Az Xpeng bemutatta új, speciális Vision Action Model-lel hajtott humanoid robotját
Xpeng demo'd its new humanoid robot, Iron, showcasing its capabilities and the technology behind its movement. The biggest part of this launch is XPeng’s “vision action model,” which is basically a "world model" trained on 650K hours of video that lets robots learn tasks way faster than traditional methods. The demonstration included both male and female versions of the robot, and featured 'under the hood' looks to verify the robot's mechanical nature.
- The robot's 'vision action model' is trained on 650,000 hours of video data to accelerate learning.
- Xpeng provided VFX of the 'under the hood' inner workings and proof it is an actual robot, not a human in a suit.
- The launch included a male version of the humanoid touring its own factory facilities.