On November 29, 2024, the 3rd International Summit Forum on AI4S "Beyond the Future" was held at the International Conference Center of Peking University Shenzhen Graduate School. The forum gathered a distinguished assembly of guests, including 1 Nobel laureate, 5 academicians from home and abroad, as well as over a dozen top experts, scholars, and enterprise leaders from China and overseas. They conducted in-depth, multi-dimensional, and cross-industry exchanges and collisions of ideas around core topics, jointly exploring the infinite possibilities of artificial intelligence empowering science.
Forum Scene
Academician Zhang Jin, Member of the Chinese Academy of Sciences, Vice President of Peking University, and Dean of Shenzhen Graduate School, together with Academician E Weinan, Member of the Chinese Academy of Sciences, Chairman of the Beijing Academy of Artificial Intelligence, Director of the International Machine Learning Research Center of Peking University, and Chair Professor of the School of Mathematical Sciences of Peking University, served as co-chairs of the forum. The forum was co-hosted by Peking University Shenzhen Graduate School and the International Machine Learning Research Center of Peking University, with strong support from institutions such as Beijing Baidu Netcom Science and Technology Co., Ltd., Beijing DeepMind Technology Co., Ltd., Beckman Coulter International Trading (Shanghai) Co., Ltd., Huawei Technologies Co., Ltd., Shenzhen Computer Society (SZCCF), CCF Shenzhen, Global Alliance for the Development of Scientific Intelligence, Guangdong Artificial Intelligence Industry Association, and Education Foundation of Peking University Shenzhen Graduate School. The forum was live-streamed on platforms including Peking University’s official accounts (Kuaishou, Bilibili, Baidu Baijiahao, Douyin, WeChat Channels), DeepMind Technology, Guangdong Artificial Intelligence Industry Association, Baidu AI and PaddlePaddle, and Shenzhen TV’s "Frontiers of Science and Innovation," attracting nearly 300,000 online views.
The opening session featured an opening speech by Academician Zhang Jin. The morning keynote reports were hosted by Professor Yang Zhen, Executive Dean of Peking University Shenzhen Graduate School; the roundtable dialogue and signing ceremony were hosted by Professor Tian Yonghong, Dean of the School of Information Engineering at Peking University; and the afternoon keynote speeches were hosted by Assistant Professor Chen Zhaolong from the School of New Materials at Peking University Shenzhen Graduate School.
At the opening of the forum, Konstantin Novoselov, 2010 Nobel Laureate in Physics, Fellow of the Royal Society, and Distinguished Professor of Materials Science and Engineering at the National University of Singapore, delivered a report titled "New Approaches in Materials Design." He shared insights on designing intelligent functional materials with biological characteristics, delved into the concept of "dream materials" with self-healing and memory functions, and introduced the application of machine learning and automated synthesis technologies in materials design. He particularly emphasized the importance of "artificial intelligence in developing new materials to address challenges in medicine, technology, and sustainable development."
Professor Tian Yonghong Introducing Professor Konstantin Novoselov
Professor Zhang Dongxiao, Member of the US National Academy of Engineering, Executive Vice President and Provost of NingboDongfang University of Technology(tentative name), shared a report titled "Scientific Machine Learning: AI-Driven Knowledge Discovery," introducing how artificial intelligence (AI) drives the development of scientific research. The report elaborated on the latest progress of scientific machine learning in the field of scientific knowledge discovery, especially the successful extraction of Burgers equations with interaction terms, KdV equations with high-order derivatives, and Chafee-Infante equations with exponential terms through the combination of symbolic mathematics and AI algorithms (SGA), verifying the accuracy and robustness of this method. In addition, he introduced cutting-edge models such as SGA, EqGPT, and the large language model LLM4ED applied to scientific exploration, providing scientists with new methods to bridge knowledge and data barriers.
Professor Zhang Dongxiao Delivering a Keynote Report
Professor Guo Yike, Fellow of the Royal Academy of Engineering, Fellow of the European Academy of Sciences, Fellow of the Hong Kong Academy of Engineering Sciences, Vice President and Chair Professor of the Hong Kong University of Science and Technology, and Director of the Hong Kong Generative AI Research Center, delivered a report titled "Cultivating Machine Scientists: An Engineering Perspective," sharing frontier research on using large model alignment technology to achieve model self-evolution. He explored how to enable models to self-optimize based on feedback and environmental changes through alignment technology, thereby enhancing their performance in scientific research. He proposed a new framework that combines adaptive learning mechanisms and knowledge transfer methods, allowing models to continuously improve their research capabilities through experimental verification and iterative learning, and even generate new scientific hypotheses. This method will provide new perspectives and directions for future scientific intelligence research.
