SHANGHAI – At the Huawei CONNECT 2024 conference, Zhang Ping’an, Executive Director of the Board of Huawei and CEO of Huawei Cloud, delivered a presentation titled “Thriving with the Cloud: Reshaping Industries with AI”. According to Zhang, different companies should seize the opportunities in the era of intelligent technology, and use AI to build competitiveness. The key is to adopt the original AI mindset.
First, the company should actively welcome the presence of AI, explore different industrial scenarios, and develop the enterprise AI platform. Thus, AI can serve major companies. Today, Huawei Cloud Pangu models have been applied in more than 30 industries and 400 scenarios.
Second, AI computing power plays a major role. The company should build a suitable AI-based cloud computing infrastructure. Zhang also officially launched CloudMatrix, which connects and aggregates all computing resources, including CPU, NPU, DPU, and memory. CloudMatrix represents the evolution from homogeneous computing concepts to arrays. CloudMatrix consists of an AI-based cloud computing infrastructure that aggregates all computing resources, makes them peer-to-peer, and performs synthesis. This provides a huge amount of AI computing power for the company.
Third, the quality of data determines the effectiveness of the AI model. We need to build scientifically oriented databases so that data can better serve AI. Huawei Cloud has updated DataArts to provide customers with a scientifically oriented and AI-oriented database. The new DataArts offers a number of features, including the AI+ data convergence engine, data development and management, knowledge services, and AI+ data application enablement services.
Fourth, build the right AI model according to business applications. We need to abandon the misconception that the best models are large-scale AI models. Not all needs can be met with just one basic AI model. Pangu Models 5.0 is available at different scales, with billions of parameters, tens of billions of parameters, hundreds of billions of parameters, and even trillions of parameters. The full range of Pangu models meets almost any business application needs.
On the other hand, Huawei Cloud’s Pangu 5.0 model improves the spatiotemporal controllable generation (STCG) in the field of multi-modal generation. For autonomous driving, the Pangu model is able to accurately produce driving scenarios that closely resemble physical models. The Pangu model can also generate various videos such as normal driving conditions, random road conditions, as well as daily and aggressive driving patterns. Therefore, automobile companies can train autonomous driving more efficiently.
Huawei Cloud officially launched the Mainframe-to-Cloud solution to help customers develop new cloud computing platforms with high availability, easy operation and maintenance, and better speed. This helps customers achieve high service availability of up to 99.999% in the financial sector. Now, most banks in China have chosen Huawei to develop new cloud computing platforms.
Tao Jingwen, Huawei’s Board Director and President of Quality, Business Operations, and IT, discusses Huawei’s digital transformation. Huawei has developed a methodology consisting of three levels, five stages, and eight steps. These three levels represent a breakthrough in intelligent technology business, developing and implementing AI models, and improving intelligent applications. Meanwhile, the five stages include identifying the right scenarios, changing different processes, transforming different enterprises, optimizing enterprise data, and adopting different AI applications.
Bruno Zhang, CTO of Huawei Cloud, gave a presentation on Huawei Cloud’s role in leveraging AI to revolutionize data centers, infrastructure, and cloud computing services. He also discussed the creation of AI-powered cloud computing that accelerates the development of intelligent technologies in various industries. The Pangu assistant system “1+N,” which Zhang reported, represents the arrival of a new era in the interaction of cloud computing services. The system covers Pangu Doer and various use cases in product R&D, data analysis, security, and office collaboration. In such applications, Pangu models are trained using data based on specific scenarios, experiences, and practices, which improve cloud computing services and their efficiency through tailored AI.
