Runway's Praxis-1 announcement proposes using video pretrainingThe first, biggest phase of building a model: training it on enormous amounts of text so it learns language, facts, and reasoning in general.Full definition → as a foundation for controlling robots. The motivation is data availability: collecting demonstrations on robot hardware is expensive, while ordinary video provides many examples of objects moving and people interacting with them. Runway says Praxis-1 builds on the pretraining behind its video models and is being tested on partners' hardware. The intended benefit is a stronger starting point for learning physical tasks. At the announcement, public release of the model's weights remained planned for the coming months.
Terms in this piece · Glossary
pretraining — The first, biggest phase of building a model: training it on enormous amounts of text so it learns language, facts, and reasoning in general.
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
