Meta’s new world model lets robots manipulate objects in environments they’ve never encountered before
1 min read
Summary
Meta has released V-JEPA 2, which can predict outcomes and plan actions in unpredictable environments, helping to create more capable robots and advanced automation in physical environments, according to a VentureBeat report.
The two-stage training enables zero-shot robot planning, allowing a robot to be deployed in a new environment to manipulate objects it has never encountered before without needing retraining.
It can be used in virtual environments to create digital twins of processes, helping to simulate new processes or train other AI in a physically accurate virtual environment.
Meta has released the model and its training code in order to “build a broad community around this research”.
V-JEPA 2’s lower training overhead could make enterprise AI applications more viable.