Roboter ermöglichen, Tools zu planen, zu denken und zu verwenden, um komplexe Aufgaben mit Gemini Robotics 1.5 zu lösen
Roboter ermöglichen, Tools zu planen, zu denken und zu verwenden, um komplexe Aufgaben mit Gemini Robotics 1.5 zu lösen
2 Kommentare
The video shows the evolution of DeepMind’s robotics to multi-step planning with Gemini. One of the trends people look at for more general MLLM robotics is how well a robot can handle new scenes and complex tasks. Being able to adapt on the fly to changing setups means less setup time in factory environments when processes change. The ability for robots to breakdown tasks given by a human means they can function in human spaces as assistants performing a wider range of tasks. (Laundry robots are an example people often give where clothes are incredibly varied).
Shouldn’t be as surprising as it is to a lot of people, myself included. An AI system that can reliably generate pictures of spaces and things should be able to navigate through them unless it has a really terrible body plan. Once AIs were able to figure out things like perspective, it’s only a matter of time before they can navigate a room or a forest using perspective data from their optics.