Google unveils Gemini 2.0: A significant step towards the era of AI agents

Google has launched Gemini 2.0, a major upgrade to its artificial intelligence model. It introduces new capabilities such as audio and image generation.

 Google Gemini (photo credit: SHUTTERSTOCK)
Google Gemini
(photo credit: SHUTTERSTOCK)

Google announced Gemini 2.0, the next generation of its AI model, approximately ten months after launching version 1.5. Although it is still in an "experimental preview" phase, Google DeepMind CEO Demis Hassabis describes it as "a flexible and reliable model with low latency and enhanced performance at the cutting edge of technology."However, it’s worth noting that Google is initially offering the smallest model in the series, Gemini 2.0 Flash, with other models expected to launch soon.

Gemini 2.0 introduces new capabilities, including built-in audio and image generation, and primarily serves as a foundation for AI agents—bots that can autonomously perform tasks. Google has already demonstrated such projects, like Project Astra, a visual system capable of identifying objects and assisting with navigation.

In addition, Google unveiled Project Mariner, an experimental Chrome extension that can "use the browser for you," which is also slated for release in early 2025. There is also Jules, an agent that helps developers find and fix code, and another Gemini 2.0-based agent designed to assist with video games. Hassabis sees 2025 as "the true beginning of the agent-based era," with Gemini 2.0 as its cornerstone.

In fact, Google plans to integrate Gemini 2.0 into all its products, including Google Search, the Gemini app, Workspace, and more. The goal is to create a general-purpose model with diverse capabilities. With the advent of the agent era, new issues are emerging, particularly around safety.

What is an AI Agent? An AI agent is a smart software capable of performing tasks independently, similar to a virtual personal assistant. It "learns" from its environment, understands instructions, and can make decisions to achieve specific goals. Think of it as an intelligent robot that can perform actions without being told exactly what to do at every moment—it understands intent and navigates independently.

How can autonomous agents be prevented from causing harm? Google is taking precautionary measures with Mariner and Astra, but Hassabis acknowledges the need for further research and new safety solutions, such as "sandbox testing." Gemini 2.0 is already available in the Gemini app online. Early next year, it is expected to roll out to additional Gemini platforms and all Google products.