022 WAIC World Artificial Intelligence Conference has recently ended in Shanghai.
HTML On September 3rd, at the WAIC 2022 · AI Developer Forum hosted by the machine, 2021 Turing Award winner represented the most influential academic leaders, technical experts and corporate executives, the content of the speech, the content of the speech content Including high -performance computing, multi -mode interaction, text generation research and application, RPA, brain -like computing, etc., including the cutting -edge topics. The theme of this conference is based on the theme of "AI Developers' True attention", and focuses on showing the advanced technological achievements and latest practical application progress of the artificial intelligence field in the year.
In addition to these wonderful keynote speeches, the WAIC · AI Developer Forum also revealed this year's Waic Yunfan Award winner to host the WAIC Hacker Song Awards Ceremony. In addition, the 2022 SODA Open Data Innovation Application Competition was officially launched.
is a wonderful review of the content of the Waic 2022 · AI developer forum.
021 Guting Award winner, Jack Dongarra, director of the Innovation Calculation Laboratory of the University of Tennessee, delivered a keynote speech. dfuirements ".
JACK DONGARRA mainly introduces the development of high -performance calculation in recent years, and high -performance computing and artificial intelligence integration faces new challenges and opportunities. Jack Dongarra believes that HPC & AI not only supplements each other in the specific application field, but also has a lot of connections in the fields of basic development methods, software and software infrastructure and AI hardware architecture. The future HPC system will be customized. After Moore's Law , there is still a lot of space to improve the ultimate performance of the HPC system through algorithm, software and hardware architecture.
Academician, academician of the American Academy of Engineering, James Demmel, a special professor at the Department of Computer Science and Mathematics at the University of California, delivered a keynote speech "Colossal-AI: A UNIFIED Deep Learning System for Big Big Model Era ".
, as the chief strategic officer of Laochen Technology, James Demmel mainly introduces the universal deep learning system Colossal-AI for the era of large models. In recent years, with the rise of deep learning and sweeping major performance lists, the size of the cutting -edge AI model has increased by 10,000 times in just a few years, far exceeding the slowly increased hardware. Colossal-AI achieves high-efficiency and rapidly deploying AI model training and inference through high-efficiency multi-dimensional merger, heterogeneous memory optimization, large-scale optimization library, adaptive task scheduling, etc., and promotes low-cost applications and landing on AI models.
Tsinghua University Huiyan Lecturer Professor, IEEE/CAAI FELLOW, Founder of Tuyuan Technology delivered a keynote speech ml6.
Zhou Bowen mainly introduced the research direction of the Tsinghua Cooperation Interactive Intelligence Research Center, including the progress of multi -mode AI, the trend of interactive AI, and recently led the team's trusted AI review thesis. At present, the progress and landing of artificial intelligence technology are particularly needed to trusted artificial intelligence in the aspects of robustness, interpretability, replication, fairness, privacy protection, responsibility and value alignment. Study in a single point. The process of trusted AI's landing needs to be ended and the full life cycle. It is not enough to pay attention to algorithms. At the same time, the landing of AI not only needs to consider the interaction of each dimension, but also requires the academic, industry and regulatory authorities to work together. The cross -value dimension of AI and the entire process also brings a lot of blank research opportunities.
Nine Chapters Yunji DataCANVAS Vice President and D-Lab Director Yang Jian delivered a keynote speech "Ylearn: Causal Learning, from predicted decision-making" .
Yang Jian mainly introduced machine learning 's dilemma in reality, from predictive decision -making, causal learning is the foundation of automatic decision -making and Ylearn causal learning algorithm tool package.Yang Jian said that the current demand for "intelligent decision -making" is becoming increasingly strong. Machine learning has problems with poor generalization ability, lack of interpretation, and insufficient decision -making support. The key to corporate needs and make up for the shortcomings of machine learning. Chapter Nine Cloud Polar DataCANVAS's self -developed one -stop processing causal learning complete process of open source algorithm tool package Ylearn is the first to solve the five key issues of "causal discovery, cause and effect identification, causal effect assessment, anti -factual inference, strategy learning" in causality learning, " It has the characteristics of one -stop, new and comprehensive, and widely used.
Vice President of Technology and the head of the smart driving business line Li Bo delivered a keynote speech "Robo Galaxy: Smart Driving Cloud Data Factory" .
Li Bo mainly introduced the company's business business and the development of Robo Galaxy. As the first ultra -luxury sports car brand to electrifying and intelligent transformation, Luther has a rich technology accumulation in the field of intelligent driving and has always been committed to providing users with high -performance intelligent driving products. The "Robo Galaxy" intelligent driving cloud data factory was born immediately. Li Bo used a vivid metaphorical metaphor to describe Robo Galaxy's positioning: "Early clothing was sewn by tailor stitches, but the POLO shirt now worn is made from the factory. In the future, smart driving development will not be. Then there is the scene of the handicraft workshop, but like the factory assembly line, this assembly line is built on the cloud data factory of Robo Galaxy.
Lanzhou Technology founder and CEO, China Computer Society CCF Vice Chairman, Innovation Workshop Chief Scientist Zhou Ming delivered a keynote speech. Application ".
