[Remember to entrust new achievements] Learn from a strong country ｜ China University of Science and Technology： ＂Machine Chemist＂ brings a new paradigm of scientific research
People’s Daily Wu Yuehui
In the "Machine Chemist" laboratory, Jiang Jun is debugging the operation instructions of the chemical experiment for the robot.
Zhang Dagang photo
As a major trend in the development of artificial intelligence, international academic circles have reached a consensus on "scientific research driven by artificial intelligence": artificial intelligence will bring about changes in scientific research paradigms and new industrial formats.
On the campus of China University of Science and Technology, a group of researchers from the School of Chemistry and Materials Science are actively engaged in this practice: they are deeply involved in the field of precise and intelligent chemistry, promoting the transformation of scientific research paradigm, and have achieved a series of remarkable scientific research results.
"Machine Chemist" shows great advantages of the new paradigm of intelligence.
How to create a Fenton catalyst? In the "Machine Chemist" laboratory of the Chinese University of Science and Technology, a large number of experiments of bottles and cans are "replaced" by artificial intelligence, and researchers do not need to try. The whole process is simple and efficient.
The reporter saw in the laboratory that researchers input questions in artificial intelligence programs: What types of non-precious metal elements are commonly used in Fenton catalysts? Soon, the program will give the answer. The answer provided by the program comes from the self-developed document machine reading system, which can quickly read a large number of documents and help researchers choose the best combination of elements based on statistical data analysis. Next, we can call up the experimental template of Fenton catalyst saved in the platform of "Machine Chemist", edit the parameters of the liquid injection station according to the combination of elements recommended by artificial intelligence, and let the platform of "Machine Chemist" named "Xiaolai" help with the experimental verification. In this way, "Xiaolai" can start its journey of creating Fenton catalyst.
"After the experimental data is processed, it is input into the unique computing brain of Xiaolai to generate an artificial intelligence model, which can help researchers optimize the experimental scheme." Jiang Jun, a professor at the School of Chemistry and Materials Science of the Chinese University of Science and Technology, told the reporter.
Artificial intelligence tools and platforms such as "Xiaolai" are the research and development results of Jiang Jun’s team from the School of Chemistry and Materials Science of the Chinese University of Science and Technology. Among them, the artificial intelligence program is driven by chemical data and combined with the knowledge of human chemists to carry out machine learning training, which can give preliminary experimental suggestions for the questions raised by users. "Xiaolai" is a "full-process machine chemist" platform independently developed by the team, which integrates reading literature, independently designing experiments and developing materials. It can find the optimal solution from hundreds of millions of possible combinations, thus accelerating the research and development of materials.
In the laboratory, the mechanical arm driven by "Xiaolai" can stretch freely and accurately grasp the prepared reagents.
How powerful is the "full-process machine chemist" platform? Take the high-entropy compound catalyst with great potential as an example: to obtain the optimal formula, it is necessary to test an extremely large combination of chemical ratios. If we rely on the traditional research paradigm, this process may take 1400 years, while "machine chemist" takes advantage of data-driven and intelligent optimization to find the optimal high-entropy catalyst from 550,000 possible metal ratios, which takes only five weeks.
Experts believe that the research work of this "machine chemist" has got rid of the limitations of the traditional research paradigm and demonstrated the great advantages of the new paradigm of intelligence.
Using artificial intelligence to digitize and code scientific knowledge
Guess, try, correct mistakes, guess again, try again … In the past 150 years, the traditional chemical research paradigm has been deeply dependent on the "trial and error method", and its limitations make the cycle of material creation long and costly, making it difficult to achieve high efficiency and energy saving.
Since then, quantum chemistry developed from quantum mechanics has become a tool used by chemists. Chemists can carry out simulation experiments on the computer to verify a certain theory, which greatly improves the efficiency. However, the research object of chemistry is becoming more and more complex and high-dimensional. Faced with the huge chemical space, the search for formulas and processes often stops at the local optimum, and it is impossible to explore the whole situation.
Jiang Jun, who has 10 years of research experience in computational chemistry, said with emotion: "Our chemical system is very complicated. Although supercomputing has evolved rapidly, it still cannot cope with its complexity."
What should we do? New methods must be found.
Like Jiang Jun, Li Zhenyu, a professor at the Chinese University of Science and Technology, has been paying attention to and thinking about this issue.
