Nvidia robotics head Amit Goel on physical AI safety, Halos and zero-trust
LuizaJarovsky · x · 2026-11-15
Luiza Jarovsky interviewed Amit Goel, head of robotics ecosystem at Nvidia, at GTC Berlin on physical AI safety:
- Halos for Robotics: Goel explained how the safety framework works and the new safety paradigm for physical AI, including layered safeguards for the physical world's unpredictability.
- Zero-trust philosophy: Agentic AI raises risk levels, demanding a zero-trust approach plus robust external infrastructure to monitor agent–physical system interactions.
- Optimism: Unlike most in AI governance, Goel believes physical AI will soon handle dangerous, strenuous, unsanitary or inconvenient work, freeing humans for more meaningful pursuits.
She also interviewed Ali Golshan, a senior engineer on Nvidia's new Open Agent Safety Platform. The full interview runs about 15 minutes.
Related event: Nvidia Exec Outlines New Safety Paradigm for Physical AI and Robotics(2 posts)→
More from Embodied
- Figure CEO shows humanoid robot making a bed in a real bedroom — adcock_brett · 2026-11-16
- MIT CSAIL shares robot dog meeting a robot duck — MIT_CSAIL · 2026-11-16
- Apple's smart glasses planned for next year will center on Siri AI app integrations — testingcatalog · 2026-11-16
- 12 robot learning researchers to follow after CoRL 2026 — KyleMorgenstein · 2026-11-16
- MuJoCo 4.0 cloth simulation demo shown in new test video — EL2031watson · 2026-11-16
- First fully autonomous vehicle offered to consumers as SpaceX catches Starship — mitchdeg · 2026-11-16