At Huawei Connect 2021, Deng Taihua, vice president of Huawei and president of the computing product line, shared the latest progress of six digital ecosystems, and brought a series of heavy contents to the computing industry together with eco-partners: the new release of the openEuler operating system; 21 cities jointly light up the "artificial intelligence computing network"; Beiming Diversity Computing Convergence Architecture was first published; The "intelligent pedestal" production-education integration collaborative education base has been fully expanded.

　　Six digital ecosystems have developed rapidly.

　　By empowering partners, the six digital ecosystems represented by Kunpeng, Shengteng, Huawei Cloud, HarmonyOS, HMS and MDC for intelligent driving have been widely recognized by the industry and made great progress: HarmonyOS 2 has more than 120 million upgraded users; More than 140 thousand applications around the world have integrated HMSCore；; 90 global partners joined the MDC ecosystem; There are more than 10,000 Kun Peng certification solutions; More than 500 Shengteng partners; The total number of Huawei cloud developers exceeded 2.3 million.

　　Deng Taihua said that facing the Kunpeng ecology, Huawei is committed to helping its partners from "using Kunpeng" to "using Kunpeng well". Facing the rising ecology, Huawei focuses on the AI basic software and hardware platform, and builds the AI technology ecology and business ecology to realize two-wheel drive. Facing the future, Huawei will continue to invest firmly, build a new computing ecology with its partners, and win a new digital era.

　　From "Use Kunpeng" to "Use Kunpeng Well"

　　Kunpeng computing upgrades two open source basic software and two development and application enabling kits to help partners make good use of Kunpeng.

　　Faced with the challenges of ecological fragmentation, repeated application development and complicated collaboration brought by the "soft chimney" of the operating system, the openEuler is newly released and positioned as an open source operating system for digital infrastructure, which can be widely deployed in servers, cloud computing, edge computing, embedded devices and other forms. The application scenario covers IT(Information Technology), CT(Communication Technology) and OT(Operational Technology). Euler creatively proposed full-stack atomic decoupling, which supports flexible version construction and free service combination, realizes unified operating system, supports multiple devices, and covers the whole scenario.

　　Euler and HarmonyOS have shared the kernel technology, and plan to build a distributed soft bus capability in Euler in the future, so that devices equipped with Euler operating system can automatically identify and connect HarmonyOS terminals. Subsequently, we will further share the secure OS, device driver framework, and new programming languages. Through capacity sharing and ecological interoperability, "Euler+HarmonyOS" can better serve the whole digital scene.

　　This conference officially released the first innovative version of openEuler 21.09, which supports the whole scene of digital infrastructure, and will be launched in the community on September 30. This version not only enhances the capabilities of servers and cloud computing scenarios, but also supports edge computing and embedded scenarios. In the first quarter of 2022, Euler Open Source Community will release the LTS version of the community that supports full scenario integration, enabling partners to release commercial releases for different scenarios.

　　The openGauss Community Council was formally established, and 18 units, as the first members of the Council, will lead the openGauss open source community to create the world’s leading new ecosystem of enterprise-level open source databases. DevKit2.0 and BoostKit 2.0, the Kunpeng development kit, are newly launched, which help Kunpeng developers to carry out native development conveniently and efficiently, improve application performance, be a "developer’s developer", innovate and empower developers with minimal development and extreme performance, and jointly prosper Kunpeng’s computing ecology.

　　Continuously upgrade the basic software and hardware platform of Shengteng AI full stack.

　　Focusing on the core design concept of "extremely easy to use, extreme performance", CANN 5.0 continues to improve its performance. With the same hardware, the training performance of typical CV model doubles, and that of typical NLP model increases to 2.5 times. MindSpore’s Chinese name "Shengsi" made its debut, and the latest version of Shengsi 1.5 was released. This version strengthened the whole scene capability, natively supported the big model, and added a new paradigm of AI scientific computing, electromagnetic simulation suite and molecular simulation suite to promote the application of AI in scientific computing.

　　From artificial intelligence computing center to artificial intelligence computing network

　　Last year, the White Paper on the Development of Artificial Intelligence Computing Centers, which was first published by the Ministry of Science and Technology in conjunction with China Institute of Science and Technology Information (hereinafter referred to as "China Information Institute"), innovatively put forward the guiding ideology and construction scheme of "one center and four platforms", and artificial intelligence computing centers such as Shenzhen, Wuhan, Xi ‘an, Zhongyuan and Chengdu successively landed, greatly accelerating the development of AI industry in various regions. The White Paper 2.0 on the Development of Artificial Intelligence Computing Center-From Artificial Intelligence Computing Center to Artificial Intelligence Computing Network, jointly issued by CITIC, AITISA (New Generation Artificial Intelligence Industry Technology Innovation Strategic Alliance) and Pengcheng Laboratory, elaborated the development path and construction guidance from artificial intelligence computing center to artificial intelligence computing network in detail.

