Scan the QR code:
(Scan the QR code to register)
The large model is triggering profound social changes and reconstructing the ecology of the information society, and the multi-dimensional and fair large model is an important factor to promote the benign development of the intelligent society.
In the development of artificial intelligence, "gender bias" is not a new word. There is a view that the gender bias in the training method and training data may be learned by the large model and reflected in its output, affecting the fairness and accuracy of decision-making. In the long run, the problem can have many undesirable effects and even trigger a "butterfly effect", such as exacerbating inequality of opportunity, offending women, or reinforcing gender stereotypes in human-computer interactions. How to achieve the governance of gender bias in large models requires dialogue between policy, industry and research.
The forum invited experts related to gender difference research in the field of AIGC to conduct in-depth discussions on the problem of gender bias in large models and their coping strategies from the aspects of large model development, large model alignment, emotional cognitive exploration and gender shaping in large models, mainly discussing the sources and effects of bias, and proposing solutions, with the aim of promoting the fairness, transparency and responsibility of AI technology and promoting the healthy development of intelligent society.
Forum arrangement
| Introduction | topic | Keynote speaker | unit |
| 1 | The Development of Large Models: Adapting to the Past or Changing the Future? | Gao Yang | Beijing Institute of Technology |
| 2 | Large model alignment | Qiu Xipeng | Fudan University |
| 3 | Affective Cognitive Exploration and Gender Shaping in Large Models | Yao Hongxun | Harbin Institute of Technology |
| Speculation 1. Is there a gender bias in AIGC, what are the root causes of it, and does it trigger a "butterfly effect"? 2. How can technology be used to "empower women" and de-gender AIGC? 3. What should be, what can and what "she power" should be, can do, and what is difficult to do | Xu Xiaomin | Tencent Cloud, Inc | |
| Yao Changjiang | Qingdao Dongting Intelligent Technology Co., Ltd | ||
| Faye Wong | China Technology and Business University |
Executive Chairman
Zhao Xiang
Professor of Yunnan Normal University
He is a senior member of CCF, the chairman of CCF YOCSEF Kunming for 2023-2024, the rotating chairman of YOCSEF Yifang Salon for 2024-2025, and the director of the Youth Artificial Intelligence Education Branch of the Chinese Association of Higher Education. His areas of expertise include learning analytics and smart education, pattern recognition and intelligent systems. He has presided over 1 general project of the Humanities and Social Sciences Planning Fund of the Ministry of Education, 4 provincial projects, and participated in more than 10 projects such as 1 National Social Science Fund. The first author has published more than 30 papers, authored 1 academic monograph, co-edited 6 books, edited 2 online learning textbooks (with a volume of more than 100,000 copies), and co-edited 3 books. 15 software copyrights have been registered. Participated in innovation and entrepreneurship competitions at or above the provincial level and won 1 national bronze award, 2 provincial gold awards, 6 silver awards and 4 bronze awards.
Co-Executive Chairman
Li Yang
Associate Professor of Northeast Forestry University and Deputy Dean of the School of Computer and Control Engineering
CCF YOCSEF Harbin 2024-2025 President. His research interests include natural language processing, artificial intelligence and bioinformatics. He has presided over 2 projects of the National Natural Science Foundation of China, the Outstanding Youth Project of the Natural Science Foundation of Heilongjiang Province, and many provincial and ministerial projects. He has published more than 20 papers in top international journals and conferences. He won the "Top Ten Graduate Tutors" award of Northeast Forestry University.
Forum Speaker
Yao Hongxun
Dean Professor of Computing, Harbin Institute of Technology
Heilongjiang Provincial Government Special Allowance Expert, Ministry of Education "New Century Outstanding Talent", the 7th and 8th Executive Director of the Chinese Society of Image and Graphics, currently the Director of the Affective Computing and Understanding Professional Committee of the Chinese Society of Image and Graphics, once served as the co-chairman of the 1st International Conference on Affective Computing ACII2005, and the International Conference on Internet Multimedia Computing and Services (ACM). ICIMCS2010 the chairman of the conference, the chairman of the 2021 World Artificial Intelligence Conference Harbin Branch, the co-chair of the 2023 Young Scientists Conference of the Chinese Society of Image and Graphics, and the chairman of the 1st Emotional Intelligence Conference in 2023. His main research areas are computer vision intelligence, multimedia data analysis and understanding, affective computing, etc. He has published more than 300 academic papers in top international conferences such as ICCV, CVPR, ACMMM and international journals with high impact factors such as IJCV, TPAMI, TIP, TMM, etc., with an H-index of >50, Google citations 10000+, into the global artificial intelligence TOP 2000 list of scholars, Chinese artificial intelligence 100 list of people. He has obtained more than 30 national invention patents and published 6 textbooks. He has presided over or been responsible for the completion of 12 key and general projects of the National Natural Science Foundation of China, 1 major national science and technology project of the intelligence of a new generation of workers, and the completion of a number of national 863 and 973 projects and international cooperation projects. He has won 4 national, provincial and ministerial natural science awards, and won the China Computer Science Outstanding Teacher Award (2021). It has cultivated and grown into 1 "National High-level Talent", 4 "National Young Talents", 6 "Youbo" and "Youshu" students, and more than 10 provincial outstanding graduates.
