Guide
In order to fully implement the strategy of innovation-driven development and rejuvenating the country through science and education, thoroughly implement the deployment requirements of the "National New Generation Artificial Intelligence Development Plan", mobilize the enthusiasm and creativity of the majority of intelligent science and technology leading talents, promote the integrated development of general artificial intelligence and future industries, accelerate the development of new quality productivity and new industrialization, and build an industrial science and technology innovation center with global influence. From April 12th to 14th, 2024, the Chinese Society of Artificial Intelligence will hold the "'Innovation-Driven • Digital Intelligence Power' - The 13th Wu Wenjun Artificial Intelligence Science and Technology Award Ceremony and the 2023 Chinese Artificial Intelligence Industry Annual Conference" in Suzhou Industrial Park.
01
Background of the event
In recent years, with the rapid development of a new round of global scientific and technological revolution and industrial transformation, generative AI, large models and general artificial intelligence have become key areas for the emergence of new industries, new forms of business and new models. It is particularly important to promote the optimization and upgrading of the industrial chain and supply chain, strengthen the research of key core technologies of artificial intelligence, focus on solving major application and industrialization problems, actively cultivate emerging industries and future industries, strengthen the national strategic scientific and technological strength, accelerate the construction of new quality productivity, and build a digital industrial cluster with international competitiveness.
Encourage innovation and take responsibility intelligently. 2024 is the 30th anniversary of the development and construction of Suzhou Industrial Park by the Chinese government and the Singapore government, and it is also a key year for the Society to implement the spirit of the National People's Congress and the National People's Congress and the National People's Congress of the People's Republic of China, empower Suzhou to actively carry out the "Artificial Intelligence +" action, give full play to the exploration, leading and exemplary role of Suzhou Industrial Park, promote the deep integration of artificial intelligence and the real economy, and take high-quality development to a new level, strive to build a world-class high-tech park with openness and innovation, and expand exchanges and cooperation between the Society and local governments.
02
Forum features
The 2023 Chinese Artificial Intelligence Industry Annual Conference China Artificial Intelligence Industry Annual Conference (CAIIAC2023) was initiated and hosted by the Chinese Industrial Intelligence Society, and has successfully held nine industry annual conferences so far. As the theme supporting activity of the Wu Wenjun Artificial Intelligence Science and Technology Award Ceremony, the conference integrates closed-door discussions, honorary commendations, high-end forums, product displays, report releases and other core sections, and is an annual artificial intelligence landmark awards event with high authority, large scale, strong brand power and far-reaching industry influence in China.
With the theme of "Innovation-driven, Digital Intelligence Power", the conference set up a "1+10+X" model, including 1 main forum, 10 special forums, as well as authoritative awards, launching ceremonies, report releases, science nights and other special activities. The organizer invites leading experts from ministries and commissions, academicians of the Chinese Academy of Sciences and the Chinese Academy of Sciences, members of Wu Wenjun's Artificial Intelligence Science and Technology Award Committee, nominating and judging experts, award-winning representatives and entrepreneurs to discuss and exchange, give full play to the academic, talent, technological and industrial advantages of the award-winning project team and the achievements of artificial intelligence colleges, and set up a special report meeting of the president of "Intelligent System - Artificial Intelligence" in the theme report link, focusing on how to promote the coordinated development of the application of cutting-edge artificial intelligence technology and industrial ecology. By looking forward to the future trend of artificial intelligence, exchanging key core technologies and future industrial innovation, the intelligent system promotes the in-depth thinking and collision of views on the integration of industry and education in colleges and universities.