　　This time, 21 cities, including Shenzhen, Wuhan, Xi ‘an, Chengdu, Beijing and Shanghai, jointly lit up the "artificial intelligence computing network", connecting artificial intelligence computing centers in various places into a network, dynamically sensing the state of computing resources in real time, realizing overall distribution and scheduling of computing tasks, forming a network for sensing, distributing and scheduling artificial intelligence computing power in the region, and realizing the "three convergence" of computing power, data and ecology with "one network"

　　Big Computing Power+Big Data Enabling New Application of Big Model Incubation

　　The large computing power and big data gathered by artificial intelligence computing network enable the training of large models, thus incubating new applications and promoting the development of artificial intelligence industry. Institute of Automation, Chinese Academy of Sciences and Pengcheng Laboratory respectively released Zidong Taichu, a multimodal large model trained by artificial intelligence computing center, and Pengcheng Shennong, a bioinformatics research platform. Zidong Taichu is the industry’s first three-mode large model, which can achieve efficient collaboration among the three modes of vision, text and voice, and has the world’s leading cross-modal understanding and generation ability of graphics, texts and sounds. It can easily complete tasks such as intelligent question and answer, picture generation, video understanding and generation, and will be widely used in industrial quality inspection, film and television creation, Internet recommendation and intelligent driving. Pengcheng Shennong bioinformatics research platform includes protein structure prediction, small molecule generation, interaction prediction between target and small molecule, design and effect evaluation of new antibacterial peptides, etc. Pharmaceutical enterprises and medical research institutions will greatly accelerate the screening and creation of new drugs by using AI capabilities provided by Pengcheng Shennong. In addition, Pengcheng Lab and Wuhan University will continue to launch more large models based on Ascension AI. In the future, these large models will be open source to the whole world and can be obtained by visiting Ascension Community.

　　The core of business ecology is to truly integrate artificial intelligence into application scenarios. In order to realize the deep integration of AI experts and industry experts and get through the last mile of the landing of AI industry, Huawei proposed a brand-new cooperation mode of "selecting scenarios, jointly defining, jointly developing, combining marketing and continuously iterating". Together with partners, Huawei provides AI scenario solutions that can be replicated on a large scale for end customers. At present, under this model, four solutions of Shengzhizao, Shengzhicheng, Shengzhixing and Shengzhixun have been launched successively, covering manufacturing, smart cities, transportation, energy and other fields to help these industries achieve intelligent upgrading. Facing the future, Huawei and its partners will also launch Shengteng Zhixue Solution and Shengteng Zhixue Developer Kit to provide landing AI solutions for biomedical industry and education industry. Make AI easier to use and make the industry smarter.

　　Beiming Diversity Computing Convergence Architecture Released

　　Facing the era of diversified computing, the system has changed from single computing power to diversified computing power, from single-machine deployment to cluster deployment, and from traditional application to integrated intelligent application, which has brought more complex challenges to software developers. In order to help developers achieve minimal development, efficient deployment and extreme performance on diverse computing clusters, Huawei officially released the Beiming Diversity Computing Convergence Architecture. The architecture is a complete software stack for diversified computing hardware and clusters, including basic enabling, application enabling and development enabling.

　　Gather talents with industry and lead the industry with talents.

　　Colleges and universities are the base of talent training and the future of talent development. Under the guidance of the Ministry of Education, the Ministry of Education-Huawei "Smart Base" collaborative education base project will be further deepened.

　　In terms of expanding teaching aids, this conference recently released seven teaching aids, such as CANN Architecture and Programming of Ascending AI Processor and Advanced Technology of MindSpore Deep Learning, and cooperated with domestic famous teachers to launch three teaching materials, such as Introduction to Artificial Intelligence. Next year, it will continue to release 16 teaching materials and seven teaching aids, and fully integrate Kunpeng Ascending Roots Technology into college courses. In terms of expanding schools, Kun Peng’s promotion plan will cover more than 500 colleges and universities in 2023, and will gradually expand to more than 2,700 colleges and universities nationwide and some overseas head colleges in the future. In terms of expanding the teaching staff, "Smart Base" plans to empower more than 20,000 outstanding teachers and train more than 5 million science and engineering students in five years by opening a virtual teaching and research section. At the same time, Huawei helps to improve the multi-dimensional teaching methods by means of massive open online course, community, online experiments, and the practice of talents, so that more students can get industrial practice opportunities and improve the learning effect.

　　Finally, Deng Taihua announced that the "smart base" will expand the industrial scope, from Kunpeng, Shengteng and Huawei Cloud, which surround the computing industry, to the communication industry including 5G, optical network and data communication, covering more technical fields, and cultivating more outstanding talents for the new digital era.