“物理AI”如何落地?自动驾驶是关键应用场景之一

Core Insights - The 2026 CES in Las Vegas highlighted the next phase of AI development, focusing on breakthroughs in computing power, the transition of "physical AI" into real-world applications, and personalized services driven by intelligent agents [1] Group 1: Computing Power Breakthroughs - The CEO of AMD, Lisa Su, stated that the number of active AI users globally has surpassed 1 billion and is expected to exceed 5 billion in the future. Current computing power is insufficient to support the vision of ubiquitous AI, necessitating a 100-fold increase in global computing power over the next few years [2] - The industry is evolving from traditional methods of enhancing chip performance to a full-stack collaborative design approach, integrating chips, networks, and storage into a unified AI computing platform [2] - NVIDIA's CEO, Jensen Huang, introduced the "Vera Rubin" platform, a system-level computing platform comprising six chips designed to significantly reduce model training time and inference costs, emphasizing the creation of an entire stack from chips to infrastructure [2] Group 2: Physical AI Implementation - Huang emphasized the evolution of AI towards "physical AI," which will understand the physical world, predicting a "ChatGPT moment" for this technology [3] - AI is transitioning from passive systems that understand text and images to interactive systems that assist in real-world interactions, with applications in autonomous driving, robotics, and industrial automation [3] - Siemens' CEO highlighted that industries are shifting towards AI-driven products and systems, indicating that industrial AI is becoming a transformative force rather than just a functional tool [3] Group 3: Development of Intelligent Agents - The co-founder of Liquid AI, Ramin Hasani, declared this year as the year of "proactive agents," moving beyond reactive AI assistants to systems that can autonomously understand complex goals and execute multi-step plans [4] - Lenovo introduced its first personal super intelligent agent, Lenovo Qira, which operates across platforms and devices, enhancing user experience through context awareness and preference prediction [4] Group 4: Hybrid AI and Human-Centric Development - Lenovo's CEO, Yang Yuanqing, emphasized that hybrid AI, which integrates personal, enterprise, and public intelligence, is crucial for promoting personalized and diverse AI development [5] - The CEO of Havas Group stressed that AI should serve as a collaborative partner in human creativity rather than a replacement, reinforcing the importance of keeping humans at the center of technological advancements [5]