03
Report Highlights:
On the afternoon of April 14, the organizing committee of the conference will grandly launch the special forum on "Intelligent Chips and Integrated Circuits", and invite Professor Wang Zhongfeng, IEEE/AAIA Fellow and Dean of the School of Integrated Circuits of Sun Yat-sen University, to serve as the chairman of the forum. The forum was also honored to invite a number of outstanding experts from well-known universities and leading enterprises in China, who have made remarkable achievements in the field of artificial intelligence and integrated circuits, including national high-level leading talents, editors-in-chief of first-class magazines in the field of integrated circuits, "foreign senior scholars" of the National Natural Science Foundation of China, winners of the National Science Fund for Distinguished Young Scholars, and members of the national "new generation of artificial intelligence" major special expert group members and other scholars and experts gathered together. This forum focuses on the key common issues in the field of artificial intelligence systems and integrated circuit design, discusses the core technologies and industrial applications in the direction of large computing power chips, AI large models, brain-like computing, storage and computing integration, intelligent robots and unmanned systems, shares the cutting-edge challenges and innovation difficulties in the field of intelligent chips and integrated circuits with the majority of professional peers, and looks forward to the future development trend.
This forum aims to provide a platform to understand the cutting-edge technologies in the field of artificial intelligence and integrated circuits, explore the development path and opportunities of artificial intelligence chips through the exchange and collision of views, share new progress in related fields, and contribute high-end intellectual resources to promote the development of China's integrated circuit industry.
Here, the organizing committee of the forum sincerely invites colleagues from all walks of life to come to this conference to discuss the new changes, new opportunities and new challenges in the era of artificial intelligence.
04
Agenda of the Forum
05
Forum Organizing Committee
Science Technology
Chairman of the Forum - Wang Zhongfeng
National Distinguished Expert, IEEE and AAIA Fellow. Ph.D., University of Minnesota, USA. Since 2023, he has been the dean of the School of Integrated Circuits of Sun Yat-sen University. Previously, he worked at Nanjing University as a distinguished professor, and from 2007 to 2016, he worked at Broadcom in the United States as chief scientist and technical director. Earlier worked at Oregon State University and National Semiconductor. He is an internationally renowned expert in the design of low-power chips for signal processing systems, has participated in the design of more than 10 commercial chips, published more than 350 international papers, and won the best paper of the year award in IEEE mainstream conferences and journals for many times. In addition, he has made important contributions to international standards for Ethernet.
Science Technology
Forum moderator - Cheng Jian
He is currently a researcher at the Institute of Automation, Chinese Academy of Sciences, a professor at the University of Chinese Academy of Sciences, a national high-level leading talent, and the vice president of the Nanjing Institute of Artificial Intelligence Innovation. His research interests include deep learning, image and video content analysis, and intelligent chip architecture design. He has published more than 100 academic papers in high-level journals and conferences such as IEEE TPAMI, TNNLS, TCAD, JMLR, NeurIPS, ICML, ICLR, HPCA, CVPR, ICCV, AAI, etc. The related achievements have successively won the Lu Jiaxi Young Talent Award of the Chinese Academy of Sciences, the first prize of Science and Technology of the Chinese Institute of Electronics, the second prize of Science and Technology of the Chinese Society of Image and Graphics, and the first prize of Jiangsu Science and Technology Award.
06
Invited speaker
Science Technology
Topic: Large Model Compression and Accelerator Design
Speaker: Cheng Jian
Summary of the report
Large models represented by ChatGPT are one of the hottest topics in artificial intelligence at present, and are changing the application paradigm of artificial intelligence. Due to the huge number of model parameters and the huge demand for computing power and data, the training and inference of large models have become a bottleneck. Model compression can greatly reduce the computational complexity and resource consumption of network models, making them better applied and deployed on devices with limited computing resources, such as mobile phones, robots, and the Internet of Things. This report first analyzes the current challenges encountered in the calculation of large models, then introduces the latest research results of the team in terms of model quantization compression, sparse pruning, accelerator architecture design, etc., and finally shares the future development trend of this field.
Science Technology
Topic: eDRAM-based cryogenic in-memory computing accelerators
Speaker: Ha Yajun
He is currently a professor and doctoral supervisor at the School of Information Science and Technology of ShanghaiTech University, and is also the Editor-in-Chief of IEEE Transactions on Circuits and Systems II: Express Briefs (2022-2023), and the "Foreign Senior Scholar" and "Foreign Senior Scholar" of the National Natural Science Foundation of China. He is the winner of the "Key International (Regional) Cooperative Research" Fund, the director of the Shanghai Engineering Research Center for Energy Efficiency and Customized Artificial Intelligence IC, and the director of the Center for Post-Moore Devices and Integrated Systems of ShanghaiTech University. He has been engaged in the research of high-efficiency circuits and systems for a long time, focusing on FPGAs, low-power digital integrated circuits and intelligent hardware, and has published more than 150 papers in relevant high-level academic journals and conferences, and has won several best paper awards.
