MYworldfix

Google made a significant announcement on Wednesday as it introduced its first-ever AI agent capable of taking actions on the web. This research prototype from DeepMind, named Project Mariner, holds the potential to revolutionize the way we interact with websites. The Gemini-powered agent takes control of the Chrome browser, moving the cursor, clicking buttons, and filling out forms, mimicking human behavior. Google is initially releasing this AI agent to a small group of pre-selected testers, marking the beginning of a new era in web interaction.

Google's AI Agent Transforms Web Interaction

How Project Mariner Works

After setting up the AI agent with a Chrome extension, a chat window appears on the right side of the browser. Users can instruct the agent to perform various tasks such as creating a shopping cart from a grocery store based on a given list. When in use, the agent navigates to the grocery store's website, in this case, Safeway, and searches for and adds items to a virtual shopping cart. However, it's important to note that the agent is relatively slow, with about 5 seconds of delay between each cursor movement. At times, it may stop its task and revert back to the chat window to seek clarification about certain items.Behind the scenes, the agent takes screenshots of the browser window, which users must agree to in the terms of service. These screenshots are sent to Gemini in the cloud for processing, and Gemini then sends instructions back to the computer to navigate the webpage.

Other AI Agents by Google

In addition to Project Mariner, Google unveiled several other AI agents for specific tasks. Deep Research aims to assist users in exploring complex topics by creating multi-step research plans. It competes with OpenAI's o1 in multi-step reasoning but is not designed for math, logical reasoning, or data analysis. The agent is rolling out in Gemini Advanced today and will be available in the Gemini app in 2025. When faced with a difficult or large question, Deep Research creates a multi-step action plan, takes a few minutes to answer, and generates a detailed report.Jules is another new AI agent designed to help developers with coding tasks. It integrates directly into GitHub workflows, allowing it to view existing work and make changes directly in GitHub. Jules is being rolled out to a select group of beta testers today and will be available later in 2025.Finally, Google DeepMind is working on an AI agent to assist in navigating video games, building on its expertise in game-playing AI. The company is collaborating with game developers like Supercell to test Gemini's ability to interpret gaming worlds such as "Clash of Clans." Although no release date was provided for this prototype, it is expected to have a significant impact on the gaming and virtual world landscape.It remains unclear when Project Mariner will be available to Google's massive userbase, but when it does, these AI agents will undoubtedly have a profound impact on the broader web. The web was designed for human use, but Google's AI agents may change that paradigm.