2024 · Google DeepMind · Advanced
Genie 2: A Large-Scale Foundation World Model
Genie 2 generates action-controllable 3D environments from a single prompt image.
Direct answer
What does Genie 2: A Large-Scale Foundation World Model contribute?
Genie 2 generates action-controllable 3D environments from a single prompt image.
Background
The system produces diverse interactive worlds with consistent motion, object interactions, animation, and viewpoints over short horizons. DeepMind presents it as a platform for training and evaluating embodied agents.
Problem
The work addresses a central constraint in World Models: building systems that learn useful representations or actions while remaining general enough to transfer beyond a single demonstration or environment.
Core idea
Genie 2 generates action-controllable 3D environments from a single prompt image.
Architecture and method
The system produces diverse interactive worlds with consistent motion, object interactions, animation, and viewpoints over short horizons. DeepMind presents it as a platform for training and evaluating embodied agents.
- Prompted 3D world generation
- Action-controllable simulation
- Agent evaluation environments
Results and impact
It shows how generative world models may become scalable environment engines for agent research.
Prerequisites
- Genie
- Video generation
- Embodied agents
Recommended reading order
Read the explanation above, review the related topic pages, then use the primary-source links below to inspect the abstract, figures, experiments, and released implementation.
Primary sources
External links are provided after the context needed to evaluate the work.
Follow-up research
Related papers and concepts
World Models
A compact latent model can let an agent learn behavior inside its own predicted environment.
Dreamer: Reinforcement Learning with Latent Imagination
Dreamer learns long-horizon behavior by propagating value gradients through imagined latent trajectories.
DreamerV3: Mastering Diverse Domains through World Models
DreamerV3 uses robust normalization and objectives to learn across more than 150 tasks with one configuration.
Genie: Generative Interactive Environments
Genie learns controllable interactive environments from unlabeled internet video.
Common questions
Frequently asked questions
What is the main idea of Genie 2: A Large-Scale Foundation World Model?
Genie 2 generates action-controllable 3D environments from a single prompt image.
Why is Genie 2: A Large-Scale Foundation World Model important?
It shows how generative world models may become scalable environment engines for agent research.
What should I learn before reading Genie 2: A Large-Scale Foundation World Model?
Recommended prerequisites are Genie, Video generation, Embodied agents.