Google also unveiled a new offering on Gemini called Deep Research. This feature is enabled by Gemini 1.5 Pro and enables Gemini to do work by a logical process that involves sequential thinking. It functions somewhat differently from other systems because before running it, it gives the research plan and then prints the same for users.
The main idea behind Deep Research has been to incorporate all the work related to research in one place. Otherwise, it involves cross-referencing a sea of open tabs, finding related information, and incorporating the same into the document.
Gemini’s AI Agentic Feature
Compared to Gemini, Gemini 2.0 is a huge leap forward with a strategy centered on agentic AI tiers, including reasoning, planning, and competent actions as described by Sundar Pichai, Google’s CEO. In contrast to the previous model, where the main focus remained on the expansion of applying aspects of data handling (text, images, etc.).
The new version enables learning AI agents such as Mariner to execute complex operations independently. For instance, its function called “Deep Research” can collect, process, and sum up data for the creation of detailed reports, which reflects the abilities of a rational human assistant.
It commences with the user entering a topic or a question. Deep Research then prepares what is called a research plan, which the user can make changes to or approve. From there, the staking system goes through the World Wide Web like a person does—finding the information, evaluating the results, and searching for additional information based on the discovered content.
Gemini, through its Deep Research, will develop a comprehensive plan that will be refined and validated over several stages of research. According to a blog post, it starts closely scrutinizing related information from the entire web once approved on your behalf.
The major sources of differentiation that can be identified for comparing the two are speed and lightness. Whereas Perplexity is built to make fast answers, Gemini has slower replies, which it believes can provide a better outcome from a search through many sources. It shows that Gemini incorporates over 50 links before giving a reaction, though doing that may take over 2 minutes and 30 seconds.
Deep Research is accessible in Gemini Advanced on the web version of the chatbot in English. Google pointed out that it will be integrated into the mobile application in the first quarter of the year 2025. Deep Research can be opened from the model selector drop-down, and the model’s name to use is ‘Gemini 1.5 Pro with Deep Research.