Last updated on
The report suggests that OpenAI is developing a search engine poised to rival Google. However, the absence of specific details in the report prompts inquiries regarding whether OpenAI is indeed crafting a self-contained search engine or if there are alternative motives behind the announcement.
According to The Information, OpenAI is in the process of developing a Web Search product intended to rival Google directly. A significant aspect highlighted in the report is its partial reliance on Bing, Microsoft’s search engine. However, beyond this detail, there is a lack of further information, leaving uncertainties about whether it will function as an independent search engine or be integrated into ChatGPT. All reports emphasize its aim to directly challenge Google’s dominance in the search engine market.
Reports indicate that OpenAI plans to utilize Bing search as an integral component of its speculated search engine, leveraging a fusion of GPT-4 and Bing Search technologies, with an intermediary layer to facilitate coordination between the two. In this envisioned framework, OpenAI is not embarking on the development of its own search indexing technology; rather, it relies on Bing for this aspect.
Consequently, OpenAI’s primary focus lies in devising the mechanism through which the search interface seamlessly interacts with both GPT-4 and Bing. This entails the development of an interface that optimizes the utilization of GPT-4’s natural language processing capabilities while effectively harnessing the search functionality provided by Bing.
And that’s a dilemma that Bing has already addressed through the implementation of what Microsoft refers to as an orchestration layer. Bing Chat leverages retrieval-augmented generation (RAG) to enhance responses by integrating web search data to provide context for the answers generated by GPT-4. For further insights into the workings of orchestration and RAG, you can view the keynote delivered by Kevin Scott, Chief Technology Officer at Microsoft, during the Microsoft Build 2023 event, starting at the 31:45 minute mark (available here).
If OpenAI is positioning itself as a competitor to Google Search, what specific tasks remain for OpenAI to tackle that Microsoft, with Bing Chat, isn’t already addressing? Bing boasts a wealth of experience and maturity in search technology, an area where OpenAI lacks expertise.
Is OpenAI directly challenging Google? A more plausible scenario is that Bing, through OpenAI as an intermediary, is mounting a challenge against Google.
ChatGPT has emerged as one of the fastest-growing applications in history, currently boasting approximately 180 million users. It achieved this milestone in just two months, a feat that took Facebook and Twitter years to accomplish.
However, despite this impressive growth, Google maintains a substantial lead that poses a significant challenge for OpenAI. With an estimated user base ranging from 3 to 4 billion worldwide, Google vastly outnumbers OpenAI’s 180 million users.
Even if we assume that all 180 million OpenAI users conduct an average of 4 searches per day, the total daily searches could potentially reach 720 million searches.
Comparatively, Statista’s estimates indicate that Google handles approximately 6.3 million searches per minute, translating to over 9 billion searches per day. This stark contrast underscores the formidable task ahead for OpenAI to catch up to Google’s dominance in the search market.
To effectively compete, OpenAI must offer a compelling and practical product that provides a clear reason for users to adopt it. Take, for instance, the entrenched positions of Google and Apple within the mobile device ecosystem, seamlessly woven into users’ daily routines both professionally and personally. Simply developing another search engine won’t suffice in this landscape.
Realistically, achieving the level of ubiquity and utility akin to these tech giants poses a considerable challenge for OpenAI. Not only does it face stiff competition from Google, but also from Microsoft, Apple, and Amazon, particularly when factoring in the proliferation of Internet of Things applications and devices.
Unlike its competitors, OpenAI lacks the established ecosystem necessary for widespread integration into users’ lives. Therefore, launching a search engine to rival Google isn’t feasible without the requisite momentum and infrastructure to seamlessly embed itself into daily routines.
In academic and patent circles, the formal term for search is Information Retrieval (IR). However, despite extensive searches in repositories like Arxiv.org, research papers authored by OpenAI researchers on information retrieval are conspicuously absent. Similarly, attempts to locate IR-related patents associated with OpenAI yield no results. Notably, OpenAI’s roster of research papers also lacks any studies focused on information retrieval.
This absence of research isn’t indicative of secrecy on OpenAI’s part. Historically, OpenAI has been transparent about its research endeavors, consistently publishing papers detailing the technologies it develops. However, there’s simply no evidence of research into information retrieval within OpenAI’s body of work.
Considering OpenAI’s purported challenge to Google, the absence of any indication of work on search technology raises questions. It seems reasonable to conclude that search isn’t a current focus for OpenAI. There are no discernible signals or indications suggesting that OpenAI is even contemplating the development of a search engine.
Microsoft is actively exploring the integration of Large Language Models (LLMs) into their search engine infrastructure, as supported by substantial evidence.
Several research papers published in 2024 fall within the domains of Information Retrieval, Artificial Intelligence, and Natural Language Computing. For instance:
One study focuses on augmenting human annotation efforts through the utilization of large language models and streamlined batch processing techniques. The aim is to enhance the classification of search queries with the aid of AI technologies.
Structured Entity Extraction Utilizing Large Language Models
This scholarly article unveils a methodology for extracting structured data from unstructured textual sources, such as webpages. Essentially, it involves converting unstructured data, like that found on webpages, into a structured format that can be comprehended by machines.
Enhancing Text Embeddings via Large Language Models (Accessible in PDF Format)
This academic paper delves into a technique aimed at enhancing the quality of text embeddings, which are essential for tasks like information retrieval (IR). Text embeddings refer to the process of generating representations of text that enable algorithms to grasp the semantic nuances and associations among words.
The aforementioned research paper elaborates on the utilization of text embeddings, which are vector representations of natural language encapsulating semantic information. They find extensive application across various natural language processing (NLP) tasks, including information retrieval (IR) and question answering. Particularly in the realm of IR, text embeddings play a crucial role in the initial retrieval stage by efficiently recalling a limited set of candidate documents from a vast corpus through approximate nearest neighbor search techniques.
Moreover, Microsoft has conducted additional research in the domain of search, specifically exploring the synergy between large language models (like GPT-4.5) and search technologies. This implies a potential link between Microsoft’s advancements in search technology and OpenAI’s speculated plans for leveraging such technology in their search engine, should the rumors hold true.
The speculation about OpenAI venturing into the search engine domain surfaced on February 14th. Interestingly, on February 15th, Google revealed the launch of Gemini 1.5, following their announcement of Gemini Advanced on February 8th.
One might ponder whether it’s mere happenstance that OpenAI’s rumored initiative stole the spotlight from Google’s Gemini announcement the very next day. The timing seems quite remarkable.
However, it’s crucial to note that as of now, the existence of OpenAI’s search engine remains speculative.
Original news from SearchEngineJournal