Summary of the report
The development of related applications such as quantum computing requires energy-efficient intelligent computing at extremely low temperatures. In-memory computing (IMC) is a promising solution for achieving greater energy efficiency in low-temperature data-intensive areas. However, existing IMC designs cannot take full advantage of the characteristics of cryogenic devices, such as ultra-low leakage current, to optimize the overall circuit at low temperatures. This lecture will introduce an eDRAM-based cryogenic IMC accelerator (eCIMC) that supports energy-efficient operations in different modes, including conventional memory operations, Boolean operations, and convolution operations.
Science Technology
Topic: New Opportunities for Integrated Storage-Computing Chips in the Era of Large Models
Speaker: Sun Tang
He has a bachelor's degree and a master's degree from Tongji University, and a doctoral student at Fudan University. Chief Scientist of Starblaze Technology, Head of the Special Project of High-tech Industry and Scientific and Technological Innovation in Lingang New Area of China (Shanghai) Free Trade Zone: R&D and Application of AI Chips with Integrated Storage and Computing. With 15 years of working experience in international GPU, image, data storage chip and other companies, he has undertaken the chip AI system architecture work in Starblaze, focusing on the AI chip hyper-converged architecture and distributed heterogeneous resource scheduling system. This paper conducts in-depth analysis and architecture deduction on the evolution direction of data embedding structure, vector similarity, KV index, storage and computing integrated stream acceleration required for AI model applications. 16 invention patents for hardware accelerators of storage and computing architecture have been authorized.
Summary of the report
The large model has led to various explorations to better match the chip architecture with practical applications. Based on the GPT algorithm framework, domestic chips are limited by the calculation efficiency ratio, process, and basic components, and need to be broken through.
The profitability basis of the existing large-scale model application can be summarized as follows: faster knowledge accumulation and condensation at marginal cost, faster cognitive query and comparison, and more refined prompting and fine-tuning method paradigm.
The information accumulation computing capability of the integrated storage and computing architecture is highly compatible with the business framework of large model applications, and it is extended from the narrow sense of integrated storage and computing AI algorithm acceleration to the generalized full-function integrated storage and computing acceleration of large model applications.
Science Technology
The theme of the report: "Wentian 1" brain-like computer achievement sharing
Speaker: Shang Delong
Ph.D., graduated from the University of Newcastle, UK, and is currently a researcher and doctoral supervisor of the Institute of Microelectronics, Chinese Academy of Sciences. He has published more than 120 academic papers in related academic journals and international academic conferences, and has won the first and third prizes of the Science and Technology Progress Award of the Chinese Academy of Sciences, the first prize of the Digital Category of the 10th PCB&MCM Technology Annual Leading Award of The Mentor Graphics in the United States, the Major Scientific and Technological Achievement Award during the National "Eighth Five-Year Plan" period, and the IET of the World Electrical Engineering Association 2005 Global Technology Innovation Award for Computer and Digital Technology.
He has participated in and led three national scientific and technological research projects, two major scientific and technological innovation 2030 projects, a number of projects of the Innovation Institute of the Chinese Academy of Sciences and horizontal projects of leading enterprises. The main achievements include the "Wentian I" brain-like computer, KSJ2300 minicomputer CPU system, KSJ2850 minicomputer CPU system, etc.
Summary of the report
The success of AI applications such as image recognition and speech recognition has catalyzed the rapid development of AI and accelerated the improvement of computing power. However, in the post-Moore era, it is becoming more and more impossible for the traditional semiconductor technology to improve computing power, so the urgent need for innovation is facing challenges.
Brain-like computing, which mimics the working mechanism of biological brains, realizes efficient information processing based on bionic spiking neurons, and has typical characteristics such as sparseness, spatiotemporal correlation, anti-noise, and event-driven.
