Latest news with #Pangu
South China Morning Post
6 days ago
- Business
- South China Morning Post
Huawei claims better AI training method than DeepSeek using own Ascend chips
Researchers working on Huawei Technologies ' large language model (LLM) Pangu claimed they have improved on DeepSeek's original approach to training artificial intelligence (AI) by leveraging the US-sanctioned company's proprietary hardware. Advertisement A paper – published last week by Huawei's Pangu team, which comprises 22 core contributors and 56 additional researchers – introduced the concept of Mixture of Grouped Experts (MoGE). It is an upgraded version of the Mixture of Experts (MoE) technique that has been instrumental in DeepSeek's cost-effective AI models. While MoE offers low execution costs for large model parameters and enhanced learning capacity, it often results in inefficiencies, according to the paper. This is because of the uneven activation of so-called experts, which can hinder performance when running on multiple devices in parallel. In contrast, the improved MoGE 'groups the experts during selection and better balances the expert workload', researchers said. In AI training, 'experts' refer to specialised sub-models or components within a larger model, each designed to handle specific tasks or types of data. This allows the overall system to take advantage of diverse expertise to enhance performance. 01:38 China a 'key market', says Nvidia CEO Huang during Beijing visit as US bans AI chips China a 'key market', says Nvidia CEO Huang during Beijing visit as US bans AI chips The advancement comes at a crucial time, as Chinese AI companies are focused on enhancing model training and inference efficiency through algorithmic improvements and a synergy of hardware and software, despite US restrictions on the export of advanced AI chips like those from Nvidia
Korea Herald
09-05-2025
- Business
- Korea Herald
Huawei: Keeping Cities Moving, Amplifying Urban Rail Intelligence
HONG KONG, May 9, 2025 /PRNewswire/ -- The Huawei Smart Urban Rail Summit, themed "Steaming Ahead with Cloud-Network Convergence and Data & AI Enablement for Smart Urban Rail", was held in Hong Kong, China. Thomas Xu, Vice President of Huawei's Smart Transportation BU and Director of the BU's Partner Development and Sales Department highlighted the importance of identifying the most suitable information, digital, and intelligent technologies for addressing specific service needs. Dr. Tony Lee Kar-yun, Operations and Innovation Director at MTR, believed that the integration of digital resources and innovative AI applications can enhance operational efficiency and passenger experience, and pointed out that this integration will serve as the foundation for more efficient urban transportation and more convenient community life. Mr. Xi Xiaodong, Chief Engineer of Shanghai Rail Transit Maintenance Support Co., Ltd., emphasized the need for metro operators to integrate external resources and enhance internal core capabilities if they are to achieve high-quality development. Nelson Huang, Rail Business Director of Huawei's Smart Transportation BU, stated that Huawei integrates innovative ICTs, such as the cloud, big data, 5G, AI, and IoT based on the digital and intelligent foundation for urban rail. Mr. Xiong Xinbin, Rotating CEO and Senior Vice President of Beijing Railway Institute of Mechanical & Electrical Engineering (BRI) noted that they have deployed Pangu models to empower rail transportation, utilized Ascend and Kunpeng to lay a solid data foundation, and integrated AI to develop a comprehensive intelligent O&M platform. During the follow-up roundtable salon, Mr. Ji Kun, Overseas Rail Solution Director of Huawei's Smart Transportation BU; Datuk Dr. Mohd Yusoff Sulaiman, Chairman of the Malaysian Rail Industry Corporation (MARIC); Chan Hing-keung, Chief of Operations Engineering Service & Innovation at MTR; Mr. Xi Xiaodong, Chief Engineer of Shanghai Metro Maintenance and Support; Mr. Xiong Xinbin, Rotating CEO and Senior Vice President of BRI; and David Xu, Vice President of Huawei's Smart Transportation BU and Director of the BU's Solution Development I&V Dept — delved into technology enablement scenarios, transformation pain points, data value mining, and user experience upgrade. Huawei has served more than 300 urban rail lines in over 70 cities around the world. Moving forward, Huawei will deepen its cooperation with industry partners by consistently adhering to the principle of "openness, cooperation, and shared success". The company is determined to realize convenient travel and smooth logistics, alongside digital and intelligent transportation. Photo - Photo - Photo - Photo - Photo -
Channel Post MEA
04-03-2025
- Business
- Channel Post MEA
Huawei Cloud Unveils New Cloud Services And Solutions At MWC 2025
