JD INDUSTRIALS (07618) and AMD Showcase Innovative World Model Application at CIIE, Making AI Your Most Understanding "Production Assistant"

The 9th China International Import Expo officially opened in Shanghai a few days ago. As a global leader in high-performance and AI computing, AMD (AMD.US) appeared at the CIIE for the sixth consecutive year under the theme of "Empowering the Agentic AI Era." At the JD INDUSTRIALS (07618) booth within AMD's exhibition area, the two parties showcased the application of the JD INDUSTRIALS production line all-in-one machine. Based on the latest industrial world model plus AMD's edge computing unit, the feasibility of human-machine collaboration on industrial production lines was preliminarily validated. This also marks the latest exploration progress following the strategic cooperation between the two parties, focusing on the deep integration of "industrial large models + intelligent computing hardware," preliminarily validating the application potential of the relevant technologies in industrial production line scenarios and laying the foundation for further exploration and application of AI edge computing in large-scale industrial terminal scenarios. In the demonstration, the model was able to determine in advance the tools likely to be needed next and the timing for collaboration before the operator issued explicit instructions or reached for a tool, using a robotic arm to grasp the most suitable tool and hand it to the operator at the most appropriate moment, helping the operator complete the work more smoothly. This demonstrated the industrial world model's multi-task understanding and forward-looking reasoning capabilities in the human-machine collaboration process. Industrial World Model JoyIndustrial-WM Builds a "ObservationState AssessmentInteractive Decision-Making" Closed Loop According to disclosures, this demonstration is based on JoyIndustrial-WM, an industrial world model newly developed by JD INDUSTRIALS. This is a real-time analysis and human-machine collaboration industrial world model oriented toward video streams. Through deep adaptation with the AMD Ryzen AI Max+ 395 processor, it can continuously understand personnel actions, tool states, and interaction processes. Through continuous visual observation, the model identifies three interaction stages"waiting, preparation, and response"and builds an "observationstate assessmentinteractive decision-making" closed loop: from determining whether personnel are in a waiting state, to identifying workpiece preparation and tool handover needs, to clarifying needs, identifying target tools, and generating interactive responses or action instructions. First, the system can predict needs in advance and prepare ahead of time. Combining the person's current actions, operation sequence, and interaction process, the model reasons about the work state of the next stage, predicts the type of tool the person is about to use and the upcoming tool handover need, and prepares the next tool in advance. Second, the system can grasp the timing of delivery so that tools arrive when needed. Efficient human-machine collaboration requires judging not only "what to hand over" but also "when to hand it over." JoyIndustrial-WM is oriented toward human-machine interaction and collaboration scenarios in industrial production. Through continuous video observation, it understands the operator's actions, tool usage, and task progress, predicts future interaction states and tool needs, and achieves continuous collaboration from demand prediction and advance preparation to timely delivery, helping operators complete work more smoothly. Third, it reduces operational interruptions and makes workflows smoother. In tool delivery and workstation collaboration, the model integrates tool preparation into the person's existing operation flow by predicting subsequent needs, reducing situations where personnel repeatedly shift attention, reach for tools, and interrupt operations. At the same time, by judging tool usage and handover timing, it helps reduce the risk of mistakenly touching tools, knocking over objects, and dropping tools during delivery. Building on the strategic cooperation between JD INDUSTRIALS and AMD, this human-machine collaborative work model demonstrates leading tool prediction capability and prediction speed. The edge-side combination of the JoyIndustrial-WM industrial world model and the AMD Ryzen AI Max+ 395 processor is at an industry-leading level in terms of tool type prediction and prediction speed. Deep Integration of "Industrial Large Models + Intelligent Computing Hardware" Creates a Model Practice for Smart Manufacturing Applications JD INDUSTRIALS and AMD previously signed a strategic cooperation agreement to carry out multi-level, wide-ranging, and all-around deep strategic cooperation in areas such as industrial large models, intelligent computing power, global promotion of industrial scenario applications, and market channels. In particular, the two parties will focus on the deep integration of "industrial large models + intelligent computing hardware," targeting factory, workshop, and other industrial terminal scenarios and edge computing environments to jointly develop AI terminal products and applications for industrial sites. The AI edge-side innovative applications jointly promoted by the two parties will bring enormous upgrades to business operations in real industrial scenarios. Wang Peinuan, CEO of JD INDUSTRIALS, gave a vivid example at the signing: on the production floor, personnel may enter the trajectory of moving vehicles, creating a collision risk. The industrial world model plus edge computing power can reason about the future states of personnel and equipment, take the actual on-site environment into account, and determine collision risks several seconds in advance, triggering alarms or slowing vehicles down ahead of time. The latest demonstration showcased a collaborative work model implemented on the edge side in real industrial production scenarios through deep adaptation between the industrial world model and edge computing power. A relevant technical lead at JD INDUSTRIALS said: "The industrial world model transforms video understanding into collaboration and automation capabilities, enabling tool handover services, workstation collaboration, and Siasun Robot&Automation responses, expanding application space for smart manufacturing enterprises to build deployable and scalable intelligent solutions." From vehicle collision risk prediction on the production floor to human-machine collaboration in tool usage, the "industrial large models + intelligent computing hardware" model created by JD INDUSTRIALS and AMD further advances industrial AI applications from centralized analysis toward edge-side real-time decision-making, enabling risk prediction, change reasoning, task understanding, and assisted decision-making in complex sites to be completed locally, reducing response latency, enhancing data security, and improving deployment flexibility. The two parties will strive to create an industry benchmark practice for the deep integration of industrial world models and high-performance edge computing power, providing a replicable and scalable practice model for AI applications in the industrial sector.