Academician E Weinan: AI empowerment will change the workshop model of scientific research, but it is necessary to avoid speculating on concepts

The wave of AI is sweeping across, changing the ecology of many industries, including the methods of academic research. However, AI-driven academic research currently faces difficulties such as lack of resource support. On May 30, at the 2023 Zhongguancun Forum "Artificial Intelligence Driven Scientific Research Forum", E Weinan, academician of the Chinese Academy of Sciences and director of the Beijing Institute of Scientific Intelligence, mentioned that AI empowerment will completely revolutionize scientific research methods, but speculation needs to be avoided. Concepts, superficial prosperity, and problems that cannot really be implemented.

E Weinan said that in the traditional scientific research system, there are two methods: data-driven and basic principle-driven. However, in practice, the former often faces the dilemma of low data collection efficiency and lack of effective data analysis methods. The latter have been basically "exhausted", "and the efficiency of using basic principles to solve practical problems is relatively low, because mathematical problems expressing basic principles are too difficult."

The problem this brings is that academic research can be done well on academic issues such as structural mechanics, but when faced with complex issues such as material design and drug design, it can only be found through experience and trial and error. right direction. This is because there are higher degrees of freedom in complex problems. "From a mathematical point of view, it is dimensionality," Eweinan explained. "Why structural mechanics is relatively simple is because it has fewer degrees of freedom and drugs are more complex. , because it is a multi-body problem with relatively high degrees of freedom and dimensionality. The increase in dimensionality brings about the disaster of dimensionality."

And AI can help solve this dilemma. E Weinan pointed out that AI provides new and efficient tools for data-driven models, which can improve the reliability and efficiency of basic principle-driven models, and can also combine data-driven and basic principle-driven models. The most classic example of this is the molecular dynamics DBM tool. "Quantum mechanical precision molecular dynamics is a very basic tool in chemical calculation materials science, but even if high-performance computing is used, it can only handle thousands of atoms before, so DBM calculations, that is, adding artificial intelligence Intelligent tools can quickly make it into hundreds of millions or even tens of billions."

Therefore, when AI empowers scientific research, it will inevitably bring about changes in productivity and production relations. E Weinan mentioned that the four basic tools for scientific research are basic principles and data analysis methods, experiments, literature, and computing power. AI will bring about four tool innovations and break the "workshop model" of previous scientific research, which is a long-cycle, inefficient operation method.

E Weinan believes that AI-empowered scientific research will comprehensively change the pattern of scientific research and industrial innovation. For this reason, E Weinan said, "AI for Science is the best opportunity in the entire history of China's scientific and technological innovation."

AI plays an important role in scientific research, so various platforms have successively launched open source tools and knowledge bases. In 2018, AI empowered scientific research was first proposed. Last year, the Beijing Institute of Scientific Intelligence was established, becoming the first international research institution with the theme of AI empowered scientific research, and launched the DeepModeling open source community platform for basic principle research.

However, E Weinan also pointed out that relevant open source platforms and research institutes are facing the problem of insufficient resources - "We are trying to find ways to use the few resources we have to do this."

In addition, the future development of AI-enabled scientific research may also face the possibility of "hype concepts, superficial prosperity, and failure to truly implement". To this end, E Weinan suggested that we need to adhere to a rigorous scientific research style and adhere to an attitude of openness, sharing, and win-win cooperation from the foundation.

, but to move towards higher-end artificial intelligence development. ". I hope we can take the lead in China this time to create a new scientific research paradigm of vertical integration of platforms." E Weinan said.

Written by: Nandu reporter Hu Gengshuo

The above is the detailed content of Academician E Weinan: AI empowerment will change the workshop model of scientific research, but it is necessary to avoid speculating on concepts. For more information, please follow other related articles on the PHP Chinese website!