The Huawei Cloud Summit, with the theme 'Accelerate Intelligence, Amplify Success', concluded today in Barcelona, bringing together hundreds of industry leaders and experts from sectors including carrier, finance, Internet, and government. Huawei Cloud unveiled a series of cutting-edge services and solutions, solidifying its commitment to driving intelligent transformation across diverse industries. Today, AI is changing everything. Countries and companies are putting AI as a core strategy, and have the chance to build their own models. AI is not a game of heavy costs. AI models can be optimized by leveraging the achievements from the open source community. Data, especially high-quality data, plays a more and more important role in making AI a reality. Jacqueline Shi, President of Huawei Cloud Global Marketing and Sales Service Jacqueline Shi, President of Huawei Cloud Global Marketing and Sales Service, pointed out that Huawei Cloud always puts AI as the company's core strategy, and by transitioning from cloud-native to AI-native, Huawei Cloud leads the industry with non-stop innovation. She elaborated on that Huawei Cloud defines AI-Native in two sectors. One is 'AI for Cloud', which means they reshape and upgrade all Huawei Cloud services into intelligent ones using AI. The other is 'Cloud for AI', where Huawei Cloud aims to build the best platform, including compute architecture, data, model tools, and professional services, to accelerate the development of AI. In 2024, Huawei Cloud experienced significant growth outside of the Chinese mainland, expanding by over 50%. The company has become a preferred partner for numerous organizations seeking for digital and intelligent upgrades, including over 140 carriers, 500 financial institutions. Jacqueline restated Huawei Cloud's determination in upgrading their cloud services and solutions to satisfy the requirements of customers in all industries, helping them accelerate intelligence. For carriers, Huawei Cloud provides a new connection solution, helping them connect more applications to drive consumer and enterprise businesses. Huawei Cloud provides a hybrid cloud to build a unified foundation for internal IT and B2B business. The Cloud Device solution connects all applications for people and families. The cloud marketplace KooGallery helps carriers establish their ecosystem. The Cloud on Cloud solution enables carriers to build their cloud and make cloud business easier. For governments, Huawei Cloud enables their shift from e-Government and digital services to smart governance. Huawei Cloud Pangu government model, aPaaS, and big data services help governments better serve citizens. In the finance sector, Huawei Cloud provides the Digital Core solution to replace mainframes and accelerate application modernization. The financial big data solution and Pangu finance model support intelligent customer service and smart data analysis. In the media and entertainment industry, Huawei Cloud provides intelligent interaction solutions for content generation and user engagement, with CDN & Cloud Live, Cloud Media Edge, and MetaStudio. For retailers, Huawei Cloud helps customers build smart stores, including real-time data insight, smart order prediction, and digital shop assistants. In the automotive field, Huawei Cloud not only provides leading intelligent IoV and autonomous driving solutions, but also helps customers realize better R&D, production, and supply chain management with smart manufacturing solutions. Bruno Zhang, CTO of Huawei Cloud, shared his insights on transitioning from cloud-native to AI-native. In this AI era, the foundation of achieving AI-native lies in being cloud native. Over the past few years, Huawei Cloud has focused on assisting businesses in effectively going to, using, and managing cloud through cloud native means. Building upon this foundation, Huawei Cloud is now advancing towards the next stages: constructing AI-native infrastructure, such as CloudMatrix and AI Cloud Service; developing AI-native cloud services, including databases, knowledge lakes, and software development pipeline; and offering AI models and tools for customers and partners to better train and run their AI models. Bruno Zhang, CTO of Huawei Cloud Huawei Cloud sticks to the 'AI for Industries' philosophy to build and upgrade their Pangu models, which now see wide adoption in over 30 industries across 400 use cases, including public services, finance, manufacturing, drug discovery, mining, steel manufacturing, railways, autonomous driving, industrial design, architectural design, and meteorology, driving real value creation everywhere. For example, robots powered by Pangu models can accurately inspect 32,000 items on a high-speed train, covering more than 350 complex fault conditions in 8 categories. They help high-speed railways improve operational efficiency and cut labor costs. Manual tasks in the early morning are significantly reduced. The Pangu Steel Manufacturing Model has been deployed at one of Baowu Group's hot rolling production lines. It can quickly and accurately predict the optimal settings for hot rolling production lines, as well as critical quality metrics of the products. This helps