Professor Guo Yike Delivering a Keynote Report
Professor Zhang Jin, Member of the Chinese Academy of Sciences, Vice President of Peking University, and Dean of Shenzhen Graduate School, delivered a report titled "Can AI Revolutionize the Research Paradigm of Materials Science?" He in-depth discussed the potential transformative role of artificial intelligence in materials science research. He pointed out that materials science faces issues such as high system complexity, low data standardization, and long research chains, exposing the limitations of traditional research paradigms. Academician Zhang emphasized that AI technology excels at processing high-dimensional, multi-scale data and can explore complex correlations between parameters, offering new possibilities to address the dilemmas in materials science research. He explored how AI technology can revolutionize materials research paradigms from three aspects: "AI for materials characterization," "AI for controlled materials preparation," and "AI for materials industrialization," aiding the standardization of materials databases, the comprehensiveness of basic materials research, and the leap from laboratory to industrialization of materials. He also proposed the vision of "AI Agent for Science," envisioning a new and rapidly developing future that AI will bring to materials science research.
Academician Zhang Jin Delivering a Keynote Report
Professor David Srolovitz, Member of the US National Academy of Engineering and Dean of the Faculty of Engineering at the University of Hong Kong, delivered a report titled "AI Models for Atomic-Scale Materials Design," exploring the application of artificial intelligence in materials science and engineering. He noted that AI methods are widely used in materials selection, performance prediction, and inverse materials design. In his speech, he focused on the role of machine learning in the development of interatomic potentials and the automation of atomic modeling, as well as the application of large language models (LLMs) in materials classification and performance prediction. Finally, Professor Srolovitz looked ahead to "future directions of AI in materials science," including its potential to reveal fundamental materials laws, demonstrating the great possibility of AI in advancing materials science.
Professor David Srolovitz Delivering a Keynote Report
In the roundtable dialogue session, Professor E Weinan, Professor Zhang Dongxiao, Professor David Srolovitz, Professor Turab Lookman, and Huang Youyuan, Executive Vice Chairman and General Manager of BETRICH New Materials Group Co., Ltd., discussed the theme "AI for Science: How to Achieve Integration of Education, Technology-Industry-Talent." Experts shared their insights, in-depth exploring how to achieve collaborative development and integrated advancement in education, technology, industry, and talent under the background of AI-science integration, providing valuable ideas and suggestions for the comprehensive, coordinated, and sustainable development of related fields in the future.
Subsequently, the forum held the release ceremony of AI4S annual achievements at Peking University Shenzhen Graduate School and the signing ceremony of the cooperation agreement for the Nanyan Scientific Computing Learning and Training Platform. Two achievements were released: "AI-empowered design and preparation of high-performanceOlefin carbon fiber" and "big data-driven efficient and controllable protein design." The signing of the cooperation agreement for the Nanyan Scientific Computing Learning and Training Platform will further promote university-enterprise cooperation in the Greater Bay Area, providing strong support for cultivating high-quality interdisciplinary talents in China’s AI4S field.
At the opening of the afternoon keynote speeches, Professor Turab Lookman mentioned in his report titledGuiding Next Experiments: Decisions, Decisions, Decisionsthat Bayesian Global Optimization (BGO) has become the preferred method in many recent studies since its application in materials science in 2015. He demonstrated how reinforcement learning can be applied to accelerate materials discovery, presented verification results of synthetic optimization functions, and clarified the advantages and disadvantages of related methods under different conditions, as well as how to identify high-value regions of targets/properties through experimental iterations.
