陈天桥罕见公开演讲：投入超10亿美金发展AGI“发现式智能”

Core Viewpoint - The article presents the concept of "Discoverative Intelligence" as a new approach to achieving Artificial General Intelligence (AGI), emphasizing its role in human evolution and scientific discovery [5][7][10]. Group 1: Concept of Discoverative Intelligence - Discoverative Intelligence is defined as the ability to ask questions, understand principles, and construct testable theoretical models, distinguishing it from existing AI models that primarily focus on extrapolation within known parameters [8][9]. - The article argues that AI's ultimate value lies in its capacity to assist humans in discovering the unknown, rather than merely replacing human jobs [7][10]. Group 2: Paths to Discoverative Intelligence - Two main paths to achieving Discoverative Intelligence are identified: the "Scale Path," which relies on large models and data, and the "Structure Path," which focuses on cognitive mechanisms akin to human brain functions [9][10]. - The Scale Path has yielded significant results in AI applications, while the Structure Path is emerging as a necessary complement to overcome the limitations of current AI systems [9][14]. Group 3: Time Structure Analysis - The article introduces the concept of "Temporal Structure," which refers to the brain's ability to process information dynamically over time, contrasting with the static nature of current AI models [10][11]. - Five core capabilities are outlined as essential for achieving Temporal Structure: neural dynamics, long-term memory, causal reasoning, world modeling, and metacognition [11][12][13]. Group 4: Opportunities for Young Researchers - The article emphasizes that the current limitations of the Scale Path present an opportunity for young researchers to redefine intelligence through structural approaches [14][16]. - The company plans to invest over one billion dollars in dedicated computing resources and establish research centers to support young scientists in exploring new theories and algorithms [19].