10/13/2026
Changing an object’s color can expose a weakness in a robot’s training.
Carolina Parada describes a robot trained on red objects, then placed in a scene where those same objects are blue. Without enough variety in its training data, it can associate “red” with the object it’s supposed to recognize.
Hear her explain why the environments robots learn in matter when they encounter something different.
Catch the full episode at automated.fm