The Chinese humanoid robot industry is gradually transitioning from technological innovation to large-scale commercial application, with a key driver of this progress being the refinement of the robots brains, meaning the increasing complexity of their algorithms, thanks to the Vision-Language-Action model robots now have a unified end-to-end architecture that enables precise motion control from visual perception to the operation of complex joints, as a result humanoid robots are capable of performing long and complex sequences of tasks such as making a bed, in Beijing the startup Spirit AI trains its robots through imitation learning, for example when preparing ingredients for hot pot the machines learn to carefully place food into bowls or containers by mimicking human actions, replicating the movements of employees using wearable devices, in Shenzhen robots assist with household chores by helping organize small items and collect trash in living rooms and dining areas, according to Hu Bowen an engineer at X Square Robot these robots can identify used items such as bottles, cans, and tissues and throw them into trash bins, while performing household tasks the machines continuously train which optimizes their models and strengthens their algorithms, over time robots are expected to perform tasks more efficiently, robots used in supermarket product sorting must memorize product layouts, input product information and constantly refine their grasping skills, Pan Guoping head of algorithms at Zerith Robotics said the company has a pre-trained model and for some products about thirty minutes is enough to collect around 20 data samples, experts hope the model will learn to generalize through analogy, for example if it learns one type of potato chips it could apply that knowledge to all similar products, this level of generalization has been achieved through data-driven approaches and large-scale models, Yang Wei a brand manager at Zerith Robotics added that developing full artificial general intelligence may take five to ten years during which robots could gain human-like multidimensional capabilities, however the industry is already making breakthroughs in commercial applications and building closed-loop systems for such tasks, and experts predict that within the next one to two years new innovative use cases for humanoid robots will emerge.