SIMA 2: An Agent that Plays, Reasons, and Learns With You in Virtual 3D Worlds
Google DeepMind
Google DeepMind released SIMA 2, an AI agent that can play video games, reason about objectives, and improve itself through self-directed play by integrating Gemini's reasoning capabilities. The agent was trained across commercial video games and demonstrated performance approaching human-level play on tasks it had never encountered before, including untrained games like Minecraft and ASKA. SIMA 2 can now explain its reasoning to users, understand complex instructions in multiple languages, and learn new skills independently without human-generated training data in subsequent iterations.
Why it matters
Introducing SIMA 2, a Gemini-powered AI agent that can think, understand, and take actions in interactive environments.
