Home Tech SearchGPT is OpenAI’s direct attack on Google

SearchGPT is OpenAI’s direct attack on Google

0 comment
SearchGPT is OpenAI's direct attack on Google

After months of speculation about its search ambitions, OpenAI has revealed Search in GPTa “prototype” search engine that could eventually help the company grab a slice of Google’s lucrative business.

OpenAI said the new tool would help users find what they’re looking for more quickly and easily by using generative AI to collect links and respond to user queries in a conversational tone. SearchGPT could eventually be integrated into OpenAI’s popular ChatGPT chatbot. In addition to broader web search, the search engine will leverage information provided by publishers who have signed agreements granting OpenAI access to their data.

Kayla Wood, a spokesperson for OpenAI, declined to provide a demonstration of SearchGPT or an interview about the new tool to WIRED, but confirmed that the company has already given access to anonymous partners and editors and has improved aspects of the search engine based on their feedback.

Microsoft, an investor in OpenAI, was one of the first companies to bring an AI-powered generative search engine to the public when it launched an AI-powered version of Bing in 2023 that was based on OpenAI’s large language models. That Microsoft AI-powered search experience has since been rebranded as Copilot.

Since then, several competitors, such as Google and Perplexity, have launched their own AI search experiences for users. Google’s AI Overviews offer AI-generated summaries of articles, often at the top of news results. OpenAI’s SearchGPT seems most similar to Perplexity’s approach, where the chatbot provides a supplemental list of relevant links and the user can ask follow-up questions.

After OpenAI first introduced ChatGPT in November 2022, early adopters saw the chatbot’s ability to extract and summarize information from the web as a potential replacement for conventional web search. However, the shortcomings of large language models make chatbots imperfect search tools. The models rely on training data that is often months or years out of date, and when they are unsure of an answer, they make up facts.

Microsoft’s first attempts with Bing were unsuccessful, as the AI-powered search engine generated strange, inappropriate and incorrect answers. Bing’s market share grew only slightly after the revamp.

When Google added AI Overviews to search results in May, the company also quickly ran into reliability issues, such as Recommend people to add glue to pizza.OpenAI’s SearchGPT can use a generative AI approach, called Recall Augmented Generation, which is an industry standard for AI search and is designed to reduce the rate of hallucinations in chatbot responses. With a RAG approach, the AI ​​tool references trusted information, such as a preferred news website, while generating its output and links back to the data’s place of origin.

There’s also the question of potential copyright violations. Perplexity in particular has been criticized by publications including WIRED for copying aspects of original journalism with its AI search tool and seemingly ignoring requests not to take content from some websites. In OpenAI’s blog post, the company mentions its commitment to publishers: “SearchGPT is designed to help users connect with publishers by citing and linking to them prominently in searches.” Several companies, including Vox media, The Atlantic, News Corporationand The Financial TimesAll have signed licensing agreements with OpenAI this year.

You may also like