Google introduced ‘Deep Search’ as the representative function of ‘Gemini 2.0’, which is claimed to be optimized for artificial intelligence (AI) agents. It’s a function that searches for complex data and writes articles on behalf of individuals, and solves tasks that will otherwise take hours in minutes.
Google launched Gemini 2.0 on the eleventh (local time) and revealed Deep Search that might be utilized in the paid plan, Gemini Advanced.
It is a feature where AI explores complex topics in your behalf and presents the leads to comprehensive, easy-to-read reports. It combines advanced inference capabilities with a million-token context window to create comprehensive reports that provide useful, easy-to-read insights.
“Earlier this yr, we shared our vision of constructing more agent capabilities into our products, and Deep Search is the primary feature that brings that vision to life,” said Dave Citron, senior director of product management at Gemini. “Utilizing our expertise, we’ve got built a brand new agent system to direct Gemini’s exploration and research,” he said.
When the user gives a command, AI creates a multi-step research plan. Once the user approves or edits it, it begins an in-depth evaluation of related information from the net.
For a couple of minutes, Gemini searches the net, finds vital information, after which starts a brand new search based on its inferences. This process is repeated several times and a comprehensive report of key results is generated.
Subsequently, he said, he connects to related web sites and firms that will not have been found otherwise, and digs deeper to search out out more details. He also said that even when there’s something so as to add or if you should change something, what would take a human being several hours might be done in a matter of minutes.
This might be seen as overcoming the restrictions of the present large language model (LLM), which cannot handle complex tasks without delay. In other words, that is an example that emphasizes that the performance of the agent can only be improved when the performance of the LLM itself increases.
Google said that deep search might be useful for research, market evaluation, and marketing. This feature has been released on Advanced Desktop and Mobile Web, and can even be available on the mobile app early next yr.
In this manner, Google emphasizes that Gemini 2.0 is “a next-generation model built for the brand new agent era.” The response of competitors equivalent to Open AI and Antropic is attracting attention.
Reporter Lim Da-jun ydj@aitimes.com