In this context, we have developed the "Wentian 1" brain-like computer, which has opened up the whole industrial chain of chip architecture, software compilation, algorithm model, software and hardware system, and the machine can support 530 million neurons and 250 billion synaptic intelligence scale brain simulation;
Science Technology
Topic: Technological progress and typical applications of artificial general intelligence
Speaker: Wang Shijin
He is currently the Vice President of iFLYTEK Co., Ltd., the Executive Dean of the AI Research Institute, and the Deputy Director of the National Key Laboratory of Cognitive Intelligence. Mainly engaged in natural language processing, cognitive large models and other directions, undertake a number of national key research and development programs, won more than 30 international competition championships, published more than 100 academic papers in ACL, AAAI, KDD, SIGIR, IEEE Trans and other core journals at home and abroad and top conferences, has won the first prize of Anhui Provincial Science and Technology Progress Award, Wu Wenjun Artificial Intelligence Science and Technology Progress Award, China Association for Science and Technology Outstanding Youth Achievement Transformation Award. The Xunfei Xinghuo large model developed by the team has reached the leading level in China, and has been planned and applied in education, medical, office, industry and other industries.
Summary of the report
This report first analyzes the three stages of artificial intelligence, proposes that the current general artificial intelligence technology represented by cognitive large models has attracted widespread attention around the world, and then analyzes the technical characteristics, development trends and application value from cognitive large models to multimodal large models. Secondly, the report reported the achievements and R&D experience of iFLYTEK in the development of the Spark large model, introduced the exploration experience of the large model in various industries, and finally thought about the future development trend of the large model.
Science Technology
Topic: Artificial Intelligence Chips Help Digital and Intelligent Power Grids
Speaker: Chen Yuzhe
He graduated from the University of Southampton with a master's degree, and now works in Beijing Zhixin Microelectronics Technology Co., Ltd., mainly responsible for the research and development of artificial intelligence chips and related algorithms. It has mainly participated in major scientific and technological projects of the State Grid such as artificial intelligence edge computing chips, high-reliability and high-security intelligent vision chips. For application scenarios such as transmission line monitoring and intelligent substation inspection, an overall video surveillance solution based on artificial intelligence is built to provide source-grid-load-storage collaborative solutions for the construction of new power systems.
Summary of the report
As an important form of the new power grid, the digital and intelligent strong power grid is the core link and main starting point for promoting the construction of the new power system and the new energy system, which not only plays an important role in transforming the form of the power grid and enhancing the function of the power grid, but also relies on the intrinsic characteristics of deep embedding, extensive connection and high-frequency interaction of digital intelligence to play a non-negligible support and empowerment function in the whole industry and society.
Among them, artificial intelligence chips are used as the underlying computing power support, providing strong support for the digital and intelligent transformation of the power grid. In particular, it has become indispensable in solving the problems of intelligent operation and inspection based on machine generation and the core dispatching of power grids. Through an in-depth analysis of the applications and challenges of AI chips in helping to build a digital and intelligent power grid, this report aims to provide guidance and reference for the digital and intelligent upgrade of power grids, and promote the digital transformation of the energy industry.
Science Technology
Topic of the report: Discussion on the development path of high-computing power chips
Speaker: Yin Shouyi
Ph.D., Professor of Tsinghua University, Vice Dean of the School of Integrated Circuits, winner of the National Science Fund for Distinguished Young Scholars, and member of the National Expert Group on "New Generation Artificial Intelligence". His research interests include reconfigurable computing and artificial intelligence chip design.
