报告题目 (Title)：智能科学-数学原理与工程应用 （Intelligence Science--Mathematical Principle and Engineering Applications）
报告人 (Speaker)： 吴焦苏 研究员（中科院人工智能联盟、华东政法大学互联网法治研究院）
报告时间 (Time)：2023年11 月22日 ( 周三 ) 15:00-17：00
报告地点 (Place)：校本部GJ303
邀请人(Inviter)：白延琴 教授
报告摘要：We develop a general theoretical framework of Intelligence Science——Spatial-Temporal Gamedynamics，including mathematical principle, i.e., Strategic Correlativity Principle (SCP), Strategic Causality Law (SCL), Fundamental Theorem of Strategic Causal Learning (FTSCL)，and show some interesting properties, such as scaling law, emergent ability, phase transition.
We apply Intelligence Science to Intelligence Engineering for large communication and action games with high order beliefs, such as Mechanism Design, Information Design, Action Design to reduce hallucination and execute superalignment for multi-modal pretraining large models, including Large Language Model (LLM)/AI Generated Content (AIGC), i.e., OpenAI’s ChatGPT/GPT-4, and Vision-Language-Action Model (VLA)/AI Generated Action (AIGA), i.e., Google’s PaLM-E and Stanford’s VoxPoser, etc.