ByteDance's AI assistant, Doubao, recently underwent a functional upgrade, introducing a new "Deep Thinking" mode currently in limited testing. This mode innovatively combines reasoning and information retrieval, allowing users to "think and search" simultaneously for a more comprehensive information gathering experience.
Traditional AI search typically involves a comprehensive information search followed by analysis. However, this single-search approach struggles with complex problems, failing to encompass all relevant factors. For instance, purchasing decisions, industry analysis, or academic research often require integrating multi-dimensional information, making a single search insufficient. Doubao's new mode overcomes this limitation.
In "think and search" mode, Doubao conducts multi-round searches while reasoning. This "retrieval + reasoning" combination enables the AI to provide more accurate and efficient answers to complex problems requiring external data. For example, when planning a trip from Beijing to Shanxi, Doubao can use three rounds of search to find not only basic attractions and transportation information, but also integrate the latest weather forecasts and traffic conditions to create a more detailed itinerary.
Furthermore, it can handle ambiguous search queries, such as "Several animals fighting, one is good at dismantling things, what old cartoon is it?", efficiently finding the answer using the "think and search" mode. This new feature provides users with richer and more accurate feedback during information retrieval.
Doubao's "Deep Thinking" mode is somewhat similar to xAI's Grok3 "Deep Search" function, but Doubao integrates reasoning and deep search for a more unified user experience. Moreover, this Doubao upgrade is completely free, offering users a more efficient search and thinking experience.
This new Doubao feature not only improves search efficiency but also provides powerful support for daily decision-making and research.
