What's the hottest thing in tech this year? It is undoubtedly a new wave of artificial intelligence technology represented by large models and AIGC.
In this wave, there is a clear feeling: the technical concept of working in downstream tasks is dizzying and will be widely adopted in a very short time. In the case of Vincent Graph, Stable Diffusion, which has been around for less than a year, has taken diffusion models by storm, and Midjourney has completely exploded its commercial value. Similar concepts are not only models, but also some methods, techniques, such as thought chains, RLHF...
At the same time, some new technology trends are quietly brewing, including multi-modality, multi-task unification, and the rise of AI agents. Representative models of the former include ImageBind, which aligns all modalities with images, and Segment Anything Model (SAM), which splits all images with one model; The latter includes AutoGPT, which can automatically decompose tasks, and "virtual towns" composed of multiple agents created by Stanford and Google. Some time ago, Andrej Karpathy, former director of AI at Tesla and who returned to OpenAI at the beginning of this year, revealed at a developer event that OpenAI has also been very interested in the development of AI agents and has high expectations, and whenever a new AI agent paper appears, OpenAI will be very interested and seriously discussed.
For ordinary researchers and practitioners, being able to keep up with the application and evolution of these technical concepts and trends is an important prerequisite for advancing the work at hand, and it is also a way to find new inspiration. To this end, at 14:00 pm on July 25, Alimama Bojian will cooperate with the Hillhouse School of Artificial Intelligence of Chinese University to hold a major event of the Alimama & Idle Fish Technology Festival, inviting many scholars from the Hillhouse Institute of Artificial Intelligence of the National People's University, including Dou Zhicheng, Zhao Xin, Lu Zhiwu, Xu Jun, Qi Qi, Lin Yankai, Chen Xu, Li Chongxuan, Huang Wenbing, etc., to talk about the big model. Let's take a look at what are the technical points that deserve our special attention today, how the big model will develop in the future, and how it will affect digital intelligence business technology.
Event schedule
Guest introduction
Dou Zhicheng: Deputy Dean, Professor and Doctoral Supervisor of Hillhouse School of Artificial Intelligence of Chinese Minmin University, Project Manager of "Intelligent Information Retrieval and Mining" of Beijing KLCII Artificial Intelligence Research Institute, Deputy Secretary-General of Big Data Expert Committee of China Computer Society, Deputy Director of Information Retrieval Special Committee of China Chinese Information Society. He worked at Microsoft Research Asia from 2008 to 2014 and taught at Chinese University since 2014. His main research interests are intelligent information retrieval, natural language processing, and big data analysis. He has published more than 100 papers in internationally renowned academic conferences and journals (such as SIGIR, WWW, CIKM, WSDM, ACL, EMNLP, TKDE, etc.), presided over 3 projects of the National Natural Science Foundation of China, 2 key research and development projects of the Ministry of Science and Technology, and more than 10 enterprise cooperation projects. He has won the first prize of the Natural Science Award of the Ministry of Education, SIGIR 2013 Best Paper Nomination Award, AIRS 2012 Best Paper Award, CCIR 2021 Best Paper Award and other awards. He has served as the Program Committee Chair of SIGIR, a top conference in the field of information retrieval (2019 short paper), the Chairman of the AIRS Conference of the Asian Information Retrieval Conference (2016), the Chairman of the CCIR Program Committee of the National Conference on Information Retrieval (2020), and the President of the General Assembly (2023). He has served as a member of the (senior) program committee of several international academic conferences.
Zheng Bo: CTO of Alimama and Xianyu, Chief Scientist of Alimama and Executive Committee Member of CCF Computational Economics Professional Group, responsible for the overall technical work of Alimama Advertising Technology Division and Xianyu's algorithms, machine learning, and engineering architecture. Graduated from Tsinghua University's Department of Computer Science, he worked at Google for 11 years before joining Alibaba in 2017, leading Google's display advertising algorithm team and China map team. Research interests: His research interests include deep learning, display and search advertising algorithms, multimodal and engine optimization.