Title: Emotional Cognitive Exploration and Gender Shaping in Large Models
Summary:
This report takes the image sentiment analysis task as an example to explore the transformation process of artificial intelligence from semantic perception to emotional cognition, emphasizes the importance of images as a carrier of rich emotional information, and introduces the differences between general image sentiment analysis and traditional image perception research. On this basis, two main challenges faced by image sentiment analysis are pointed out: affective gap and subjectivity. The affective gap is reflected in the inconsistency between the image features and the user's expected emotional state, while subjectivity involves the phenomenon that different viewers have different emotional responses to the same image. This paper introduces the research directions of feature extraction and label noise processing, and looks forward to the development direction of image sentiment analysis, emphasizes the importance of multimodal large models, and puts forward the potential impacts of sentiment information inference and feature fusion of multimodal large models.
Qiu Xipeng
Professor, School of Computer Science, Fudan University
CCF YOCSEF Shanghai 2024-2025 Chairman, the main research direction is the basic technology and basic model of natural language processing, published more than 100 CCF-A/B papers, more than 19,000 citations, selected into Elsevier's "China Highly Cited Scholars" and "Lifetime Scientific Influence Ranking (Top 2% of the World's Top Scientists)", has been awarded the China Association for Science and Technology Young Talent Lifting Project, National Excellent Youth and other projects, won the first prize of Qian Weichang Chinese Information Processing Science and Technology Award (the first completer), was selected into the "Outstanding Teacher Award Program for Computer Science in Colleges and Universities" of the Ministry of Education, and won the first prize of the Teaching Achievement Award of Shanghai Computer Federation twice; the MOSS developed by him has become one of the most influential open source large language models in China. The book "Neural Networks and Deep Learning" has been used as a textbook by hundreds of colleges and universities.
Title: Large Model Alignment
Summary:
In the current era of rapid technological development, large models are playing an increasingly important role in our daily lives. However, these technological systems often exhibit gender biases, which not only reflect bias in training data, but can also exacerbate gender inequalities. This report focuses on how to reduce and eliminate these biases by increasing the representation and participation of women in AI and gender-aligning large models.
Gao Yang
Associate Professor, School of Computer Science, Beijing Institute of Technology
He has been engaged in generative language modeling for more than ten years, published more than 60 high-level papers, presided over 2 projects of the National Natural Science Foundation of China, 1 project of patent achievement transformation, and participated in 7 national, provincial and ministerial key/major research plans. The course "Large Model: Algorithm and Practice" was selected as the core course of the Ministry of Education for master's and doctoral degrees in key areas of artificial intelligence. He presided over the construction of the MindLLM and domestic ecological model trained from scratch, and open-sourced the "end-to-side" dialogue model, which was downloaded more than 4,000 times in the first month of its launch, with a total of more than 10,000 downloads, and launched the intelligent government service of retrieving and enhancing the large model. The results won the first prize of scientific progress of the Institute of Electronics and the second prize of national defense scientific progress. He has served as an editorial board member of Web Intelligence and other journals, the chairman of EMNLP, CCL and other conference fields, and a member of the program committee of high-level international conferences for many times.
Title: The Development of Large Models: Adapting to the Past or Changing the Future?
Summary:
The gender bias presented by large models is largely caused by human past data and cognition, and the existing model method adapts to this data distribution. In the face of social discrimination and gender bias, what kind of energy can AI technology exert and what changes can be made? This report will discuss the above issues and discuss the "she power" of AI empowerment from the aspects of data construction, model training methods, and evaluation systems. Next, we look forward to the advantages of women's power in the AI era and the possible benefits it can bring to society as a whole.
Speculative guest
Xu Xiaomin
Regional General Manager of Yungui of Tencent Cloud
Vice Chairman of the 3rd (2023~2025) CCF Kunming Branch, member of the Big Data Professional Committee. Graduated from Tsinghua Dahua with a degree in environmental science and engineering, he joined Tencent in December 2022 and served as the general manager of Alibaba Cloud Yunnan before joining Tencent. With more than ten years of experience in the IT industry, he pays attention to the cutting-edge trends of digital technology, and is committed to using leading technologies such as cloud computing, big data, artificial intelligence and blockchain to promote the digital transformation and industrial innovation of the industry, and help the integration and development of the digital economy and the traditional real economy.
Yao Changjiang
Product Director of Qingdao Dongting Intelligent Technology Co., Ltd
Obtained the International Product Manager Qualification Certification, the International Project Manager Qualification Certification. He has nearly ten years of work experience in the field of artificial intelligence, and has rich experience in data analysis, NLP, large model application, and intelligent transformation of enterprises. He has been responsible for a number of enterprise information transformation projects, has unique insights into the fine-tuning and application of large models, and has planned and designed a number of practical application systems of large models.
Faye Wong
CCF YOCSEF One Party Salon Gender Consultant
Ph.D. in Management, graduated from Adam Smith Business School, University of Glasgow. He is currently a postdoctoral fellow at the School of Economics, Beijing Technology and Business University, and his main research areas are digital economy, AI development and labor.