Google Genie 2
A large-scale foundation world model
Listed in categories:
VideoArtificial Intelligence
Description
Genie 2 is a foundation world model developed by Google DeepMind that generates an endless variety of action-controllable playable 3D environments for training and evaluating embodied agents. It allows users to create diverse interactive experiences based on a single prompt image, enabling both human and AI agents to interact with the generated worlds using keyboard and mouse inputs.
How to use Google Genie 2?
Users can prompt Genie 2 with a single image to generate a 3D environment. They can then interact with this environment using keyboard and mouse inputs, allowing for both human and AI agents to explore and perform actions within the generated world.
Core features of Google Genie 2:
1️⃣
Generates diverse 3D environments from a single image prompt
2️⃣
Supports action-controllable interactions for both human and AI agents
3️⃣
Models complex object interactions and character animations
4️⃣
Remembers and accurately renders parts of the world that are no longer in view
5️⃣
Enables rapid prototyping of interactive experiences for researchers and designers
Why could be used Google Genie 2?
| # | Use case | Status | |
|---|---|---|---|
| # 1 | Training AI agents in varied and rich 3D environments | ✅ | |
| # 2 | Rapid prototyping of interactive experiences for game developers and researchers | ✅ | |
| # 3 | Evaluating AI agents' performance in unseen environments generated by Genie 2 | ✅ | |
Who developed Google Genie 2?
Genie 2 was developed by Google DeepMind, a leading AI research organization known for its innovative work in artificial intelligence, including breakthroughs in game-playing AI and generalist agents.