On Jan 6, 2025, Google DeepMind announced the formation of a new team to work on massive generative AI which would ‘simulate the world.’ In fact, it is the next step in the development of AI.
World models, as Google calls it, are computational frameworks that enable the AI system to understand and simulate the real or virtual world. AI systems would learn to navigate the environment. These models will be useful in gaming, autonomous systems and robotics. Autonomous cars make use of world models to simulate traffic and road conditions. AI robots which are generalist can be trained to operate in different environments.
Such scaling of AI models will evolve the technology further. Scaling related to video and multi-modal data brings us closer to AGI. World models will power different domains — visual reasoning and simulation, planning for embodied agents, real-time interactive environment.
Former OpenAI employee Tim Brooks who left in October 2024 has come on the rolls of DeepMind and would lead the world models team. It will work further on Gemini, Veo and Genie. DeepMind has already developed world models of Genie1 and Genie 2.
Leave a Reply