Expanding Technical Application Scenarios to Promote Intelligent Upgrades in the Automotive Industry

The automotive industry is witnessing rapid advancements in embodied intelligence technology. A key question facing the industry is how to embrace cutting-edge technologies for further development and upgrades. Recently, a series of interviews titled 'China Automotive Innovation Technology Tour: Insights into New Productive Forces in the Automotive Industry' was held, focusing on the topic 'From Intelligent Vehicles to Embodied Intelligent Entities: Mutual Empowerment and Ecological Restructuring.' This event featured in-depth discussions among industry professionals on how automotive technology can drive the application of embodied intelligence across various scenarios, and how embodied intelligence can continuously break through in the direction of automotive intelligence. Embodied intelligence is generally understood to be a form of AI that, after simulating language intelligence through multimodal large models, further increases its action space to interact with both virtual and physical worlds in a human-like manner. The director of the Visual Intelligence Research Center at Tsinghua University, Deng Zhido, emphasized that embodied intelligence focuses on simulating human interactions in the real physical world, rather than just understanding and reasoning in the internet or pure digital spaces. The characteristics of embodied intelligence include 'human-like actions, human-like interactions, and wide applicability,' enabling it to achieve humanized perception, prediction, decision-making, and control. Industry insiders believe that the application of embodied intelligence in the automotive sector holds significant potential. Cars can serve as excellent carriers for embodied intelligence due to their numerous cameras and sensors, which can directly gather information from inside and outside the vehicle to serve as input for various algorithms. The automotive industry, often referred to as the 'jewel in the crown of industry,' constantly absorbs new technologies to serve humanity. The development of embodied intelligence relies heavily on high-quality training data; hence, those who possess more data for training and who can solve unsolved or poorly addressed problems will take the lead. This is where the automotive industry can facilitate the rapid implementation of embodied intelligence. Autonomous driving emerges as a crucial application scenario for embodied intelligence. Zhao Lijin, deputy secretary-general of the China Society of Automotive Engineers, pointed out that the perception, decision-making, and control involved in autonomous driving closely align with the technological evolution of embodied intelligence. Furthermore, cockpit interaction and power system optimization are also seen as important potential application scenarios for embodied intelligence. For electric or hybrid vehicles, energy consumption is a critical metric, and embodied intelligence can learn from factors such as road conditions and the environment to form optimal driving algorithms, thereby optimizing energy consumption. As the automotive industry is poised to accelerate the realization of embodied intelligence, the introduction of this technology can inject new vitality into the high-quality development of the automotive sector. Zhao noted that embodied intelligence, as a common technology, is expected to further enhance the intelligence of automotive products. It can facilitate improvements in vehicle performance, enabling higher levels of autonomous driving and optimizing interactive experiences in smart cockpits. On the manufacturing side, humanoid robots can undertake repetitive and hazardous tasks, improving production efficiency and ensuring product quality. Looking ahead, the automotive industry and embodied intelligence are set to empower and develop collaboratively. Both sectors share high similarities in their supply chain processes involving sensing, decision-making, execution, and new energy. However, challenges remain in evolving vehicles into embodied intelligent entities capable of autonomous decision-making. The primary challenge lies in transitioning from a one-dimensional text-based language model to a four-dimensional world model, requiring extensive training data that encompasses interactions in both real and virtual environments. In summary, the development of embodied intelligence as a carrier of AI's influence on reality necessitates comprehensive integration from the chip level to application layers. This ensures that vehicles equipped with embodied intelligence can respond quickly and effectively to demands during operation. However, this poses high requirements on the R&D capabilities of companies, with only a handful currently able to meet these standards. Strengthened collaboration between software development and application layers is also essential for creating targeted and effective software solutions tailored to the intricacies of the automotive industry.