steelheadfish| The Shanghai Station of the 2024 Global Machine Learning Technology Conference concluded successfully to present a new era of AGI transformation
Source: CSDN
From April 25 to 26, the ML-Summit 2024 Global Machine Learning Technology Conference, which attracted much attention from the industry, opened at Hyatt Hotel, the Global Port of Shanghai. The conference is co-sponsored by CSDN, a Chinese developer network, and Boolan, a high-end IT consulting and education platform.
The gathering of this grand meetingSteelheadfishExperts and scholars from the field of large models, such as Microsoft, Ali, Xiaomi, Baidu, Ant Group, Zhisu, 010 things, Zhiyuan, face Wall Intelligence, Chinese Academy of Sciences, Peking University, Fudan, Shanghai Jiaotong University, Southern University of Science and Technology, etc., cover all fields of industry-university-research, and jointly discuss the cutting-edge trends of large-scale model technology and its application in various industries. The conference attracted more than 1000 AI technical elites from various industries to participate, focusing on 12 major topics of technology stack and industry landing practice at all levels of the large model.
Li Jianzhong, senior vice president of CSDN and chief technology expert of Boolan, discussed in detail the three paradigm shifts brought by large models to the computing industry: computing paradigm, development paradigm, interaction paradigm, and their far-reaching impact on infrastructure, development process and application form, including infrastructure, development process and application form, in the keynote speech "AGI 2024 technological Innovation Research and thinking: large Model driven Computing Industry Paradigm shift". Dr. Daniel Povey, father of Kaldi and chief voice scientist of Xiaomi Group, shared how AI viewed the current large model technology throughout its history. Zhang Qi, a professor at Fudan University and head of Mu Sida Model, focused on analyzing the reasoning ability of the big model in his keynote speech "is the big language model the only way to achieve AGI", saying "forget AGI, emerge, align, inspire …" And other online celebrity terms, call for "re-departure" from the perspective of statistical machine learning. In addition, the speech of Dr. Qiao Mu, co-founder of Shuimu molecule, focuses on multimodal biomedical models and their industrial applications, indicating the deep integration of general large models and industrial applications. During the summit dialogue session, experts conducted in-depth discussions around "the evolution of the big model and the road to AI innovation", providing participants with valuable industry insights and answers. The successful holding of the conference not only shows the latest research results and technical applications, but also provides new ideas and framework for the future development direction of AI technology.
The 12 major themes were held in parallel, and the guests were enthusiastic.
The 12 major topics of large language model technology evolution, multimodal large model frontier, large model engineering and architecture, large model application development practice, GenAI product innovation and exploration, large model application framework and tools, AI intelligent agent, code big model and software change, ML/LLM Ops large model operation and maintenance, AI Infra large model infrastructure, computing infrastructure and performance optimization and large model + industry landing practice were held simultaneously in two days.
The rise and development of large-scale models are completely changing the face of the software industry, bringing developers more powerful tools and technical support, enabling them to build smarter and more complex software systems. In the special session of large model enabling software development and change, we specially invited Microsoft Chief Applied Scientist Manager Huang Yufan, Alibaba Tongyi Lab Senior algorithm Director Li Yongbin, Baidu Comate architect Zhang Lili, Zhisu AI-CodeGeeX algorithm engineer Zhang Xinpeng and many other industry leaders to discuss the application prospect and innovative practice of large model in the field of software development. They share their valuable experience and cutting-edge technology in the development of large model enabling software, bringing us a unique technological feast.
Multimodal technologies such as large model video generation are leading us to general artificial intelligence (AGI) by simulating everything. Both Vincent audio model and Vincent video model are refreshing industry cognition again and again. In this conference, Pan Xin, co-founder of 010 things, deeply discussed the application of multimodal technology and pointed out the importance of large-scale multimodal data. At the same time, Liu Shujie, chief researcher of Microsoft Asia, shared the help of the pre-training model in speech recognition and generation. In addition, the new method of codec brings innovative breakthroughs for speech processing and indicates a wide range of application prospects. Professor Yuan Ji of Peking University introduced Open-Sora Plan in cooperation with Rabbit Exhibition AIGC Lab, an open source project designed to promote the development of video generation models and encourage the participation of the open source community. At the same time, Ni Bingbing, professor of Shanghai Jiaotong University, Zhang Junlin, head of new technology research and development of Sina Weibo, and Li Tong, vice president of Luchen Technology, had a wonderful sharing on the native multimodal and multimodal architecture and specific applications.
