丁宁：大模型是“智能基建” 资本与技术融合重塑AI版图

Core Viewpoint - The event "Scientists Meet Investors" highlighted the significance of the fourth industrial revolution, emphasizing that artificial intelligence (AI) is likely to become an indispensable core technology in the future world [1]. Group 1: AI Development Trends - AI is entering a "multimodal fusion" stage, evolving from single-modal models to integrating text, images, speech, and code [2]. - The performance of large language models (LLMs) is not solely dependent on parameter size; structural design, training methods, and data quality also play crucial roles [2][3]. - The industry is shifting from blind expansion of model size to structural innovation and refined training due to factors like cost, energy consumption, and data cleaning [2]. Group 2: Capital and Technology Interaction - The rapid development of large models relies on both capital investment and industry collaboration, where capital acts as a magnifier for technology, and technology enhances capital effectiveness [4]. - High initial costs for training large models include computing power, data, algorithms, and talent, making capital intervention essential for developing high-quality foundational models [4]. - China leads globally in AI-related patents, accounting for 69% of the total as of 2023, while the U.S. maintains a lead in top enterprises and computing power centers [4]. Group 3: Future Trends in AI - Future AI development will feature trends such as multimodal integration, parallel advancements in large-scale and lightweight models, and embodied intelligence that interacts with the physical world [5]. - The exploration of artificial general intelligence (AGI) aims for systems with general cognitive and self-learning capabilities, while superintelligence remains a theoretical concept [5].