Benchmarking Bing Chat: Baidu Search’s small-scale public beta “conversation” function, based on the Wenxin Yiyan language model

According to news from IT House netizens on May 9, Baidu Search has recently begun a small-scale public test of the generative AI "dialogue" function, which is based on Baidu's Wen Xin Yi Yan Da Language model.

This product is based on Baidu Knowledge Enhanced Large Language Model Wen Xinyiyan, benchmarking Microsoft search engine Bing and integrating OpenAI company ChatGPT New Bing after service.

According to news from the Brand Public Relations Laboratory, the current testing channels for Baidu AI dialogue are Baidu main website and Baidu App, and the independent website is Chat.Baidu.com. Users who use this service need to have and log in to Baidu account . Currently, users who are not included in the test cannot access the URL normally. After entering the page, "404 Not Found" will be displayed, and the corresponding entrance will not be displayed when accessing Baidu main website and Baidu App.

Baidu AI dialogue provides two input options: voice input and text input, and also provides a wealth of online applications. The AI ​​Dialogue Inspiration Center shows that the product supports capabilities including Excel formula editor, Vincentian diagrams, classical Chinese translation, recipes, dream interpretation, AI poetry composition, legal consultation, company naming, AI signature requesting and other capabilities.

Based on Baidu AI dialogue's own answers, it integrates the search results of the entire network based on the large model, provides more credible answers, helps you create and provide inspiration, including social media copywriting/generated pictures /Write weekly reports, and assist in decision-making based on users’ individual needs, such as providing travel guides, etc.

On May 4, Robin Li, founder, chairman and CEO of Baidu, revealed at the internal "New Mission Sixth Anniversary and Baidu Pride Award Ceremony" that Wenxinyiyan was under internal testing for more than a month and completed After four major technology upgrades, the cost of large model inference has been reduced to one-tenth of the original, or the inference performance has been improved by nearly 10 times.

The above is the detailed content of Benchmarking Bing Chat: Baidu Search’s small-scale public beta “conversation” function, based on the Wenxin Yiyan language model. For more information, please follow other related articles on the PHP Chinese website!