Does Generative AI Really Bring Us a New Generation of Robots?

Within the past years, the already very strongly evolving field of artificial intelligence (AI) and machine learning has experienced a paradigm shift, where artificial models are trained for the ability to generate new data samples. These so-called generative models have since been extended, refined, and applied to various fields from text document generation to image and video synthesis. In robotics, GenAI models are particularly interesting for reinforcement learning settings and, more recently, also in the context of so-called vision-language-action (VLA) models. This talk will give a brief overview of some recent GenAI methods, highlighting their potentials for robotics in general and showcasing some of the challenges they come with. In particular, examples from robot perception will be shown, where generative AI models help to detect objects for robotic manipulation.