中国信通院云大所所长何宝宏：数字原生，点亮未来智能化社会

Core Insights - The article emphasizes the transition from passive digital transformation to proactive digital-native approaches, aiming to fully unleash the potential of AI technologies [2][16] - It discusses the challenges and breakthroughs in the application of AI large models and AI agents in industrial digital transformation [2][3] Group 1: Digital Transformation and AI - The forum highlighted the need for industries to escape homogenized competition and low-level replication, focusing on how AI can create real business and social value [2][3] - Dr. He Baohong proposed a "digital-native" framework to guide industries in overcoming traditional transformation paradigms and unlocking the value of AI technologies [2][16] Group 2: Historical Context and Evolution - The article draws parallels between the current AI era and the past internet era, noting that just as TCP/IP established the foundation for the internet, the Transformer architecture is now central to AI infrastructure [3][4] - The evolution from the internet's initial text interactions to multimedia and specialized networks is mirrored in AI's progression from general models to domain-specific applications [4][12] Group 3: AI Model Implementation - The implementation of large models involves three core stages: model preparation, application, and operational maintenance, with a focus on continuous iteration to adapt to business changes [6][8] - Data processing has evolved from simple purification to a systematic data strategy, emphasizing the need for specialized data annotation in high-value industries [8][9] Group 4: Agent Development and Market Potential - Agents are identified as the key to transitioning from a scale-first approach to an efficiency-first model in AI, enhancing reasoning capabilities and multi-modal integration [12][13] - The agent market is projected to grow significantly, reaching $216.8 billion by 2035, with a compound annual growth rate of 40.15%, indicating its role as a major driver of digital economic growth [13][18] Group 5: Challenges and Future Directions - The article outlines challenges faced by agents, including the complexity of goal transmission and the inherent limitations of AI models, which may lead to misunderstandings in user interactions [14][18] - The concept of "digital-native" is proposed as a guiding principle for future intelligent societies, advocating for a shift from traditional digital transformation to a more innovative and integrated approach [16][18]