深耕數字化，華爲首推分佈式雲原生產品華爲雲UCS

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"9月23日，在Huawei Connect 2021"},{"type":"link","attrs":{"href":"https:\/\/www.huawei.com\/cn\/events\/huaweiconnect","title":"xxx","type":null},"content":[{"type":"text","text":"華爲全聯接大會"}]},{"type":"text","text":"上，華爲輪值董事長徐直軍發表了主題演講《深耕數字化》，並宣佈正式"},{"type":"text","marks":[{"type":"strong"}],"text":"發佈華爲分佈式雲原生產品——華爲雲UCS"},{"type":"text","text":"。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"“隨着雲原生應用深入企業各個業務場景，雲原生正在走向分佈式，跨雲跨地域統一協同治理，保證一致應用體驗等新的需求日漸突出。”徐直軍表示，UCS的意思是"},{"type":"text","marks":[{"type":"strong"}],"text":"無處不在的雲原生服務"},{"type":"text","text":"，致力於讓企業使用雲原生應用時感受不到地域限制，感受不到跨雲限制，感受不到流量限制，把雲原生能力帶入企業的每一個業務場景，加速千行百業擁抱雲原生。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了雲服務，徐直軍還從人工智能、網絡以及低碳發展等方面分享了華爲最新的創新進展和產業思考。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"人工智能方面，MindSpore的成績尤爲顯著。自從"},{"type":"link","attrs":{"href":"https:\/\/www.infoq.cn\/article\/Nk06aiVTKvvpjYyzGdGA","title":"xxx","type":null},"content":[{"type":"text","text":"MindSpore"}]},{"type":"text","text":"於2020年3月開源後，截至2021年8月底的1年多時間內，其社區累計下載量超過60萬，有超過100家高校選擇MindSpore進行教學，"},{"type":"text","marks":[{"type":"strong"}],"text":"MindSpore已經成爲中國主流AI計算框架"},{"type":"text","text":"。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此外，Atlas 900集羣和基於Atlas 900集羣的雲服務已經服務於300多企業客戶，目前成功訓練出了一系列大模型，包括：華爲雲盤古NLP中文大模型；華爲雲盤古計算視覺大模型；華爲雲盤古藥物分子大模型；華爲雲盤古科學計算大模型；遙感專用框架LuojiaNet；鵬城盤古NLP中文大模型；鵬城生物製藥大模型等。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"針對華爲雲"},{"type":"link","attrs":{"href":"https:\/\/xie.infoq.cn\/article\/be2dfc0ad09543d9603774d2a","title":"xxx","type":null},"content":[{"type":"text","text":"ModelArts"}]},{"type":"text","text":"，徐直軍用了較多的篇幅進行介紹。ModelArts屬於AI應用使能層，其全流程、場景化服務大大簡化了AI應用開發，最終實現讓每一位工程師都能開發AI應用。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"對於企業來講，開發AI應用有三個階段，針對不同階段和場景，ModelArts均提供針對性的服務。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在初始階段，也就是探索性AI試驗階段，主要訴求是開發出一個模型，驗證初步可行性，處於這個階段的企業的AI能力往往最爲欠缺。針對性地，ModelArts通過領域套件、場景樣例、盤古大模型、預訓練模型等服務和開發工具，使得一般的工程師在簡單學習之後，基本不用開發代碼就能夠快速訓練出AI模型，快速驗證，快速上手。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二個階段是Quick Win，即在試驗成功的基礎上，期望快速通過AI爲企業創造價值，此時的AI開發不再是研究性模型開發，而是要針對一個或多個特定生產任務，結合特定部署場景，結合特定行業規範，滿足可信等生產性要求。ModelArts通過增加可信組件與安全算法、ModelBox、AutoSearch、盤古大模型等使AI工程師能適應多樣性的部署環境，快速開發生產性實戰型AI應用。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第三個階段是系統性AI應用或智能子系統的開發，此時往往涉及多種應用、工具和系統的配合，ModelArts通過進一步增加MLOps、OptVerse求解器、科學計算、盤古大模型和異構分佈式調度器以及來自生態夥伴的豐富的行業性組件與工具等支持企業實現極簡、高效的智能系統開發。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"企業網絡方面，華爲提出"},{"type":"text","marks":[{"type":"strong"}],"text":"自動駕駛網絡願景"},{"type":"text","text":"，即未來的網絡應該與自動駕駛汽車一樣都能自己運維自己，而不是靠人。“我們期望未來的自動駕駛網絡應該是：支持自動，即根據用戶意圖業務自動部署，最終目標是業務全自動部署；支持自愈，預測預防故障並基於事件自我恢復，最終目標是實現全自動運維；支持自優，根據用戶體驗自適應調整優化，最終目標是實現全自動優化；支持自治，在自動、自愈、自優的基礎上，網絡功能自適應、自學習、自演進；這是自動駕駛網絡的願景，也是終極目標。”徐直軍說道。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"目前華爲自動駕駛網絡既支持端到端的網絡質量可視，還對75類典型故障實現了3分鐘定位根因，5分鐘給出修復建議，今年基於知識圖譜的自學習能力，現網數據實現了在線增量學習，可覆蓋97%的故障。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"數字技術是低碳化發展的基本要素，因此，在低碳發展方面，華爲通過三個方向幫助客戶實現低碳發展，包括：投資創新節能技術，持續提升ICT產品能效，促進ICT產業自身低碳發展；投資電力電子技術及與數字技術融合創新，推動清潔能源發展與傳統能源數字化；把數字技術帶給每個行業，支持各行各業通過數字化促進低碳發展。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"徐直軍表示：“數字化走到今天，取得今天的進步，因爲有不斷的創新。數字化走向未來，實現更加宏偉的目標，還要靠不斷的創新，堅持創新永不止。”"}]}]}