Google DeepMind unveils RT-2 vision-language-action model

RT-2 is a first-of-its-kind vision-language-action (VLA) model that learns from both web and robot data and translates that knowledge into generalised robot control instructions.