"at present, we are at the starting point of the fifth industrial revolution, and artificial intelligence is profoundly changing our research and development paradigm in the field of chemistry and materials." At the artificial Intelligence (AI) Symposium in the Field of Chemistry and Materials Science held in Yantai on August 17, Huang Qishan, vice president of Wanhua Academia Sinica, pointed out that with the large model of chemistry and materials as the entrance, AI assisted literature reading and experimental design, and the AI model was created through high-throughput experimental platform, robotic chemists with various innovative methods of computational chemistry, and dry-wet experiments to build large model data sets. It is helping us to better carry out large-scale material screening and new molecular discovery.

Shen Xiangjian, director of the Chemical Theory and Mechanism Program of the Chemistry Department of the National Natural Science Foundation of China, also pointed out that the arrival of the AI era has changed the original "trial and error" model, greatly shortened the research and development time of molecular materials, and reduced labor costs.

It is understood that machine learning is the core of AI and the fundamental way to make computers intelligent. It uses numerical algorithms and statistical models to guide the computer system to find rules in a large number of data, and use this knowledge to make predictions or judgments. "now, without machine learning, we basically can't do anything." Zhang Donghui, academician of the Chinese Academy of Sciences, said in an interview with reporters, "through massive data, machine learning can predict the molecular configuration of new materials." It helps us to reduce dimensionality, simplify very complex problems in high-dimensional space, and make things that cannot be done in the past possible. "

However, there are still difficulties in the development of AI in the field of chemistry and material science. "the key bottleneck of AI application is data, and the industry has basically excavated the dividend of public data. The way to solve data problems is the combination of automation and intelligence, and the core is to change the way data is generated. " Yang Mingjun, chief research scientist of Shenzhen Jingtai Technology Co., Ltd., said, "from the perspective of enterprises, I think a lot of exploration needs to be done from the aspects of business model, engineering optimization and multi-element combination decision-making. form an effective iteration of the business-data-model."

In the face of the two problems of scarcity of data and oversimplification of material theoretical model, Dr. Chen Yue of Yantai Capital Institute of material Genome Engineering said that the institute has developed three core technologies: computational chemistry-assisted material data reorganization (DRCS), which removes the limitation of lack of material data, and theoretical chemistry-enhanced AI model (AI-TEML) to solve the disadvantage of insufficient extrapolation ability of machine learning. The material evolution system based on small samples can realize the automatic and intelligent iteration of new materials. The related technology has been successfully applied to the research, development and industrialization of OLED optical extraction materials, electronic transport materials and hole transport materials of Yantai Xianhua Science and Technology Group Co., Ltd., which has been incubated into a "Hessen AI platform" and plans to be applied and promoted for the majority of new materials enterprises.

What is the prospect of commercialization of material AI? Jiang Jun, a professor at the University of Science and Technology of China, said that the industry is still in its infancy and has many opportunities, but there are also blind spots in research. He stressed that in the future, the academic emphasis on model development should shift from "absolute accuracy" to "demand docking", think from the point of view of the application side, grasp the two keys of "pre-training" and "active learning", and enhance the practicability of AI model in the field of new chemical materials through man-machine cooperation.

This meeting is hosted by Yantai Capital material Genome Engineering Research Institute and Yantai Base of Guotou Science and Technology Innovation Co., Ltd. Chen Xuebo, professor of Beijing normal University, Yang Huai, professor of Peking University, Zhu Tong, professor of East China normal University, Zhou Yanhua, professor of Southeast University, ao Yufei, associate researcher of Institute of Chemistry, Chinese Academy of Sciences, Shen Lin, professor of Beijing normal University, and Ouqi, associate researcher of Sinopec Petrochemical Research Institute Co., Ltd. Feng Peichuan, chairman of Yantai Xianhua Technology Group Co., Ltd., and other enterprise experts attended to share the report and conducted in-depth discussions on the topics such as machine learning pre-training, model transfer, and future industrial application direction. Experts at the meeting said that, as Academician Zhang Donghui said, it is difficult to rely on manual models for innovation on the basis of the vast amount of data available. It is believed that the future material AI will shine in the fields of medicine, electronic chemicals and polymer materials, and become an important solution for the innovative development of new productivity. Source: Sinochem New Network