Beijing, China, May 10, 2024 – The second Ascend AI Developer Summit was held at the Zhongguancun International Innovation Center in Beijing. With the theme of "Ascend Together, Dream the Future", the summit brought together industry leaders, technology experts, academic leaders, enterprises, universities, and individual developers to discuss cutting-edge AI technologies, share industry innovation cases, and excellent developer practices. At the summit, Ascend AI announced a comprehensive upgrade of its basic software to fully support native innovation and help developers pursue their dreams for the future.
Zhou Jun, President of Huawei ICT Marketing, said in his speech that more than 200 Ascend-based AI hardware products have entered various industries. More than 3 million developers and more than 1,600 industry partners have gathered in the Ascend AI ecosystem, jointly incubating more than 2,900 industry AI solutions. In terms of talent training, Ascend has worked closely with more than 300 universities and research institutes to jointly develop more than 20 textbooks and teaching aids, empowering more than 3,000 teachers and training more than 300,000 students, and building a solid talent foundation for the rapid development of China's AI industry. Ascend is committed to building a more open, efficient, and easy-to-use AI-native development technology system and ecosystem, enabling partners and developers to achieve business success and realize their dreams.
Zhou Jun, President of Huawei ICT Marketing
Gao Wen, Academician of the Chinese Academy of Engineering and Director of Pengcheng Laboratory, pointed out in his speech that Ascend's continuous innovation in cutting-edge technologies in the field of artificial intelligence, long-term planning and investment in native innovation, and extensive cooperation and layout in talent training have injected momentum into native innovation in the era of large models. Native innovation is related to the sustainable development of the field of artificial intelligence. It is necessary for every one of our developers to work together to open a new chapter in the era of rapid development of artificial intelligence and contribute our wisdom and strength to the progress of human society.
Gao Wen, academician of the Chinese Academy of Engineering and director of Pengcheng Laboratory
Developers are the backbone of the Ascend industry, and Zhang Dixuan, President of Huawei's Ascend Computing Business Group, said that Ascend is committed to focusing on basic software and hardware, building a solid AI computing foundation, and growing together with developers.
For operator innovation, the heterogeneous computing architecture CANN has been further opened and the Ascend C programming language has been fully upgraded, enabling developers to quickly build differentiated and competitive operators. For model innovation, the new version of MindSpore 2.3 supports the unification of dynamic and static, taking into account development efficiency and execution performance, and supporting efficient native model development. At the same time, we continue to maintain close cooperation with industry frameworks such as PyTorch, PaddlePaddle, and TensorFlow to improve the model development experience. For application innovation, the Ascend inference engine and scenario-based SDKs are provided to support developers to quickly deploy inference applications and accelerate the innovation and implementation of large models. For efficient cluster O&M, the CCAE cluster autonomous engine is provided to support capabilities such as system physical examination, fault detection, and cross-border location, improving O&M efficiency.
Ascend firmly focuses on "native" and drives ecological prosperity with the two-wheel drive of "technology + business". For the technology and business ecosystems, Ascend has released and upgraded a series of ecosystem plans to motivate and enable developers to build operators, models, and applications natively based on Ascend through diversified incentives, accelerating the cultivation of native talents. At the same time, it empowers partners to create more competitive hardware products and solutions to support their business success.
Zhang Dixuan, President of Huawei's Ascend Computing Business Unit
At the conference, a number of outstanding developers from enterprises and universities shared their innovative practices based on Ascend.
Zhang Peng, CEO of Zhipu AI, shared the theme of "Zhixing Voyage, Working Together to Create - Accelerating the New Era of AGI", and said that Zhipu AI is working with Ascend to achieve rich large-scale model practices, and the training and inference performance is rapidly improving.
Zhang Peng, CEO of Zhipu AI
Zeng Guoyang, CTO of Facewall Intelligence, shared the theme of "The Road to Agi Leap, a New Journey of Efficient Large Models", and pointed out that Facewall Intelligence continues to lead the "efficient large model route", combined with the basic software and hardware of Ascend AI, has done a lot of cutting-edge exploration and development, and proposed adaptive algorithms in terms of fine-tuning, which has significantly improved the efficiency compared with traditional methods.
Zeng Guoyang, CTO of Face Wall Intelligence
Wang Bibo, Senior Director of JD Cloud, delivered the theme sharing of "JD Technology's Ascend Practice", in which he said that JD Technology has full-stack AI capabilities, including intelligent computing infrastructure, AI computing power virtualization and hybrid scheduling, AI development and application platform, all of which have in-depth cooperation with Ascend, and have been successfully implemented in the fields of code assistance and operation and maintenance models.
Wang Bibo, Senior Director of JD Cloud
The Ascend AI ecosystem is booming, with many outstanding developers emerging. At the meeting, Li Ying, the gold medal winner of the developer kit track of the Ascend AI Innovation Competition 2023 National Finals, came to the scene and shared his developer growth story.
Li Ying, winner of the Gold Award in the Ascend AI Innovation Competition 2023 Developer Kit Track
The summit also held the release ceremony of the Ascend Large Model Native Development Results and the Ascend AI Innovation Competition 2024 Launching Ceremony.
Ascend Large Model Native Development Results Release Ceremony
Ascend AI Innovation Competition 2024 Kick-off Ceremony
At a time when AI technology is rapidly evolving and AI will reshape thousands of industries, Ascend AI is firmly focused on "native development" to provide developers with easy-to-use development tools and platforms to grow together. We look forward to more developers joining us to build a sustainable and resilient AI ecosystem.