Tag
#physical AI
Robots that learn a task from one demonstration: inside 2026's vision-language-action models
In one week of November 2026, Google DeepMind and NVIDIA released robot foundation models that learn from human video and a handful of demonstrations. How vision-language-action models work, what a humanoid actually did on a Mercedes-Benz line, and what is still hard.
AI that imagines whole worlds: how world models like Genie 4 and Cosmos 3 are training robots and self-driving cars
World models generate interactive, physics-aware video that responds to actions. In late 2026 Google DeepMind previewed Genie 4 and NVIDIA released open Cosmos 3 checkpoints for driving and robot simulation. What they are, how they are used, and where they still drift.