Google has released a new AI model, Gemini 2.0. It has multimodal capabilities that can process various types of data, such as text, images, and videos, and is optimized for the era of AI agents.
This announcement includes updates to **Project Astra**, a prototype exploring the future of universal AI assistants, and a new approach to exploring human-agent interaction starting in the browser. 'Project Mariner', an AI-based code agent that supports developers.Jules' was included.
Project Astra is an AI assistant based on Gemini 2.0 that understands multiple languages and accents, provides useful help using Google Search, Lens, and Maps, remembers conversations to provide personalized experiences, and boasts fast response times. This feature will be implemented on various form factors, including Google products and Glasses, and the test program will expand.
Project Mariner is an early research prototype based on Gemini 2.0, an agent that understands and utilizes web elements such as pixels, text, code, images, and forms to complete tasks within the browser. It has shown high performance on the WebVoyager benchmark, and Google is conducting extensive research to build it safely. It is currently being tested as an experimental Chrome extension.
The new AI-based code agent, Jules, integrates into the developer workflow to handle Python and Javascript coding tasks. Jules works asynchronously and handles time-consuming tasks like bug fixes, allowing developers to focus on what they actually want to build. Jules creates multi-step plans, efficiently modifies multiple files, and pushes bug fixes directly to GitHub.
Google DeepMind also used Geminai 2.0 to build an AI agent that helps navigate the virtual world of a video game. The agent can reason about the game based on actions on the screen and suggest what to do next through real-time conversation. Google is working with Supercell to test the agent’s ability to interpret rules and tasks in a variety of games.
With Gemini 2.0, Google plans to expand its AI search and assistant to more countries and languages. Google CEO Sundar Pichai said, “The advancement of multimodal technology is a step toward Google’s universal assistant.”1 Gemini 2.0 is now available for web applications, and will be rolled out to more Google platforms and products in the future.