Professor Turab Lookman Delivering a Keynote Speech
Professor Pan Feng, Vice Dean of Peking University Shenzhen Graduate School and Founding Dean of the School of New Materials at Shenzhen Graduate School, delivered a report titledMaterials Gene Exploration Based on Graph Theory and AI, exploring research methods for material genes and their structure-function relationships in lithium-ion batteries. Through the establishment of graph theory-based chemical methods and a material big data system, he aimed to address fundamental questions such as "What is a material gene?" and "How to study material genes in lithium-ion batteries?" Dean Pan introduced several cutting-edge interdisciplinary fields, including graph theory-based structural chemistry, material big data, lithium-ion battery material genes, and structural characterization using large scientific facilities such as synchrotron radiation and neutron radiation.
Professor Pan Feng Delivering a Keynote Speech
Professor Peng Shaoliang, Deputy Director of the National Supercomputing Center in Changsha, delivered a report titledMultimodal Biomedical Big Data and Large AI Models, discussing how life sciences have fully entered the digital intelligence era driven by big data and artificial intelligence, covering areas from single-cell omics to population data, from electronic medical records to medical imaging, and from Western medicine to traditional Chinese medicine. The report focused on multimodal data, thought chains, and intelligent agents in the biomedical field, as well as the storage and computing challenges they face, along with the latest research progress of his team and the global community.
Professor Peng Shaoliang Delivering a Keynote Speech
Tng Tai Hou, Innovation Lead of the Computational Intelligence Department at ASTAR’s High Performance Computing Institute, delivered a speech titledCo-creating the AI Healthcare Future Now. He emphasized our responsibility to bring benefits to humanity and discussed how to co-create the future of AI healthcare at the current stage. The speech covered interdisciplinary collaboration among AI scientists, engineers, clinicians, nursing professionals, communities, government agencies, and enterprises, exploring how such collaboration can drive innovation and development in the healthcare field.
Tng Tai Hou Delivering a Keynote Speech
The scientific community still faces challenges such as high computational costs for multi-spatiotemporal scale phenomena and low efficiency in multi-source data fusion applications. The rapid development of deep learning technology has brought new research ideas. Ma Yanjun, General Manager of Baidu’s AI Technology Ecosystem, delivered a report titledDeep Learning Open Source Platform Empowering Scientific Exploration, sharing the important role of deep learning technology in scientific exploration and detailing key technologies, tool components, and application cases of the PaddlePaddle deep learning platform for large model and scientific computing research and development. In addition, the report discussed how to promote ecological cooperation in the field of AI scientific computing through open-source initiatives, facilitating integration and innovation between science and AI.
General Manager Ma Yanjun Delivering a Keynote Speech
Ke Guolin, Partner and Senior Vice President of AI at Beijing DeepMind Technology Co., Ltd., delivered a report titledMolecular Design Based on the Universal Molecular Foundation Model Uni-Mol, introducing breakthrough progress of Uni-Mol in the field of molecular design. The report also covered specific achievements of Uni-Mol in drug discovery, OLED material design, MOF gas adsorption prediction, and coolant design, demonstrating its broad prospects in scientific and industrial innovation.
Vice President Ke Guolin Delivering a Keynote Speech
Dr. Qiao Nan, Chief Healthcare Scientist at Huawei, delivered a speech titledHuawei’s Exploration and Practice in the Field of AI4Science, introducing Huawei’s strategic deployment and practical experience in artificial intelligence. Since releasing its artificial intelligence development strategy and full-stack, all-scenario AI solutions in 2018, Huawei has firmly advanced its AI strategy, aiming to build a solid computing power foundation for China and provide a second option for the world. In the field of AI for Science, Huawei’s research teams have collaborated with industry experts to actively carry out exploration and practice, covering fields such as gene sequencing, drug research and development, clinical research, and meteorological research, and have achieved a series of remarkable results.
Chief Scientist Qiao Nan Delivering a Keynote Speech
Finally, Professor Tian Yonghong, Dean of the School of Information Engineering at Peking University and Boya Distinguished Professor of Peking University, delivered the closing speech and expressed gratitude on behalf of the forum organizers to all experts, scholars, and participants for their attendance. In this forum, reports from numerous academic experts and industrial elites at home and abroad not only demonstrated that the AI4S scientific research paradigm is continuously breaking boundaries in frontier exploration but also presented the vigorous development of AI4S in industrial applications. It is hoped that everyone will continue to pay attention to the "Beyond the Future" AI4S series forums at Peking University Shenzhen Graduate School and jointly discuss and share the transformation and development of AI4S.
