Disturbing Video Shows Man Jerking Robot Around by Chain Around Its Neck
Briefly

Disturbing Video Shows Man Jerking Robot Around by Chain Around Its Neck
"A video showing a robotics engineer violently yanking a humanoid robot around by a chain wrapped around its neck is making its rounds on social media, prompting fears that he may be accelerating the inevitable robot uprising. Footage shows the robot stumbling helplessly as it struggles to stay upright. A different clip shows the human engineer kicking the robot's torso in an apparent attempt to knock it down to the ground."
"The latest video was originally shared by Tsinghua University PhD student Zhikai Zhang, who has been developing a humanoid motion tracker system called Any2Track in collaboration with Chinese robotics company Galbot, the same firm that recently opened a robot-run bodega. The system "needs to operate stably in real-world scenarios against various dynamics disturbances, including terrains, external forces, and physical property changes for general practical use," Zhang and his colleagues wrote in the project's documentation."
"The team deployed a "two-stage reinforcement learning framework" to teach the robot how to adapt to these disturbances - such as being kicked or jerked around with a chain - on the fly. The results speak for themselves. The Unitree G1 bipedal robot refuses to lose its footing, despite its human master's considerable abuse, maintaining an impressive degree of balance and composure."
A video shows a robotics engineer yanking a humanoid robot by a chain around its neck and kicking its torso while the robot stumbles but resists falling. Viewers reacted with jokes about a robot uprising, yet the demonstration highlights notable advances in balance and resilience. Tsinghua University PhD student Zhikai Zhang shared the clip and has been developing the Any2Track humanoid motion tracker with Chinese robotics company Galbot. Any2Track must handle real-world dynamic disturbances across terrains, external forces, and changing physical properties. A two-stage reinforcement learning framework enables on-the-fly adaptation, allowing the Unitree G1 to maintain footing despite deliberate abuse. Similar footage shows another Unitree G1 shrugging off kicks.
Read at Futurism
Unable to calculate read time
[
|
]