11月25日上午10：30，我院邀请国防科技大学理学院王琦副研究员作题为“Thoughts and Practice on Large-Scale Generative Simulation Artificial Intelligence”的学术报告。
报告摘要：
Recent advances have brought a new focus on generative artificial intelligence (GenAI), which paves promising paths to exploring the creation of texts, images, videos, or other contents, rather than simply performing discriminative learning tasks. GenAI's emergence, e.g., the large model, has changed the landscape of deep learning research, influenced individuals’ work and life, holding tremendous potential to reshape robotics research, national governance, and life sciences. This talk nominates large-scale generative simulation artificial intelligence (LS-GenAI) as the next hotspot for GenAI to connect. We will share some thoughts and recent practice on enabling LS-GenAI. Specifically, the identifiable simulation systems or scenarios need to be generated with a few observations, and the decision-making modules can afford fast adaptation utility in time-sensitive scenarios, e.g., autonomous driving or decision-making in defense. The primary goals of LS-GenAI are to assist in meaningful experimental design and enable fast adaptation of learned skills. Achieving these will ultimately enrich the utilities of GenAI in a broader range of real-world scenarios.
报告人介绍：
王琦，国防科技大学理学院副研究员，从事随机优化与智能博弈的研究，博士师从Max Welling教授和Herke van Hoof博士，长期担任AI顶级会议审稿人，受邀组织NeurIPS EcoRL2021 Workshop。以第一作者在ICML、NeurIPS和ICLR上发表多篇文章，涵盖条件生成模型、元强化学习、鲁棒优化和近似推断算法，获得2023年中国多智能体系统优秀博士论文奖。目前主持国家级与省部级科研课题3项，聚焦生成式仿真智能的研究框架，受邀在中国科学院数学与系统研究院、中国科学院理论物理所、清华大学、南开大学、山东大学等知名研究机构作学术报告。
