Microsoft Research Asia Establishes Theory Center to Strengthen Theoretical Foundation of AI
On December 11th, Microsoft Research Asia (MSR Asia) held the 2021 Theory Workshop, where theoretical research experts from academia and the industry came together to share their latest research results. At the workshop, MSR Asia announced the establishment of its Theory Center. The mission of this center is to promote in-depth integration of theoretical research, big data, and artificial intelligence technology by building international academic exchange and cooperation hubs. While driving progress of theoretical research, the center aims to strengthen interdisciplinary research cooperation, helping break through the bottleneck currently encountered in AI development, and realizing further substantial development of computer technology.
The inauguration ceremony of MSR Asia Theory Center
Dr. Wei Chen, principal researcher at MSR Asia, serves as the chair of the Theory Center, and Dr. Wei Chen, principal research manager at MSR Asia, serves as the co-chair. The center has established an advisory committee composed of members including Peking University Chair Professor and Academia Europaea academician Dr. Xiaotie Deng, Gordon McKay Professor of Computer Science at Harvard University and Distinguished Scientist of Microsoft Dr. Cynthia Dwork, Cornell University professor and Turing Award winner Dr. John Hopcroft, Professor of Mathematics at University of Oxford and Royal Society Fellow Dr. Terry Lyons, mathematician and academician at the Chinese Academy of Sciences Dr. Zhi-Ming Ma, and Professor of Computer Science at the University of Southern California and Gödel Prize Winner Dr. Shang-Hua Teng (Alphabetical order by last name). Moving forward, they will offer suggestions on the developmental direction of the Theory Center, provide support for related programs and events, and evaluate the progress of the center.
Issue appointment letters to members of the advisory committee (from left to right: Outreach director at MSR Asia Xin Ma, assistant managing director of MSR Asia Dr. Tie-Yan Liu, mathematician and academician at the Chinese Academy of Sciences Dr. Zhi-Ming Ma, Cornell University professor and Turing Award winner Dr. John Hopcroft, Peking University Chair Professor and Academia Europaea academician Dr. Xiaotie Deng, principal research manager at MSR Asia Dr. Wei Chen, senior research program manager at MSR Asia Lily Sun, principal researcher at MSR Asia Dr. Wei Chen)
Dr. Lidong Zhou, managing director of MSR Asia, stated: “Theory is the foundation of computer science and the driving force of computer technology. Microsoft Research Asia has always set its goal on fostering curiosity-driven innovations with long-term significance and foundational impact through a diverse and inclusive research environment. We hope that the establishment of the Theory Center can create a theoretical core in the lab for solving fundamental problems in computer science and artificial intelligence, advocate the principled thinking behind scientific research, and strengthen our connections broadly with the international theory community.”
Dr. Lidong Zhou, managing director of MSR Asia
Right now, artificial intelligence technologies such as deep learning and reinforcement learning driven by big data and empirical validation are hitting a developmental bottleneck. Super-sized models trained by large-scale data and large computing power have yielded results comparable to or even surpassing human abilities in certain fields. However, this method can often draw researchers into the misconception that breakthrough results can be achieved so long as there is enough data and computing power, causing them to ignore the support of algorithms and theories and bringing challenges to deep learning and reinforcement learning in terms of robustness, reliability, interpretability, fairness, and many other aspects.
On the other hand, in the theoretical computer science field, if researchers are only focused on achieving more intricate algorithm theories, which causes divergence from practical applications, the scope of theoretical research would become increasingly narrow, and the resulting impact on the larger world would diminish. This does not comply with the original intent of theoretical research to support all fields of computer science.
Dr. John Hopcroft expressed: “Artificial intelligence, in recent years, has achieved great success in cross-industry and interdisciplinary applications due to continuous breakthroughs that have been made in deep learning technology. However, there still remain lots of problems for us to explore and understand surrounding artificial intelligence systems. Training an AI system, for example, requires that we understand how much data is needed and how you carry it out in order to be more effective and efficient, but unfortunately, as of now, we still do not have much theoretical support for this. We have to develop the relevant theoretical research.”
Dr. John Hopcroft, Cornell University professor and Turing Award winner
In order to tackle these new problems and challenges, researchers need to re-examine their method for carrying out scientific research. They need to strengthen the foundational support for mathematical theory, enhance the interpretability of artificial intelligence, and broaden the research scope of computer science theory, discovering new theoretical problems with the merging of biology, environmental science, and other subjects, and improving the versatility and impact of theoretical research.
“The development of artificial intelligence, big data, and machine learning has brought greater possibilities to mathematics,” said Dr. Zhi-Ming Ma, “and at the same time, artificial intelligence also increasingly depends on applied mathematics. To write an algorithm, you need to understand the mathematical mechanism behind it in order to have the opportunity to produce major innovative thinking.” He added, “I hope the establishment of the Microsoft research Asia Theory Center will better promote the integration of computer science theory and mathematics, construct a solid mathematical foundation for artificial intelligence, allow mathematics to more effectively support the development of computer technology, and at the same time, allow computer technology to better promote the development of theoretical research.”
Dr. Zhi-Ming Ma, mathematician and academician at the Chinese Academy of Sciences
Currently, the Theory Center has already gathered members from different teams and of different research backgrounds at MSR Asia and is focusing on solving fundamental problems in fields such as deep learning, reinforcement learning, dynamic system learning, and data-driven optimization. In the future, the center will systematically introduce post-doctoral programs, visiting scholar programs, intern programs, and regular academic lecture series, etc. In addition, the Theory Center also plans to hold summer programs and winter workshops to help scholars, researchers, faculties, and students from related fields to interact with each other and strengthen international academic exchanges.
The establishment of the MSR Asia Theory Center will help enhance consciousness of theoretical thinking among scientific researchers and improve the connection between academia and the industry. With the rise of fields such as computational science and data-driven optimization, new artificial intelligence technologies have created new opportunities for traditional sciences and optimization, but these technologies also urgently need theoretical support to realize their full potential. The Theory Center welcomes more researchers and students to participate in theoretical research and related discussions and jointly promote the development of computers and related studies.
The theory Center is looking for talents who are committed to theoretical research to join our research team and jointly promote new theoretical research in the era of big data and artificial intelligence. Please contact us at [email protected]