中国科学院院士梅宏：当前人工智能热潮需要一场“冷思考”

Core Viewpoint - The current AI wave, particularly in large models, requires rational thinking as it faces significant limitations despite impressive advancements in deep learning [1] Group 1: AI Technology and Limitations - AI technology, represented by deep learning, is fundamentally reliant on data and computational power, achieving only perceptual intelligence rather than true cognitive ability [1][2] - The hype surrounding AI, including concepts like "replacing humans" and "general AI," overlooks critical issues such as energy consumption, data depletion, and legal-ethical challenges [1] Group 2: Model Framework and Data Dependency - Large models operate within a "probability statistics" framework, and their performance improvements do not alter the fundamental reliance on data [2] - The capabilities of AI agents are limited by the underlying large models, and their effectiveness is constrained by computational resources [2] Group 3: Future Directions in AI Research - The academic community is urged to embrace diversity in AI research, moving beyond a singular focus on deep learning to include symbolic representation, which is crucial for knowledge exchange [3] - AI should remain a controllable tool for humans, aimed at enhancing work efficiency and quality, while maintaining human oversight in knowledge discovery and value judgment [3] - Current applications of large models in text, image, and video are only a small part of industry needs, emphasizing the necessity for effective solutions to real production and business problems through data accumulation [3]