Google DeepMind is expanding its agent-based artificial intelligence research by embedding its SIMA game-playing system inside the new Genie 3 simulation platform, according to remarks by DeepMind Chief Executive Officer Demis Hassabis at a research briefing on 12 August. SIMA can now take direct control of a wide range of commercial video games, while Genie 3 dynamically generates the virtual environments in which the agent operates. Hassabis said the combination builds on earlier breakthroughs such as AlphaGo and AlphaZero and is intended to give models the ability to plan, reason and learn the physics of materials, liquids and living beings rather than rely solely on language or mathematical data. DeepMind researchers have also released a technical paper, “Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation,” describing a modular architecture that links video generation, policy learning and simulation, positioning Genie 3 as a crossroads for both digital gameplay and real-world robotics. The push underscores a wider industry shift toward AI-driven content creation. Tesla and X owner Elon Musk said the future of video games lies in “dynamic generation by AI,” echoing DeepMind’s strategy of using synthetic environments to accelerate progress toward more general forms of intelligence.