Dyna-2 Pre-Trained on 1M Hours of Video Reveals New Embodied Scaling Laws
JasonMa2020 · x · 2026-08-11
Dyna Robotics introduced Dyna-2, a world-action model pre-trained on an impressive 1 million hours of egocentric human video data.
This research uncovers several groundbreaking scaling laws in embodied AI:
- Human Data Scaling: World-action models exhibit scaling laws across four orders of magnitude on human data (from 1,000 to 1,000,000 hours).
- Robot Data Generalization: The scaling law derived from human data implies a scaling law for unseen robot data, demonstrating strong transfer capabilities.
- Objective & Data Matter: Both the scaling on video data and the world modeling objective are essential for cross-embodiment scaling transfer to emerge.
More from Embodied
- Kyle Vogt's Bot Company Sued for Turning Airbnb into Robot Lab — MarwaEldiwiny · 2026-08-11
- Dyna-2 Model Transfers to WUJI2 Dex Hand, Adapts in 13 Minutes of Data — JasonMa2020 · 2026-08-11
- MIT's GeoPT Model Cuts Simulation Data Needs by 60% with Built-in Physics — nordicinst · 2026-08-11
- MIT and Tsinghua introduce GeoPT: physics as AI's third modality — MIT News AI · 2026-08-11
- Amazon Gutting Nova AI Models, Pivoting Hard to Robotics — max_paperclips · 2026-08-11
- Big Tech Robotics Engineers Make $400K+, Startups Face a Math Problem — ATTlKA · 2026-08-11