Bing’s new ‘Deep Search’ feature offers more comprehensive answers to complex search queries

Microsoft Bing is getting a caller “Deep Search” characteristic powered by OpenAI’s GPT-4. The characteristic is designed to springiness users much applicable and broad answers to analyzable hunt queries. Microsoft notes that Deep Search is not a replacement for Bing’s existing web search, and is alternatively an enhancement that allows for deeper exploration of nan web.

In a blog post, Microsoft explains that nan caller characteristic builds connected Bing’s existent web scale and ranking strategy and enhances them pinch GPT-4, which takes nan hunt query and turns it into a much broad explanation of what nan results should include.

For instance, opportunity a personification is looking into loyalty programs successful different countries and enters nan query: “how do points systems activity successful Japan.” Deep Search would return nan query and grow it into nan following:

Provide an mentation of really various loyalty paper programs activity successful Japan, including nan benefits, requirements, and limitations of each. Include examples of celebrated loyalty cards from different categories, specified arsenic convenience stores, supermarkets, and restaurants. Show a comparison of nan advantages and disadvantages of utilizing loyalty cards versus different costs methods successful Japan, including existent rewards and benefits. Highlight nan astir celebrated services and participating merchants.

With this expanded description, you are capable to explicate your intent amended than you could pinch conscionable a fewer words.

In instances wherever your hunt query is much ambiguous, Deep Search will find each of nan imaginable intents and creates a broad explanation for each of them. Deep Search past displays these intents to you, allowing you to prime nan correct one.

Once nan expanded explanation has been created, Bing will propulsion applicable results that often don’t show up successful emblematic hunt results, Microsoft says. Deep Search finds pages that mightiness lucifer nan expanded query, rewrites nan query and past searches for those variations too.

Sticking pinch nan aforesaid loyalty points query illustration above, Deep Search whitethorn besides hunt for things for illustration “loyalty paper programs Japan,” “best loyalty cards for travelers successful Japan,” “comparison of loyalty programs by class successful Japan,” “redeeming loyalty cards successful Japan” and “managing loyalty points withe telephone apps.”

“By doing this, Deep Search tin find results that screen different aspects of my query, moreover if they don’t explicitly see nan original keywords,” Microsoft wrote successful nan blog post. “Regular searches connected Bing already see millions of web pages for each hunt and Deep Search does 10 times that to find results that are much informative and circumstantial than nan ones that rank higher successful normal search.”

Once Deep Search has collected web pages, it ranks them based connected really good they lucifer nan expanded description. When ranking, Deep Search considers really good nan taxable matches nan query, whether it has an due level of details, really trustworthy nan page is and really caller and celebrated it is.

Microsoft notes that Deep Search is optional and tin return up to 30 seconds to complete, which is why it’s not meant for each query aliases user. For users who don’t want much broad answers, they tin get regular Bing hunt results instantly.