Redefining Search: How Emerging Conversational Engines Overcome Outdated LLMs and Context-Less Traditional Search Engines

The advent of conversational search engines is redefining the way we retrieve information online, shifting from traditional keyword searches to more natural conversational interactions. By combining large language models (LLMs) with real-time web data, these new systems address key issues common to both legacy LLMs and commodity search engines. In this article, we explore the challenges facing LLMs and keyword-based searches, and explore how conversational search engines offer a promising solution.

Outdated knowledge and reliability challenges in LLMs

Large Language Models (LLMs) have significantly improved our methods of accessing and interpreting information, but they face a major limitation: their inability to provide real-time updates. These models are trained on extensive datasets containing text from books, articles and websites. However, this training data only reflects knowledge up to the time it was collected, meaning LLMs cannot be automatically updated with new information. To address this, LLMs must undergo retraining, a process that is both labor intensive and expensive. This includes collecting and managing new data sets, retraining the model, and validating its performance. Each iteration requires significant computing power, energy and financial investments, raising concerns about environmental impacts due to significant carbon emissions.

The static nature of LLMs often leads to inaccuracies in their answers. When faced with questions about recent events or developments, these models can generate answers based on outdated or incomplete information. This can result in ‘hallucinations’, where the model produces incorrect or fabricated facts, undermining the reliability of the information provided. Furthermore, despite their vast training data, LLMs struggle to understand the full context of current events or emerging trends, limiting their relevance and effectiveness.

Another major shortcoming of LLMs is their lack of citation or source transparency. Unlike traditional search engines, which provide links to original sources, LLMs generate answers based on aggregated information without specifying where it comes from. This absence of sources not only hinders users’ ability to verify the accuracy of the information, but also limits the traceability of the content, making it more difficult to discern the reliability of the answers given. As a result, users may find it challenging to validate the information or explore the original sources of the content.

Context and information overload challenges in traditional web search engines

While traditional web search engines remain critical for accessing a wide range of information, they face several challenges that impact the quality and relevance of their results. A major challenge with this Internet search is the difficulty in understanding the context. Search engines rely heavily on keyword matching, which often leads to results that are not contextually relevant. This means users receive a flood of information that doesn’t directly relate to their specific question, making it difficult to sift through and find the most relevant answers. Although search engines use algorithms to rank results, they often fail to provide personalized answers based on an individual’s unique needs or preferences. This lack of personalization can mean that generic results do not match the specific context or intentions of the user. Additionally, search engines are susceptible to manipulation via SEO spam and link farms. These practices can skew results, causing less relevant or lower quality content to appear at the top of search results. As a result, users may be exposed to misleading or biased information.

Rise of a conversational search engine

A conversational search engine represents a paradigm shift in the way we interact and retrieve information online. Unlike traditional search engines that rely on keyword matching and algorithmic ranking to deliver results, conversational search engines use advanced language models to understand and respond to user queries in a natural, human-like way. This approach aims to provide a more intuitive and efficient way to find information by engaging users in a dialogue rather than presenting a list of links.

Conversational search engines leverage the power of large language models (LLMs) to process and interpret the context of queries, enabling more accurate and relevant answers. These search engines are designed to dynamically interact with users, asking follow-up questions to refine searches and offering additional information as needed. In this way, they not only improve the user experience, but also significantly improve the quality of the information retrieved.

One of the key benefits of conversational search engines is their ability to provide real-time updates and contextual understanding. By integrating information retrieval capabilities with generative models, these search engines can pull and integrate the latest data from the Internet so that answers are current and accurate. This addresses one of the key limitations of traditional LLMs, which often rely on outdated training data.

Additionally, conversational search engines offer a level of transparency that traditional search engines lack. They connect users directly to credible sources and provide clear quotes and links to relevant content. This transparency promotes trust and allows users to verify the information they receive, promoting a more informed and critical approach to information consumption.

Conversational search engine versus retrieval augmented generation (RAG)

Nowadays, one of the widely used AI-based information retrieval systems is known as RAG. Although conversational search engines share similarities with RAGs, they have important differences, especially in their objectives. Both systems combine information retrieval with generative language models to provide accurate and contextually relevant answers. They extract real-time data from external sources and integrate it into the generative process so that the answers generated are timely and comprehensive.

However, RAG systems, such as Bing, focus on merging retrieved data with generative output to deliver accurate information. They don’t have follow-up capabilities that allow users to systematically refine their searches. Conversational search engines, such as those from OpenAI, on the other hand SearchGPT, engage users in a dialogue. They use advanced language models to understand and respond to questions naturally, and provide follow-up questions and additional information to refine searches.

Real world examples

Here are two real-world examples of conversational search engines:

• Bewilderment: Bewilderment is a conversational search engine that allows users to interact with online information naturally and contextually. It offers features such as the ‘Focus’ option to limit searches to specific platforms and the ‘Related’ feature to suggest follow-up questions. Perplexity runs on a freemium model, with the basic version offering standalone LLM capabilities and the paid Perplexity Pro offering advanced models such as GPT-4 and Claude 3.5, along with improved search refinement and file uploads.
• SearchGPT: OpenAI recently introduced SearchGPT, a tool that combines the conversational capabilities of large language models (LLMs) with real-time web updates. This helps users access relevant information more intuitively and easily. Unlike traditional search engines, which can be overwhelming and impersonal, SearchGPT provides concise answers and engages users in conversation. It can ask follow-up questions and provide additional information if necessary, making the search experience more interactive and user-friendly. An important feature of SearchGPT is its transparency. It connects users directly to credible sources and provides clear quotes and links to relevant content. This allows users to verify information and explore topics more thoroughly.

It comes down to

Conversational search engines are changing the way we find information online. By combining real-time web data with advanced language models, these new systems address many of the shortcomings of outdated large language models (LLMs) and traditional keyword-based searches. They provide more timely and accurate information and improve transparency by linking directly to credible sources. As conversational search engines like SearchGPT and Perplexity.ai evolve, they offer a more intuitive and reliable approach to searching, going beyond the limitations of older methods.