He has published more than 200 academic papers, including ISSCC, VLSI, ISCA, MICRO, HPCA, DAC, IEEE JSSC, TPDS, TCSVT, TVLSI, TCAS-I/II and other academic conferences and authoritative journals in the field of integrated circuits and architecture. He has published 2 monographs on "Reconfigurable Computing" and "Artificial Intelligence Chip Design". He has won the second prize of the National Technological Invention Award, the first prize of the Technological Invention Award of the Chinese Institute of Electronics, the Gold Award of the China Invention Patent, the first prize of the Technological Invention Award of the Ministry of Education, the second prize of the Jiangxi Provincial Science and Technology Progress Award, the Outstanding Scientific and Technological Worker Award of the Chinese Institute of Electronics, and the Outstanding Scientific and Technological Paper Award in the field of electronic information in China. He is currently a member of the technical committee of MICRO and A-SSCC, an international conference in the field of integrated circuits, and an associate editor of the international journals "ACM Transactions on Reconfigurable Technology and Systems" and "Integration, the VLSI Journal".
Summary of the report
At present, the new generation of information technology, represented by artificial intelligence, is developing rapidly, and the demand for computing infrastructure is growing at a rate far exceeding Moore's Law. The development of high-computing power chips in mainland China is facing severe challenges. This report will analyze the development trends of international high-computing power chips, summarize the technical trends of high-computing power chips, and propose a development path based on the current industrial foundation of the mainland.
Science Technology
Panelist: Huang Haocheng
He has a bachelor's degree and a master's degree from the University of Electronic Science and Technology of China, and is currently the chief architect of Starblaze. He has more than 11 years of experience in chip development, teamwork, and project management at Huawei and Marvell, and is the chief architect at Starblaze, responsible for the overall chip architecture design, power consumption design, performance tuning, prototype verification platform, and IP selection. It has authorized 21 invention patents for chip architecture and 1 US patent.
Science Technology
Panelist: Mao Wendong
Assistant Professor of the "Hundred Talents Program" of Sun Yat-sen University. His main research interests are artificial intelligence system and integrated circuit design, intelligent visual computing, and optimization acceleration for Transformer/large models. He has published nearly 20 papers in mainstream IEEE journals and conferences, and applied for more than 10 patents. Many of these results have been published in journals in the fields of IEEE TNNLS and TMM as the first author. He has won the Best Paper Award of the IEEE ISVLSI International Academic Conference, and the Doctoral Dissertation Excellence Award of the Chinese Electronic Education Society. He has presided over or participated in a number of projects such as Shenzhen City, Jiangsu Province/Ministry of Science and Technology, National Natural Science Foundation of China, and Huawei and Kuaishou.
Science Technology
Panel Panelist: Maggie Wang
Assistant Professor of the "Hundred Talents Program" of Sun Yat-sen University. His main research interests are the design of integrated circuits and intelligent systems for artificial intelligence, focusing on the ideas of software and hardware collaboration, storage and computing integration, and applications such as large models, robots, and autonomous driving. He has published nearly 20 academic papers (including many first-class/important journals in the IEEE field), won the IEEE International Conference Best Paper Award twice, and applied for more than 20 patents. Participated in/guided students to win many national awards in important competitions in the field of integrated circuits (first prize twice). He has presided over or participated in many projects such as the CUHK Hundred Talents Program, the Ministry of Science and Technology/Jiangsu Province Key R&D, the National Natural Science Foundation of China, and the Military Science and Technology Commission, Huawei, etc.
07
Topics and guests of the roundtable dialogue
Conversation topics
Moderator: Wang Meiqi, Assistant Professor of the "Hundred Talents Program" of Sun Yat-sen University
Mao Wendong is an assistant professor of the "100 Talents Program" of Sun Yat-sen University
Topic of dialogue: 1. What are the development opportunities and challenges of cloud-edge-end intelligent chips in the era of large models?
2. How to design robot smart chips to better promote the development of general artificial intelligence
3. An agile path to the integration of modern chip design and AI algorithms
4. How to promote the construction of chip ecology from the perspectives of technological innovation, talent training and market demand
Panelist: Ha Yajun, Professor, ShanghaiTech University, Editor-in-Chief of TCAS-II
Cheng Jian is a researcher at the Institute of Automation, Chinese Academy of Sciences
Shang Delong is a researcher at the Institute of Microelectronics, Chinese Academy of Sciences
Huang Haocheng is the chief architect of Beijing Starblaze Technology Co., Ltd
Chen Yuzhe, Senior Engineer of AI R&D at Beijing Zhixin Microelectronics Technology Co., Ltd