Zhao Xin: He is currently a professor at the Hillhouse School of Artificial Intelligence at Chinese Minmin University. He received his Ph.D. from Peking University in July 2014 and has been working at Chinese Minmin University since then. His research fields are information retrieval and natural language processing, especially basic technology and application research based on large language models, with a total of more than 100 papers published, more than 10,000 Google Scholar citations, and has led the development of open source tools such as Bole (recommendation system library RecBole) and Miaobi (text generation library TextBox). He won the 2020 Wu Wenjun Artificial Intelligence Outstanding Youth Award, ECIR 2021 Test of Time Award, RecSys 2022 Best student paper runner-up, CIKM 2022 Best resource paper runnerup, etc., and was selected as a young talent support project of China Association for Science and Technology and a young scientist of Beijing KLCII, CCF-IEEE CS Young Scientist.
Lu Zhiwu: Dr. Lu Zhiwu is a professor and doctoral supervisor at the Hillhouse School of Artificial Intelligence of Chinese Minmin University. In 2005, he graduated from the Department of Information Science, School of Mathematical Sciences, Peking University, with a Master of Science degree. He graduated from the Department of Computer Science of City University of Hong Kong in 2011 with a PhD degree. His main research interests include machine learning, computer vision, etc. Design the first publicly available Chinese universal graphic-text pre-training model Wenlan BriVL. He has published more than 90 academic papers as the main author, including more than 50 papers in international journals such as Nat Commun, TPAMI, IJCV and international conferences such as ICML, ICLR, NeurIPS, CVPR, ICCV. The students mentored won the 2021 CCF Youbo and 2021 Baidu Scholarships. He is a member of the CCF Bioinformatics Committee. Served as a (senior) program committee member of NeurIPS, ICML, ICLR, ICCV, CVPR, AAAI, IJCAI and other top international conferences.
Lin Yankai: He received his bachelor's and doctoral degrees from Tsinghua University in 2014 and 2019. After graduating with a Ph.D., he worked as a senior researcher at Tencent WeChat and joined Chinese Minmin University as an assistant professor in 2022. His main research interests are pre-trained models and natural language processing. He has published more than 40 papers in top international conferences on natural language processing and artificial intelligence such as ACL, EMNLP, NAACL, AAAI, IJCAI, NeurIPS, etc., and Google Scholar has been cited more than 9,000 times, with an H-index of 27. His knowledge-guided natural language processing research results are summarized as three representative works in "Structured Knowledge Representation Learning Methods" and won the first prize of natural science of the Ministry of Education, and the open-source toolkits OpenKE and OpenNRE have won more than 6400 stars on Github, the world's most influential open source platform, becoming the mainstream tools of knowledge-driven natural language processing in the world. He has served as a field chair for conferences such as EMNLP, ACL ARR, and others.
Li Chongxuan: Associate Assistant Professor and Doctoral Supervisor of Hillhouse School of Artificial Intelligence, Chinese Minmin University. His research interests are probabilistic machine learning. His representative works include: Triple-GAN, an optimal semi-supervised GAN method under consistency theory; The optimal inverse variance estimation of the diffusion probability model in the maximum likelihood sense Analytic-DPM. Li Chongxuan won the 2022 Outstanding Paper Award of ICLR, an important international conference in the field of machine learning, the first prize of the 2021 Wu Wenjun Artificial Intelligence Natural Science Award, the 2019 Outstanding Doctoral Dissertation of the Chinese Computer Society and the 2017 Microsoft Scholar. Li Chongxuan was selected as the 2021 Beijing Science and Technology Rising Star, the 2019 China Postdoctoral Innovation Talent Support Program, and presided over the National Natural Science Foundation of China.
