Official announcements from Google or the corporate partner

Google Unveils Gemini 2.0 with Multimodal AI Agents and Trillium TPU Support

Dec 12, 2024, 05:07 PM

Google has launched Gemini 2.0, its latest artificial intelligence model, introducing a new era of what the company calls 'agentic AI.' This model is designed to perform a variety of tasks autonomously, leveraging advanced multimodal capabilities that allow it to generate images, audio, and text. Key features include native image and audio generation, real-time voice interactions, and the ability to access tools like Google Search, Lens, and Maps. The launch includes several prototype AI agents such as Project Astra, which aims to be a universal AI assistant, Project Mariner, a research prototype for browser-based tasks, and Jules, an AI coding agent. Google's CEO Sundar Pichai emphasized that Gemini 2.0 is the company's most capable model to date, focusing on making information more useful by enabling AI agents to understand and act on behalf of users. The model is being integrated into Google's ecosystem, including its search engine, YouTube, and Android operating system, with plans to expand its reach further. Google also introduced the Multimodal Live API, allowing real-time audio and video-streaming input, and showcased projects like Colab for data science, Deep Research for complex topic exploration, and Gemini 2.0 for Games, which applies AI to video game navigation and strategy. The new model is supported by Google's Trillium TPU, enhancing its training and inference capabilities.