Microsoft Bing is getting a brand new “Deep Search” characteristic powered by OpenAI’s GPT-4. The characteristic is designed to offer customers extra related and complete solutions to complicated search queries. Microsoft notes that Deep Search just isn’t a substitute for Bing’s current internet search, and is as a substitute an enhancement that permits for deeper exploration of the online.
In a blog post, Microsoft explains that the brand new characteristic builds on Bing’s present internet index and rating system and enhances them with GPT-4, which takes the search question and turns it right into a extra complete description of what the outcomes ought to embody.
For example, say a consumer is wanting into loyalty applications in numerous nations and enters the question: “how do points systems work in Japan.” Deep Search would take the question and increase it into the next:
Present a proof of how varied loyalty card applications work in Japan, together with the advantages, necessities, and limitations of every. Embody examples of widespread loyalty playing cards from totally different classes, comparable to comfort shops, supermarkets, and eating places. Present a comparability of the benefits and drawbacks of utilizing loyalty playing cards versus different cost strategies in Japan, together with present rewards and advantages. Spotlight the most well-liked companies and collaborating retailers.
With this expanded description, you’ll be able to clarify your intent higher than you could possibly with just some phrases.
In cases the place your search question is extra ambiguous, Deep Search will discover the entire attainable intents and creates a complete description for every of them. Deep Search then shows these intents to you, permitting you to pick out the correct one.
Picture Credit: Bing
As soon as the expanded description has been created, Bing will pull related outcomes that always don’t present up in typical search outcomes, Microsoft says. Deep Search finds pages which may match the expanded question, rewrites the question after which searches for these variations too.
Sticking with the identical loyalty factors question instance above, Deep Search can also seek for issues like “loyalty card programs Japan,” “best loyalty cards for travelers in Japan,” “comparison of loyalty programs by category in Japan,” “redeeming loyalty cards in Japan” and “managing loyalty points withe phone apps.”
“By doing this, Deep Search can find results that cover different aspects of my query, even if they don’t explicitly include the original keywords,” Microsoft wrote within the weblog submit. “Regular searches on Bing already consider millions of web pages for each search and Deep Search does ten times that to find results that are more informative and specific than the ones that rank higher in normal search.”
As soon as Deep Search has collected internet pages, it ranks them primarily based on how effectively they match the expanded description. When rating, Deep Search considers how effectively the subject matches the question, whether or not it has an acceptable degree of particulars, how reliable the web page is and the way new and widespread it’s.
Microsoft notes that Deep Search is non-obligatory and may take as much as 30 seconds to finish, which is why it’s not meant for each question or consumer. For customers who don’t need extra complete solutions, they will get common Bing search outcomes immediately.
