Runway, best known for AI video generation, has introduced Praxis-1, a world action model for robots. It is built on the same large scale video pretraining behind Runway's world models and turns that understanding into robot control.
CTO Kamil Sindi says most robot policies are held back by scarce robot data, while Praxis-1 learns mostly from third person video. Runway plans to release it with open weights after more testing across robot types.
Why it matters: if video can stand in for expensive robot demonstrations, the companies with the most video may have an edge in robot brains.
As an Amazon Associate, RobotNewsToday.com earns from qualifying purchases.