Some observations to the recent progress of AI

Prof. Gong Ke is currently the President of the World Federation of Engineering Organizations (WFEO), the Executive Director of the Chinese Institute for the New Generation Artificial Intelligence Development Strategies, and the Chairman of the Academic Committee of Nankai University. In Jan. 1982, he obtained his bachelor's degree in Electronic Engineering from Beijing Institute of Technology and then studied further in Graz University of Technology, Austria, and obtained PhD in Technological Science in Nov.1986. He returned to China and joined Tsinghua University in 1987. During his service at Tsinghua University, he had been the Dean of the Department of Electronic Engineering, Director of Tsinghua Aerospace Research Center, Deputy Dean of the Graduate School, Director of the Science & Technology Office, Director of Chinese State Key Laboratory on Microwave and Digital Communications, and Director of Chinese National Research Center for Information Science and Technology (formerly known as the National Lab) at Tsinghua. From 1999-2006 he was vice president of Tsinghua University for research and strategic planning. From 2006-2011 he was President of Tianjin University, then the President of Nankai University from 2011 to 2018.

Prof. Gong has led a number of research projects in wireless communications including the initiative on developing China’s wireless transmission standard for digital television as well as the research on experimental microsatellite, through which he has received a number of awards including the National Technological Invention Award. He is foreign member of the Russian Academy of Cosmonautics. He has served as member the Experts Committee of the National High-tech R&D Program (2001-2010) for space technology, and served as member of standing committee of Chinese Association of Science and Technology (2006-2016), etc.


Abstract: The COVID-19 has driven wide applications of AI and its R&D. Base on the works of the institute and referring to the opened studied of other influential think tanks, some progresses of AI are shown in different aspects. It is found that, (1) Having fast adopting the AI technologies, the “non-core sector of AI” or “AI application sector” is going to replace the position of the core AI sector in the AI industrial symbiosis, which shows the AI is penetrating into wide industrial sectors; (2) additional to the various application of AI in combating COVID-19, AI has been more widely and effectively used to empower fundamental scientific research, such as to solve Schroedinger equation and to predict possible protein structures; (3) Regarding the current technical bottle-necks of widely used data-driven DL algorithms, such as its data dependency, non-interpretability, low energy efficiency, etc., though there are some exciting research progresses  have been made, breakthroughs yet to be expected; (4) Based on the AI governance principles proposed,  notable affords are made to turn these principles into trustworthy AI systems, but the urgent needs for global consensus and coordinated works on establishing guiding ethical rules and technical standards have not yet been met.



Honggang-Zhang-2018Toward Native Intelligence in 6G Networks: A Paradigm Changing for Future Networking Architectures

Prof. Zhang Honggang

College of Information Science & Electronic Engineering

Zhejiang University



Abstract: The mobile communication system has transformed to be the fundamental infrastructure to support digital demands from all industry & society sectors, and future 6G is envisioned to go far beyond the communication-only purpose. There is coming to an expectation that 6G will treat Artificial Intelligence (AI) as the cornerstone and has a potential capability to provide “intelligence inclusion”, which implies to enable the access of AI services at anytime and anywhere by anyone.

Particularly, the intelligent inclusion vision generates far-reaching influence on the corresponding network architecture design in 6G and deserves a fundamental rethink. Accordingly, we initiate a native intelligence network architecture design for 6G, and analyze the necessity to incorporate an independent data plane as well as a novel intelligent plane with specific emphasis on AI workflow management and orchestration within 6G networks. We also highlight the advantages to provision converged connectivity and computing services at the network function plane. Benefiting from these approaches, we believe that 6G will become an intelligence inclusive platform.






Data-poisoning Attacks Threat in Semi-supervised Learning

Prof. Li Shenghong is a Tenured Full Professor in Shanghai Jiao Tong University, and the Deputy Director of Key Laboratory for Shanghai Integrated Information Security Management Technology Research. He is an IEEE Senior Member, a New Century Talent of Chinese Education Ministry, and a Shanghai Dawn Scholar. His research interests include network and information Security, information processing, artificial intelligence. In the past years, he has taken charge of a set of national/provincial level research projects including Chinese National Natural Science Foundation Project, National Basic Research Program of China (973 Plan), Chinese National Key Technologies R&D Program Project, etc., has authored or co-authored hundreds of papers published on journals and conferences including IEEE Trans. on CyberneticsIEEE Trans. on CSVTIEEE Trans. on TETCI IEEE Trans. on Commu.IEEE IOT JournalIEEE Trans. on SPIEEE JSAC and AAAISIGKDDICIPGlobecom, etc., and has more than 20 Chinese invention patents authorized by SIPO of P.R. China. Dr. Li has also received some academic awards and honors, including The 1st Award of Shanghai Science and Technology Progress in 2003 and 2013 respectively, The 2nd Award of Science and Technology Progress of Chinese Institute of Electronics in 2016.


AbstractSemi-supervised Learning (SSL) is a key approach toward more data-efficient machine learning (ML) by leveraging extra unlabeled data. Due to the availability of a vast amount of unlabeled data, breakthroughs in SSL can dramatically advance the application of ML in many fields. However, also caused by the vast availability of unlabeled data, the notorious data-poisoning attacks is extremely likely to be conducted on the unlabeled data in SSL. Here, we overview the typical methods of data-poisoning attacks in the context of SSL. Moreover, we identify a new insidious threat of SSL where unlabeled training data are poisoned by backdoor methods migrated from supervised settings. To further exploit this threat, a Deep Neural Backdoor (DeNeB) scheme is proposed, which can achieve the stronger illegal backdoor manipulation effectiveness on the trained model by poisoning less unlabeled training data. Our investigation clearly demonstrates that SSL is extremely vulnernable to data-poisoning attacks, and courtermesures should be developed to enable robust and secured semi-supervised learning.



mmexport1519395449873_mh1519396094164Brain Computer Interface technology: Recent Progress and Challenges

Dr. Jianguo Jia, Previous expert of Huawei technology (Retired at March 2021). Dr. Jia had worked in telecommunication industry for 22 years (12 years at Huawei technology, and 10 years at Alcatel/Alcatel Lucent). He has experience in fixed network and mobile network commutation system. Recent years, his interest shifted to the AI and Complex Intelligent System research, He is the expert of Huawei Intelligent Computing System and the certified moralist of Huawei Ascend Computing System. Dr. Jia is also the expert of International System Dynamics Society, he is the vice director of System Dynamics Society of China, and has worked for complex dynamic system modeling about 30 years. Dr. Jia had his Ph.D. in Management Science and Master Degree in Physics at Fudan University. He had his Bachelor degree in Engineering Physics at Tsinghua University.


Abstract: Brain Computer Interface (BCI) Technology devoted to connect the human brain and external artificial intelligent system has being developed for many years. Recently, with the progress of Brain Science, and also the great progress of AI and complex Intelligent system technology, BCI has become an international research hotspot and also an industry hotspot. However, the existing BCI technology and its application still face many problems and challenges. In this paper we try to summarize the recent progress of BCI technology and also the related industry, and analyze the big problems and challenges encountered.