improve the yield rate of steel plates. At MWC 2025, Huawei Cloud introduced an array of cloud services and solutions to the global market, aiming to expedite the evolution from cloud native to AI-native. CloudDC Solution provides one-stop data center-to-cloud services, boosting three major capabilities: deterministic operations, cloud DC, and full-stack AI. Enterprise IT Cloud Migration Solution offers over 240 cloud services, backed by unified O&M services, flexible billing options, and professional services for cloud migration and more to smooth cloud migration. Backup & Disaster Recovery Continuity Center Solution ensures seamless continuity across multiple layers, including management, application, host, and storage, as well as cloud data centers. It provides comprehensive coverage for three critical scenarios: cloud native backup and DR, IDC-to-cloud backup and DR, and cross-cloud DR. With advanced backup and DR capabilities, such as risk detection within seconds, rapid backup and recovery, and centralized management, this solution empowers businesses to create highly available architectures and safeguard their core data and applications. Large Model Security Solution is an end-to-end solution that covers environment security, LLM corpus security and compliance, inference security and compliance, and unified security operations. Cloud Multilayer Ransomware Protection Solution provides comprehensive ransomware protection against intrusions, attack spread, leakages, and tampering to defend against ransomware attacks all round. GaussDB & TaurusDB: GaussDB's innovative three-layer pooling architecture enables layered and on-demand resource scaling, while its intra-city dual-cluster disaster recovery capability isolates software and hardware faults between clusters. The AI DevOps assistant automates 90% of fault diagnosis and rectification. TaurusDB, fully compatible with MySQL, requires zero refactoring for business system migration. Featuring AI-based serverless capabilities, TaurusDB is also the only serverless cloud service that supports write scaling. AI Cloud Service, compatible with more than 100 mainstream models including DeepSeek and Llama, supports efficient, uninterrupted training of foundation models for up to 40 days and quickly rectifies faults within 10 minutes. Cloud Search Service (CSS) – KooSearch: CSS offers a high-performance platform for searching and analyzing logs, able to handle hundreds of millions of search requests in milliseconds, a powerful service facilitating data-driven O&M and operations. KooSearch is a comprehensive knowledge chat solution with over 90.9% search accuracy, speeding up data monetization efforts. Cloud Device, providing KooPhone, KooDrive, Cloud Gaming, Cloud STB, and Cloud PC, leverages boundless cloud compute and storage to overcome device limitations, paving the way to everything cloud-enabled. KooVehicle, an intelligent vehicle-cloud synergy solution, enables 24/7 secure, autonomous driving in open-pit mines under different circumstances, achieving significant efficiency and safety gains. Over the past year, Huawei Cloud has helped numerous top-tier customers achieve sustained business success through its innovative solutions. Iñaki Fuentes, COO of TravelgateX, shared how they developed world-class travel services with Huawei Cloud's cloud native technologies. TravelgateX leveraged Huawei's cloud native solution to accelerate program development, simplify deployment, and scale its infrastructure. This resulted in significant cost savings of 20% and improved overall performance by 35%. With Huawei Cloud, TravelgateX was able to deliver personalized services and make more informed business decisions. Iñaki Fuentes, COO of TravelgateX Fajrin Rasyid, Chief Digital Innovation Officer of Telkom Indonesia, said that they have signed a strategic cooperation memorandum with Huawei on data centers, cloud services, digital technologies, and big data to transform into a digital carrier for a thriving digital ecosystem in Indonesia. Fajrin Rasyid, Chief Digital Innovation Officer of Telkom Indonesia Mustafa Kemal Erișen, the IT head of Turkish e-commerce player Hepsiburada said that they have seen significant improvements after migrating their search service to Huawei Cloud, achieving a 50-fold increase in platform throughput, a 95% reduction in service response time, and a 150% boost in bandwidth usage. Mustafa Kemal Erișen, IT Head of Hepsiburada At the Summit, Huawei Cloud and its partners unveiled over 20 joint solutions catering to industries such as finance, automotive, retail, media, and public services, aimed at accelerating digital transformation in these industries. Huawei Cloud also held a Cloud Native Elite Club (CNEC) roundtable with the theme of 'Accelerate Data-AI Convergence, Build Differentiated AI Capabilities'. Visionary leaders gathered to explore the building of robust data foundations for the AI era, and to share best practices for unleashing data value and building differentiated strengths. 0 0
