Cosmos 3 is designed to process and generate language, images, video, audio and actions
MLStreetTalk says Cosmos 3 takes in and generates video and outputs robot actions, letting developers test policies in simulation. Its interview with Nvidia Cosmos research lead Ming-Yu Liu is sponsored by Nvidia.
TLDR
The Cosmos 3 technical report describes a family of models designed to jointly process and generate language, images, video, audio and action sequences. Announcing an interview with Ming-Yu Liu, who leads Cosmos research at Nvidia, MLStreetTalk says Cosmos 3 lets developers test robot policies in simulation rather than in the real world. MLStreetTalk disclosed Nvidia sponsorship and shared links to the technical report and project pages.
Combined views
464
1 Source, first seen 15d ago