Genie: Generative Interactive Environments
This addresses the problem of creating scalable and generalizable virtual environments for AI agents, offering a novel approach to world modeling without domain-specific requirements.
The authors introduced Genie, a generative interactive environment model trained unsupervised from unlabelled Internet videos, which can generate diverse action-controllable virtual worlds from prompts like text or sketches, with 11B parameters enabling frame-by-frame interaction without ground-truth action labels.
We introduce Genie, the first generative interactive environment trained in an unsupervised manner from unlabelled Internet videos. The model can be prompted to generate an endless variety of action-controllable virtual worlds described through text, synthetic images, photographs, and even sketches. At 11B parameters, Genie can be considered a foundation world model. It is comprised of a spatiotemporal video tokenizer, an autoregressive dynamics model, and a simple and scalable latent action model. Genie enables users to act in the generated environments on a frame-by-frame basis despite training without any ground-truth action labels or other domain-specific requirements typically found in the world model literature. Further the resulting learned latent action space facilitates training agents to imitate behaviors from unseen videos, opening the path for training generalist agents of the future.