Google DeepMind introduces Gemini 2.0

Google DeepMind has announced the release of its new artificial intelligence model — Gemini 2.0. The model is capable of performing a wide range of tasks, from understanding complex instructions and planning to navigating websites and providing strategies in video games.

According to the CEO and CTO, the practical application of such assistants opens up exciting new opportunities, being explored through a series of prototypes.

Key projects based on Gemini 2.0

DeepMind has announced several experimental projects, each with unique features:

• Deep Research: a tool for research that helps users explore complex topics. The agent creates multi-step research plans, searches the Internet, and generates detailed reports on findings.
• Project Astra: a universal AI assistant for everyday tasks. It can respond to queries, such as how to wash clothes or find information about a landmark.
• Project Mariner: an agent for navigating the Chrome browser. It can control the cursor, click buttons, fill out forms, and navigate between websites. While the technology is still under refinement, significant improvements in its performance are expected.
• Project Jules: a developer assistant that integrates into GitHub workflows, assisting with coding, planning, and task automation.

AI for video games

DeepMind is also developing agents for video games that can analyze real-time situations, suggest next steps to players, and search for relevant information online. This will greatly enhance gamers' capabilities and make gameplay more efficient.

Future development and prospects

According to DeepMind’s leadership, all projects are in the early stages of development. The company plans thorough testing and refinement before integrating these agents into its products. They expect technologies based on Gemini 2.0 to become widely available in the near future.