Analysis of competing views on integrating large models into robot control at industry events

This piece examines the debate between deterministic control systems and generalist AI models in robotics, highlighting claims from Zhiyuan Innovation regarding latency constraints and Unitree's approach to leveraging pre-trained models for physical tasks. It contrasts these viewpoints with industry data on deployment readiness and cost implications, noting that while simulation success rates are reported, real-world application faces significant challenges in force control and timing alignment.