Dyna-2 Robots Predict the Future: Million-Hour World Models and the Bitter Lesson of Robotics
JasonMa2020 · x · 2026-10-14
Latent Space podcast talks with Dyna Robotics' Jason Ma about world-action model Dyna-2, which predicts the future to guide physical robot actions.
- Training on over 1 million hours of human video lets robots adapt to new tasks with as little as 45 minutes of demonstrations
- Synthetic data and self-improving world models could accelerate robotics training
- Agentic harnesses enable hour-long autonomous workflows
- Scaling's Bitter Lesson, full-stack hardware, and real deployments at Din Tai Fung are turning physical AI from demos into useful robots
Related event: Dyna-2 Learns to Predict the Future from Millions of Hours of Human Video(2 posts)→
More from Embodied
- Raven Prism AI glasses launch set for October 28 in San Francisco — smtabatabaie · 2026-10-14
- DeLorean joins Airspeeder International Series, first American team debuting in November — evilrabbit_ · 2026-10-14
- Gemini 4 Argon taught Apptronik's Apollo humanoid cornhole from video, sinking 7 straight throws — GoogleDeepMind · 2026-10-14
- Box2D dev unveils contact recycling, making contact generation almost free in physics sims — EricLengyel · 2026-10-14
- Startup Squad Turns Idea Into Robot-Drawn Brooklyn Bridge Sketch in 30 Minutes — davidfromkansas · 2026-10-14
- Stream Ring, an AI wearable three years in the making, starts shipping next month — pzakin · 2026-10-14