MIL-OSI Banking: Huawei Yang Chaobin: Unleashing Intelligence with New Infrastructure

6
Recommended Sponsor Painted-Moon.com - Buy Original Artwork Directly from the Artist

Source: Huawei

Headline: Huawei Yang Chaobin: Unleashing Intelligence with New Infrastructure

[Shanghai, China, September 20, 2024] At HUAWEI CONNECT 2024, Huawei’s Director of the Board and President of ICT Products & Solutions Yang Chaobin delivered a keynote speech entitled “Unleashing Intelligence with New Infrastructure”. During his keynote, Yang said: “Huawei is committed to providing the ICT products and technologies needed to build a new type of intelligent infrastructure for numerous industries. We will continue working with our partners to provide high-quality enterprise connectivity, help break down data siloes, jointly develop enterprise AI solutions, and accelerate AI application in industries.”
Yang Chaobin delivering a keynote speech

Foundation model performance has seen remarkable improvement over the last year, and AI is rapidly becoming more affordable for enterprises. This has led to an increase in new inference applications and accelerated industry intelligent transformation. However, Yang stated that industry intelligence is still quite complicated given the number of challenges enterprises still face in infrastructure, data, scenarios, applications, ecosystems, and more, cautioning that a single chokepoint can cripple an enterprise’s productivity.
Yang said that data connectivity and storage are essential to addressing these problems as, “Success with intelligence first means success in digitalization. If data is the foundation of AI, then model training is its heart.” Foundation model pre-training and incremental industry-specific model training both depend heavily on the efficiency and reliability of the training system being used.
Huawei is also known for its premium networking, storage, and computing products and solutions. The company launched a range of new products and solutions under these categories as part of their infrastructure designed for industry intelligence, including their newest generation of OceanStor Dorado all-flash storage products, the iMasterCloud management platform, the Atlas 900 A3 SuperCluster AI cluster, and the iMaster CCAE cluster O&M platform. These products and solutions cover all enterprise data flows, spanning data connectivity and storage, model development and training, and application deployment.
Huawei has also recently created a reliable and easy-to-use AI training platform that uses innovative communication algorithms, parallel acceleration algorithms, development toolchains, and system O&M to cover multiple types of training, including pre-training and incremental training. This platform is intended to help customers develop models best suited to their own needs quickly and efficiently.
Huawei also offers a full range of AI inference products to meet customers’ scenario-specific deployment needs, which include modules, PCIe cards, edge stations, and servers. Inference applications tailored to specific industries are essential to unlocking the value of AI. According to Yang, flexible, easy-to-use inference systems “extend the reach of intelligence”.
Huawei provides R&D, ecosystem, and sales support for its open hardware products, and has worked with multiple partners who work on specialized scenarios to develop over 200 products and solutions made to meet specific industry needs.
Huawei also provides the Ascend inference engine MindIE software to these customers. MindIE integrates industry-leading inference acceleration technologies, such as communication-compute fusion algorithms, tiered KV-Cache, and prefill-decode disaggregation to make inference systems more efficient and enterprise AI deployment much more affordable.
The last set of offerings Yang discussed during his speech were Ascend AI hardware and software products that partners can combine with their own different-sized models to develop out-of-the-box, scenario-based enterprise AI solutions. These offerings include AI frameworks, fine-tuning kits, toolchains, and inference engines. Over 20 partners have already used these offerings to develop more than 100 enterprise AI solutions for different scenarios.
Yang concluded his speech by saying that the company plans to further collaborate with partners to build an intelligent industry foundation for shared success, and that, “by working together, we will create a more thriving ecosystem and help all enterprises keep pace with intelligent trends.”

MIL OSI Global Banks