Xu Jun: Professor (long-term associate professor) of Hillhouse School of Artificial Intelligence, Chinese Minmin University, Distinguished Scholar Distinguished Professor of Chinese Minmin University, and KLCII Scholar of Beijing KLCII Institute of Artificial Intelligence. He has worked at Microsoft Research Asia, Huawei Noah's Ark Lab (Hong Kong) and Institute of Computing Technology of the Chinese Academy of Sciences, and joined Chinese Min University in September 2018. His research fields include Internet search and recommendation models and systems, and he has published more than 100 papers, 2 monographs, and more than 10 authorized patents, and some of his research results have been included in information retrieval textbooks and applied to search and recommendation products of Microsoft and Huawei by European and American scholars. He won the ACM SIGIR 2019 Test of Time Award Honorable Mention, CIKM 2017 Best Full Paper Runner-up, CCIR 2022, AIRS 2010 and ICMLC 2005 Best Paper Award, and won the second prize of Beijing Natural Science Award (ranked 2nd). He presided over the National Key Research and Development of China and the National Natural Science Foundation of China.
Qi Qi: Tenured associate professor at the School of Artificial Intelligence of Hillhouse, Chinese Minmin University, doctoral supervisor, national overseas high-level young talent, secretary-general of the computational economics professional group of CCF China Computer Society. He graduated from Stanford University with a Ph.D. under the supervision of Professor Yinyu Ye. He was an assistant professor and doctoral supervisor at the Hong Kong University of Science and Technology. His research interests include algorithmic game theory, mechanism design, optimization and multi-agent systems. He has published more than 40 papers in world-class computer, artificial intelligence, management journals and conferences, including OR, MOR, GEB, TR-B and other famous journals and STOC, WINE, CCC, IJCAI, NEURIPS and other top computer conferences. He has presided over the National High-level Talent Programme and a number of research projects of the Hong Kong Science Foundation. He has served as a senior program committee member and co-chair of several international conferences in the field of artificial intelligence, Internet and gaming. At the same time, he has served as a long-term reviewer for more than 10 international first-class journals. The scientific research results also have strong application value, and the research and application results in Internet advertising have obtained two US patents.
Chen Xu: He graduated from Tsinghua University with a Ph.D. and joined Chinese University in 2020 as an associate assistant professor. His research interests include recommender systems, reinforcement learning, and causal inference. He has published more than 60 papers in famous international conferences / journals such as TheWebConf, AIJ, TKDE, SIGIR, WSDM, TOIS, etc., and Google Scholar has cited more than 4000 times. His research has promoted the development of explainable user behavior analysis in a certain sense, and his research results have won the CCF Class A Conference TheWebConf 2018 Outstanding Paper Nomination Award, CCF Class B Conference CIKM 2022 Best Resource Paper Runner Up Award, and the AIRS 2017 Best Paper Award of the famous Asian information retrieval conference. He has also won the second prize of CCF Natural Science Award (ranked second), ACM - Beijing Rising Star Award (Beijing Three), Beijing Outstanding Graduate, etc. The research results have been implemented in many enterprises, and the related achievements have won Huawei's "Innovation Pioneer" President Award. He has presided over / participated in a number of National Natural Science Foundation of China and enterprise cooperation projects.
Huang Wenbing: Assistant professor and doctoral supervisor at the Hillhouse School of Artificial Intelligence, Chinese Minmin University. He used to work as an assistant researcher at the Intelligent Industry Research Institute of Tsinghua University and a senior researcher at Tencent AI Lab. His research interests include geometric machine learning theory and its application in cross-domain tasks such as intelligent drug discovery, physical scene understanding and simulation, and agent perception and decision-making. Representative work includes: DropEdge, a method for training deep graph neural networks; AS-GCN, an efficient training method for graph neural networks for large-scale graphs; Multichannel isovariable attention network MEAN for antibody generation. He has published more than 40 papers in top conferences or journals in the field of artificial intelligence (NeurIPS, ICLR, TPAMI, etc.), with more than 6,000 Google Scholar citations and a single citation of more than 900 times. He has applied for more than 10 invention patents and authorized 5 patents. He has won the ICLR 2023 Outstanding Paper Nomination Award, Tencent Rhino Bird Special Research Excellence Award, NeurIPS 2022 Open Catalyst Competition Champion, IROS 2020 OCRTOC Robot Challenge Third Runner-up, NeurIPS Outstanding Reviewer and other awards.