AnalysisRoboticsNovember 9, 2026

UC Berkeley and NVIDIA pitch agentic robotics to replace data-hungry robot training

Read original source →spectrum.ieee.org

Announced at the 2026 Conference on Robot Learning, the approach uses AI coding tools to write human-readable robot code instead of training on massive demonstration datasets. The researchers argue data-driven policies are black boxes that are hard to make flexible, reliable, and safe in commercial deployments.

1 source

More stories today

Open the live feed