Deep Research: Google's Agentic Feature Transforming Online Search
On Wednesday, December 11, 2024, Google not only announced the launch of Gemini 2.0 Flash, the first model in the Gemini 2 family, but also introduced Deep Research, an agentic feature integrated into Gemini Advanced designed to enhance online searches.
Google describes this new feature as "a personal research assistant." It leverages Google's expertise in finding relevant information on the web to guide the navigation and queries of the Gemini model, which boasts advanced reasoning capabilities. Its pop-up window has been expanded to accommodate up to 1 million tokens, allowing it to process large amounts of information all at once.
How to Use Deep Research?
Subscribers to Gemini Advanced can access this feature from the dropdown menu of the model selector under "1.5 Pro with Deep Research." Users simply submit a query, and the AI creates a multi-step plan that the user can approve or modify before conducting a thorough analysis. In just a few minutes, it generates a comprehensive report, complete with links to original sources.
For instance, Google illustrates this with a scenario involving a student preparing a presentation on autonomous vehicle sensors. Once the plan is approved, the user initiates the search, and the tool begins an in-depth analysis of relevant information from across the web.
Google explains:
"In just a few minutes, Gemini continuously refines its analysis, navigating the web like you would: searching, discovering interesting information, and then starting a new search based on what it has learned. It repeats this process several times and, once complete, generates a comprehensive report of the key findings, which you can export to a Google Doc."
Initially, Deep Research will be available in English on Gemini Advanced via the web, with plans to launch on the mobile app in early 2025.