AI Infra is a solid foundation to support the development of artificial intelligence. It is like a highway extending in all directions, providing a stable and efficient environment for the operation of large models. At the site of the AI Infra large model infrastructure project, Liu Xuan, vice president and technical partner of Yuanrong Qixing, revealed the application cases and technical breakthroughs of end-to-end models in intelligent driving; Xing long, technical director of MiniMax, shared the challenges and practices of large models in AI Infra, as well as coping strategies; Wang Wenyu, co-founder and CTO of PPIO, discussed how to achieve "hard and soft" and innovative breakthroughs in computing infrastructure and reasoning optimization. Yang Zhilei, deputy general manager of dawning Storage products Division of China Science and Technology Co., Ltd., showed how dawning storage self-developed XDS technology can improve the training efficiency of AI model and accelerate the landing of industry AI application.
Wang Qinlong, an ant group AI Infra technology expert, introduced DLRover training fault self-healing technology to greatly improve the computational efficiency of large-scale AI training, present a feast of cutting-edge technology for participants, and lead the rise of new quality productivity.
For more topics to share, please follow: 2024 Global Machine Learning Technology Conference | official website
Top 10 new products and solutions in AI area are unveiled, attracting attendees to stop by
In the exhibition hall outside the main venue, the booths such as Shenzhen Yuanrong Technology, dawning Storage, Jinmao Green Construction, PPIO, Alipay Mini Program Cloud, Xiyun Computing, Beijing Rong Digital Technology, Dell & Invida, Asynchronous Community, Beijing Chuxin Open Source and other booths attracted the attention of more participants, and the scene was bustling.
Shenzhen Yuanrong Technology has demonstrated its end-to-end model smart driving solution, which makes driving more possible, and the "super AI driver" is no longer a dream. Dawning Storage shows how its high-performance AI data infrastructure helps AI enterprises unleash their computing potential and improve training efficiency. Jin Maolujian, as a full-stack intelligent computing service provider, demonstrated its service advantages in training, reasoning, rendering computing and other areas; PPIO showed how OPEX Computing Cloud can help AI model enterprises reduce costs and speed up the implementation of AI applications; Alipay Mini Program Cloud on-site demonstrated its leading server-side solutions based on cloud native high availability, cost reduction, operation and maintenance-free, and high concurrent business support. Dell teamed up with Nvidia to show off its Precision workstation, deeply digging the "new momentum" behind "intelligence"; NWN unveiled Danmo Intelligence platform, an industry-oriented integrated software and hardware computing infrastructure for industry applications for the first time, and demonstrated how its commercial Lustre parallel file system provides high-speed storage systems for the AI platform.
At the same time, the asynchronous community also brought the newly published books on artificial intelligence technology to the participants and the on-site recruitment of the 2024 China Internet Development Innovation and Investment Competition brought by Beijing's original Open Source, which made the event splendid. The experience of the participants was more colorful.
The ML-Summit 2024 Global Machine Learning Technology Conference came to a successful conclusion. Through wonderful sharing and in-depth discussions, participants discussed in depth the frontiers and applications of large model technology, and explored the technical challenges and solutions of large model engineering and architecture, multimodal and product innovation. This meeting not only shows the latest progress in the field of artificial intelligence, but also provides valuable communication and learning opportunities for people in the industry. We look forward to further promoting the innovation and development of machine learning, artificial intelligence, large model, generative AI and other technologies in the future research and practice, and jointly create a new chapter in the intelligent era. Finally, I would like to thank all the participants for their support and contribution, and look forward to seeing you again!
At the same time, the 2024 Global Software Development Technology Conference (http)Steelheadfish: / / sdcon.com.cn/) will be held at the Westin Hotel in Beijing from July 4 to 5. Please follow us.