"Precision is a dream of all chemists. I hope to put this thing in, and what I want it to be transformed into can be transformed into 100%, and nothing else will be produced in the process. " Li Zhenyu said, "This requires that the whole chemical research can be accurately designed, characterized, prepared and regulated, and that the whole process is transparent and controllable, and the mechanism inside is clear. To achieve this goal, the entire research paradigm must change. "
Big data and artificial intelligence technologies, which are gradually emerging and rapidly evolving iteratively, have made Jiang Jun and Li Zhenyu see the hope of solving these problems.
Jiang Jun believes that big data and artificial intelligence technology can digitize, code and migrate scientific knowledge. "That is to say, we can call the code invented by excellent researchers at any time. The crystallization of their wisdom, as long as I use a’ sub-function’ can be called. If you only rely on the human brain to learn, practice and train, the whole process will be very long and the transfer of knowledge will become inefficient. "
So, an idea came to Jiang Jun’s mind: Can a new tool be developed with the help of artificial intelligence technology? In his conception, this tool can help scientists break through the limitations of thinking and use data to establish effective and complex models, thus guiding chemical practice.
In 2014, Jiang Jun’s team put forward the concept of "machine chemist" and carried out related scientific research work. After eight years of research, the team successfully developed a "full-process machine chemist" driven by data intelligence by developing and integrating technologies such as mobile robots, chemical workstations, intelligent operating systems and scientific databases in 2022.
"Machine Chemist" Helps Promote a New Paradigm of Chemical Research
Nowadays, "full-process machine chemist" plays an active role in scientific research. When Zou Gang team of Chinese University of Science and Technology screened optically active film materials, in order to find the target materials, it was necessary to mix a variety of molecules to control the film thickness, stress, gray level and other technological conditions, and there were millions of possibilities. After 10 years of hard work, the team finally raised the asymmetry factor to 1.2, but there is still a big gap from the theoretical limit of 2.0. With the help of "Xiaolai", they found the technological conditions with an asymmetry factor of 1.95 within two months, which is highly close to the theoretical limit.
Jiang Jun successfully took the first step, which greatly encouraged the chemical scientists and strengthened their determination to promote the transformation of chemical research paradigm with artificial intelligence technology.
In January this year, the Key Laboratory of Precision and Intelligent Chemistry of China Academy of Sciences was officially approved for construction, with Li Zhenyu as the laboratory director. "The laboratory is mainly oriented to the forefront of world science and technology, focusing on how to change the paradigm of chemical research, and exploring the establishment of a precise and intelligent dual-drive model for chemical research." Li Zhenyu introduced.
Li Zhenyu believes that the biggest difficulty and challenge of applying artificial intelligence technology to chemical research at present comes from data. "The existing large amount of data has complex sources and uneven quality. These data are mixed together and let artificial intelligence learn, and it is likely to learn some wrong knowledge. Therefore, we hope to develop some new technologies that can be characterized with higher accuracy, and at the same time, we can form a set of data standards, and do data-driven intelligent chemistry on this basis. "
The emergence of "machine chemists" freed chemists’ hands. Some people worry that chemists may have nothing to do in the future. Jiang Jun said that there is no need to worry about this: "A good technical tool will also give more possibilities for researchers to do more things and discover more cutting-edge theories."
The appearance of "machine chemist" has promoted the change of chemistry research paradigm and put forward new requirements for the future development of chemistry. Jiang Jun said, for example, in personnel training, it is necessary to have a solid chemical foundation, an open mind, and be good at and dare to learn all kinds of new knowledge.
In the future, Jiang Jun hopes to build a "machine chemist" scientific apparatus: in a whole building, there are hundreds of robots and thousands of intelligent chemical workstations. Based on such a large platform, the experimental data of various research groups can meet and share, generate massive data, automatically extract digital knowledge maps and artificial intelligence models, and then guide robots to automatically optimize the production of better and more efficient chemicals or new materials, and realize a new paradigm of chemical research driven by data intelligence.
For the first key applied research in the Key Laboratory of Precision and Intelligent Chemistry, Li Zhenyu and his colleagues now have a clear goal: based on the new paradigm of chemical research driven by precision and intelligence, comprehensively evaluate the existing reactions related to nitrogen resource transformation, propose a new path of green and low energy consumption transformation, create a new catalytic system, and achieve a breakthrough in the field of comprehensive utilization of nitrogen resources.
Li Zhenyu said: "Efficient transformation of nitrogen resources is a very challenging problem. I hope that a new research paradigm can help us."
Article source: https://article.xuexi.cn/articles/index.html? art_id=1625621815117764427&item_id=1625621815117764427&study_style_id=feeds_default&pid=&ptype=-1&source=share&share_to=wx_single