Viewpoint
World action models are computationally expensive, a barrier to deployment
World action models are computationally expensive, a barrier to deployment
Bill and Guanming from General Instinct introduce infrastructure for optimizing world action models (WAMs), which predict future frames and kinematics simultaneously, highlighting their computational expense (requiring expensive GPUs) as a key scalability barrier for real-world deployment.
- Speaker
- Bill Jiao
- Source timestamp
- 83:52
More from this interview
- Four technical walls block scaling of robotics according to Francois
- MAM system combines short-term visual and long-term textual memory for robots.
- Simulation-trained policy achieves zero-shot real-world tool manipulation
- R&B Encore uses self-supervised chain-of-thought reasoning for VLAs.
- Robotics application companies should own problems and build operational moats.
- Distillation and separate transformers speed up world action model inference