We developed an AI model based on predictive processing (the free-energy principle), a computational framework for flexible human cognition and behavior. By directly integrating high-dimensional visual and proprioceptive information, the model enabled a robot to perform multiple care-related tasks, demonstrating the scalability of brain-inspired computation to complex real-world control.