While this research is still in its early stage, we believe Genie 2 has huge potential for creating novel environments for training and testing embodied AI agents.

We used Genie to create a world from the image below and gave our SIMA agent instructions to explore it. 🤖👇

Excited to introduce Genie 2: our most capable and general foundation world model. It can generate a diverse array of consistent and interactive 3D worlds from a single image, playable with keyboard and mouse inputs for up to a minute. 🧞🎮🌎