Google has announced a new initiative to develop AI world models capable of simulating physical environments. The team will be led by Tim Brooks, a former co-lead on OpenAI’s Sora project, who joined Google DeepMind in October.
In a post on X, Brooks shared the project’s ambitious vision. “DeepMind has ambitious plans to create massive generative models that simulate the world,” he said. “I’m hiring for a new team with this mission.”
Building on Google’s AI Foundation
The Google AI world models project will build on the work of existing AI teams like Gemini, Veo, and Genie.
- Gemini: Google’s flagship AI model for image analysis and text generation.
- Veo: Specializes in video generation.
- Genie: Focuses on creating real-time 3D environments and interactive games.
Genie’s latest version, showcased in December, demonstrated the ability to generate diverse, playable 3D worlds.
Scaling AI Training for General Intelligence
Brooks emphasized the importance of scaling AI training with multimodal data. He explained that this step is crucial for achieving artificial general intelligence (AGI)—AI that can perform any task a human can.
The new team will integrate its models with existing tools like Gemini, focusing on applications such as:
- Real-time simulation.
- Visual reasoning.
- Planning for autonomous agents.
- Interactive entertainment.
Collaboration vs. Controversy
The rise of AI world models has attracted attention from startups and large tech firms. Companies like Odyssey aim to collaborate with creative professionals, but it remains unclear whether Google will adopt a similar approach.
At the same time, copyright concerns loom. Questions persist about whether AI models are trained on unlicensed video game content or other protected materials. While Google claims permission to use YouTube videos, it has not disclosed which videos are included in the training datasets.
Google’s Vision for AI World Models
Google’s commitment to advancing AI world models highlights its focus on revolutionizing fields like 3D simulations, video games, and robotics. However, challenges such as collaboration ethics and copyright concerns could shape the trajectory of this groundbreaking project.