The robot in the open-plan office uses Google's Gemini language model to navigate and respond to commands, showcasing the potential for large language models in physical environments.
DeepMind's system, powered by Gemini, significantly improves human-robot interactions and enhances robot usability, achieving up to 90% reliability in navigation, even with complex commands.
Gemini's multimodal capabilities allow the robot to combine video and text-based information, enabling it to understand its environment and react effectively to commands.
This advancement demonstrates large language models' capability to perform practical tasks in the physical world beyond traditional web browser or app environments.
Collection
[
|
...
]