close
close

China has released an open source-met fu robot and we are not sure why

Video about the Chinese official news agency Xinhua.

We have a new entry in the constantly growing “robot and algorithms that do things they asked for it”. Introduction of the G1 robot, a humanoid machine with impressive balance, smooth mobility and unexpected -Kung Fu skills. Yes, it can beat. Yes, it can pull through a roundhouse kick. And yes, that raises the question: Why?

Every robot fought kung fu

G1 performed for the first time on the global stage of the China's Spring Festival Gala, one of the most viewed television events in the world. However, the debut was somewhat uncomfortable and showed a humanoid figure that seemed more of a novelty than a serious technological breakthrough.

That changed quickly. In January, UNTREE, the company behind G1, demonstrated the robot effortlessly over uneven terrain. Now it shows complex martial art movements and proves how much humanoid robots are progressing in terms of coordination, stability and accuracy.

The G1 shows remarkable coordination and flexibility with 23 degrees of freedom (DOF). Photo credits: Unitree.

The robot costs 99,000 yuan ($ 13,622) and is equipped with 43 joints and mimicry learning. It is able to learn complex movements and shows an exact maneuverability.

The G1 has a price of 99,000 yuan ($ 13,622) and is equipped with 43 joints and advanced mimicry learning functions. This enables him to carry out highly dynamic actions with remarkable precision. But what is even more fascinating is How It learns.

Learn before passing

Before G1 ever takes a physical step, his training begins in a virtual world. UniTree uses the Isaac AI-controlled robotics platform from Nvidia, on which the robot first learns the basics of movement.

From there, G1 is shown a digital twin – another humanoic robot that performs actions. Using motion recording and video data, the skills are observed and gradually refined and reflects the process to use people when learning through imitation. Over time, the movements of the robot become more complex, fluid and more customizable.

An advertising picture that shows some characteristics of the G1 robot. Photo credits: Unitree.

This mix of simulation -based training and real refinement is an important step in robotics, which makes humanoid machines much more adaptable.

In addition to his acrobatics, one of the most remarkable aspects of the G1 is that it is open source.

Open source robot

Similar to Deepseek, UNTREE has made its training data, algorithms and hardware specifications publicly for its affordability, but also because of his open source accessibility available on Github. This means that developers, researchers and robotics enthusiasts can explore, optimize and improve G1's skills in a way, even if his creators may not expect.

The data record contains an interactive network and inverse kinematics gorithms, with which G1 can monitor joint positions, restrictions and motion efficiency. In short, it can learn and optimize movements autonomously, regardless of whether it does a kung -Fu routine or something more practical.

Despite the striking martial arts demonstration, Unitree insists that G1 was built for practical applications – not for the fight. The company introduces itself to helping its humanoid robot, which helps workplaces, houses and even hospitals and does repeated and complex tasks.

However, since the robot is open source, its potential goes far beyond the view of Unitree. Researchers, engineers and AI developers now have a powerful humanoid platform that they can adapt for various real applications-from rehabilitation and older care up to industrial automation and education.

Despite its impressive skills, G1 raises a bigger question – are we ready to accept robots like this in our daily life?

At the moment, G1 is an impressive representation of engineering, a symbol of how far the robotic has got. But whether we accept robots as employees, assistants or even companions remains an open question-a one that society has to answer in the coming years.