Robotics Action Spaces
Exploring the implications of large action spaces in robotics, Ashley discusses how increasing the number of latent actions can enhance model fidelity. However, mapping these actions to real-world applications presents challenges, particularly when considering the continuous nature of robotics. The conversation also touches on the potential of foundation models in robotics, emphasizing the need for diverse data sources and the inclusion of sensor data to capture critical interactions like haptics.In this clip
From this podcast
Machine Learning Street Talk (MLST)
Ashley Edwards - Genie Paper (DeepMind/Runway)
Related Questions