【时  间】2026年1月22日（星期四）10:30

【地  点】量子院526报告厅

【主  持】王正安助理研究员  智能量子计算与模拟团队

【题  目】Some examples on the interplay between AI and science

【摘  要】Artificial intelligence (AI) has been the most significant variable in contemporary technology. While it appears as a computational approach for modeling mappings, it has an intimate connection with scientific concepts and tools. This talk starts with a brief review on scientific concepts that inspired key origins of AI breakthroughs, as awarded by the Nobel Prize in Physics in 2024, and AI-for-science examples corresponding to the chemistry prize, followed by our attempts to overcome the classical accuracy-efficiency trade-off in quantum chemistry. The talk goes on with our recent explorations on leveraging scientific laws to push AI capabilities beyond data, from improving the generalizability of electronic Hamiltonian, to correcting data bias in molecular structures. Combining the power from the two sides, deeper scientific understanding and discovery could be expected in the future.

【报告人简介】刘畅，现任北京中关村学院研究员，博士毕业于清华大学计算机科学与技术系，曾在微软亚洲研究院、微软研究院科学智能中心工作。主要从事采样方法，生成式模型，以及AI赋能的科学研究等方面的工作，包括使用AI突破量子化学算法精度与效率的矛盾，使用生成模型实现统计物理系综建模以及分子体系设计等。相关工作发表在顶级学术会议和期刊上，包括《自然-计算科学》和《自然-机器智能》等，并在AI主流学术会议ICML，NeurIPS，ICLR等担任领域主席。