data must be consistent and of good quality.

How much does Perplexity AI cost?

The Perplexity Pro version is available at a price of 20 dollars per month (around 18.30 euros) or 200 dollars per year (183 euros) if you opt for early payment.

The free version has some limitations but it is a good way to test its qualities.

Comments