Nvidia experts taught robots to repeat Cristiano Ronaldo’s movements

The technique is called real alignment simulation and physics (ASAP). The scientists used videos with people’s movements and created a computer simulation in which artificial intelligence learned to repeat after a person.

As a result, anthropomorphic robots of the G1 of the Chinese company Unitree have been able to reproduce the movements of famous athletes, for example, the Ronaldo brand jump and the launch of Bryant. ASAP made it possible to reduce the number of errors in follow -up movements by 52.7% compared to the previous methods, because steel robots can perform complex and previously inaccessible maneuvers, DECRYPT notes.

Robots movements are still quite uncomfortable, but now this is due to the imperfection of devices themselves, not software. The fact is that anthropomorphic robots remain less articulation points than a person writes the newspaper.

On December 30, Nvidia said that in the first half of 2025, the latest generation Jetson Thor for humanoid robots will launch an integral solution with a series of components, from individual chips to a complete controlled training and software software. According to the Vice President of Nvidia for Department of the Robotics, the revolution in the field of artificial intelligence and the arrival of the “physical AI” are arriving. In this “turning point” for the Nvidia market, it wants to be among the leaders, he added size.