Runway's Praxis-1 robot brain learns from ordinary videos, cutting demo data from thousands of hours to minutes
c_valenzuelab · x · 2026-10-03
Spotted at Runway's conference: the video-gen lab is applying its world model research to robotics' biggest bottleneck — training data. Praxis-1, a robot "brain," learns how the physical world works from ordinary videos of humans doing tasks, so robotics companies need only minutes of demos on their own robot instead of thousands of hours of teleoperated recordings. CEO Cristóbal Valenzuela adds that "the most consequential updates in the next 18 months are coming."
More from Embodied
- Robotics' ChatGPT moment isn't needed: specialized bots land $600M deals now — Rewkang · 2026-10-03
- Tsinghua and Infinigence Open-Source RLark, a Cloud-Native Orchestration Platform for Robot Fleets — jiqizhixin · 2026-10-03
- Neuralink pretrained AI models on 50,000 hours of unlabeled brain data, betting on the scaling playbook — Dr_Singularity · 2026-10-03
- Nat Friedman launches Muse Gadgets: open source ESP32 firmware and SDK to build hardware for Muse — andrew_n_carr · 2026-10-03
- Nat Friedman gives away 5,000 open-source Muse Home Link dongles to connect AI to smart homes — altryne · 2026-10-03
- YC-backed Applied Electrodynamics builds cameras that see through walls with radio waves — ycombinator · 2026-10-03