Zhou Ming mainly introduced it from four parts: controllable text generating background, controllable text generating research progress, Lanzhou -controllable text application practice, summary and prospects. Speaking of the research progress of controllable text generation, Zhou Ming summarized three technical points: controllable text to generate neural network model, controllable text generation model training and decoding, improvement of the facts of the generation text. In addition, Zhou Ming also introduced the Lanzhou -controlled text generation platform and introduced it from the perspective of the application, such as the content generation of keyword -based content, the content generation of the formwork of writing, and the renewal of modern text novels.
Professor at the Department of Computer Science at the University of Rochester, Academician Rogo delivered a keynote speech "The Covid-19 Pandemic: What Social Media and Machine Learning Can Inform us at scale and in real-time ".
How to use artificial intelligence and social media big data to obtain information in real time in order to adopt countermeasures to resist the various problems brought by new crown , which is a question that Rogo is more concerned. Rogo believes that during the epidemic, many people spend time on social media, and the corresponding data will increase. These data can help researchers make auxiliary decisions. Rogo mainly introduced from four aspects: the impact of vaccine problems, epidemic conditions on human mood, the depth of the epidemic, and the epidemic increased the phenomenon of social polarization.
Real intelligent founder and CEO Sun Linjun delivered a keynote speech "Digital Employees, AI Applications and Land in the RPA" .
Sun Linjun believes that the prospects of digital employees are worth looking forward to and optimistic. With the future landing of artificial intelligence, it will gradually penetrate into thousands of industries. And transform this industry. He mentioned that RPA is a base that executes a regular process, but only by integrating AI technology and AI products can it become real digital employees.Therefore, it is really intelligent with "AI × RPA" as the core technology engine. It is committed to promoting the development of RPA to the IPA, creating a four -in -one product system and AI product matrix of designer, robot, controller, and a real cloud brain, and constructing various various various various various various various various, constructing various various various various various various various various various. Sample practical digital employees, create a super automation platform with digital employees as the base, simulate human operation to complete a large number of repeated rules, so that people can do a higher value and happier job. Platform, this is the future trend, and it is an irreversible trend.
Recalling the founder and CEO of Recalling Hai Hara, Ren Hualong, delivered a keynote speech "Development and Key Technology of Calculation of Brain -like Calculation" .
Ren Hualong introduced the characteristics of brain intelligence and comparison with traditional technologies. There are still many shortcomings in artificial intelligence, such as limited generalization capacity, strong data sample dependence, and cannot be flexibly recognized. Two key features of brain -like computing are technical characteristics, and the other is cognitive function. In terms of technical characteristics, the key working principles of the biological brain are used to rely on mathematical optimization for training. In addition, Ren Hualong also introduced the Ocean -class brain computing platform independently developed by Yi's original consciousness, which provides a large number of basic models, supports flexible modeling methods and free expansion, and covers research and development, application deployment and other needs.
Unicom (Shanghai) Industrial Internet Co., Ltd. AI technical director Hu Chao delivered a keynote speech "From Supervision Learning to Self -Learning: Non -Supervision Target -Level Nodes" .
has developed rapidly in deep learning in recent years. In the speech, Hu Chao mainly introduced their application and development in the field & MIM (Masked Image Modeling) & GAN, image repair and contrast learning, flow models and diffusion models, and unsupervised learning scenarios. When introducing the application scenario, Hu Chao talked about the application of unsupervised learning in the field of smart cities. Based on massive label data, different from the data bottlenecks of strong supervision, unsupervised learning is booming on this basis, and the scale is continuously expanding. Just like the wave of the Yuan universe, unsupervised universal For those.
Next is WAIC · Yunfan Prize winner forum . The guest of dialogue includes Tsinghua University postdoctoral researcher Dai Guohao and Harbin University of Technology Associate Professor Feng Ruyi, Shanghai Artificial Intelligence Lab youth scientist Xu Jing Xu Jing Jing, Zhejiang University Public Health Yang Jie, a researcher at the college hundred people plan & doctoral supervisor, Yang Yaodong, assistant professor at the Peking University Artificial Intelligence Research Institute & Beijing General Artificial Intelligence Research Institute. At the round table forum, five guests shared some of their experience summary.
was finally from the Assistant Professor of Computer Science, the Department of Computer Science, the University of North Carolina University, delivee "Building Machine Learning Models Like Open-Sourc E Software ".
How can we make the collaboration and sustainable development of machine learning models? Colin Raffel believes that the contributor needs to be able to pass the Patches to the model cheaply. The defender needs to merge the update from different contributors. Users who lack resources can train and run large models. Realize new functions.
is the operator of the developer ecology of the WAIC. The heart of the machine invites the Microsoft Asia Research Institute visual computing group Hu Han, University Computing Center Director Fan Chun, , Tsinghua AMINER engineer Huang Meiling, Shanghai Artificial Intelligence Laboratory Young Researcher Chen Li, Huawei Lito CANN Chief Engineer Jin Huijia, Alibabada Institute of Intelligent Computing Laboratory Algorithm, Lin Junzheng, Bybal Beat Unicorn Algorithm Team Member Yan Bin and other 7 front -line researchers jointly explore the difficulty of "AI Development Last N Kile", and based on the new version of SOTA! The model resource stations and partners such as Peking University High Performance Computing Center, Tsinghua AMiner and Huawei Shengteng Cannn have briefly reported the results of the co -construction of the community.
After a day of high -intensity knowledge baptism, I believe everyone has endless aftertastes. Please continue to pay attention to the heart of the machine. The Waic · AI Developer Forum Speech guest shared the content of the content will be released in recent days.
