Google DeepMind has launched SIMA 2, an advanced AI agent designed to mimic human learning and gaming in virtual environments. This release marks a significant advancement towards artificial general intelligence (AGI), with SIMA 2 achieving a 65% task completion rate, a substantial increase from the previous version's 31%. The AI agent can interpret high-level goals, collaborate in games, and apply learned concepts across various environments. Powered by the Gemini system, SIMA 2 can process text, voice, and image inputs, and autonomously generate tasks for iterative learning. Despite its advancements, the AI still faces challenges with long multi-step tasks and visual interpretation in 3D environments, as noted in the research findings.