【观点】徐文伟：深耕工业数字化，加速智能制造高质量发展

2023世界智能制造大会在南京开幕。华为Fellow、科学家咨询委员会主任、CCF中国数字经济50人论坛委员徐文伟发表了主题演讲《深耕工业数字化，加速智能制造高质量发展》。他表示，大模型重塑制造业，将快速进入生产系统，赋能行业。运力，存力和算力是迈向智能时代的基础设施底座，当前我国算力和存力仍有较大提升空间，我们需要升级技术供给系统，实现新跨越发展。 CXO UNION-CXO联盟（cxounion.cn）

以下是徐文伟演讲全文

尊敬的各位领导和来宾，大家好！

今天会议的主题是“让制造更聪明”，“让制造更聪明”一定会涉及到人工智能。当前，大模型主要是对公开数据集数据进行训练；未来6-12个月，将会对行业专业数据进行训练；未来12-24个月，对行业核心数据进行训练。在这种情况下，大模型使能制造业真正成为智能制造。

大家知道，大模型一定涉及到算力。根据统计，算力对于经济具有极强的杠杆效应，每1人民币的算力投入在制造业至少会产生10人民币的产出。当前，中国的整体算力已经接近美国，但人均算力只是美国的1/5左右。同时，数据存储容量/数据算力体量的存算比也仅位列全球第11位，还有较大提升空间，所以我们不仅要投算力，还要投存力。

当工业进入智能制造时代，我们的目标是实现智能制造的多快好省。为此，华为携手产业伙伴使能新型工业化，从产品的研发、制造，到销售，构建智能制造全流程体系。这其中的工业智能体包含三个部分：其一是AI行业大模型作为“大脑”；其二是先进工业网络作为贯通全身的“神经”；其三是数字工业装备和工业边缘引擎作为“四肢”。

智能工业装备作为工业智能体的“四肢”。当前，产业面临着OT系统和IT系统的分离，智能制造的未来一定是融合计算，也就是通用计算和智能计算一体化，IT系统和OT系统一体化。我们提供大、中、小算力的全场景系列化解决方案。华为也是一个制造业公司，拥有三十多年的制造业经验和200多条生产线，我们把经验总结凝练后和社会、产业进行分享，打造一站式AI质检方案，通过和自动化设备、AI推理子系统和AI训练子系统的协同，实现了一站交付、快速换线的方案部署，在钢铁、冶金、封装等领域实现大规模应用。 CXO UNION-CXO联盟（cxounion.cn）

工业网络作为工业智能体的“神经”。没有联接，数据就上不来，智能就下不去。当前，设备网联化率只有30%左右。过去五年，每个车间的无线联接数量只有70多个，现在达到了1500多个，五年内已增长20倍以上。全联接系统只有加强网络装备的智能化，才能真正变成一个智能制造。以南京华瑞德智能物流为例，传统的物流都是堆叠的模式，这种模式好处是简单，但库存的容量只有60%，随着未来的发展，从60%提升到90%以上的库存率，将带来联接的屏蔽和安全可靠问题，通过与华为的联合创新，南京华瑞德智能物流的方案实现库存容量达到90%以上。

在未来，我们希望整个工厂能实现联接IP化，让设备讲普通话，用开放架构去实现互联互通。这需要解决IP网络延、抖动以及丢包等问题，所以我们发展了TSN（时延敏感网络）等创新技术，保证IP网络稳定可靠、时延可管可控。

“让制造更聪明”的核心是以人工智能为代表的大模型。华为盘古大模型实际上分成三层：L0层是我们提供盘古基础大模型，包括自然语言处理、计算机视觉、科学计算等。基于这个技术平台，我们提供L1层，也就是行业的数据经过训练以后，在矿山、气象预报、医疗等领域提供的行业大模型。再上一层是L2层场景大模型，针对特定应用场景进一步地进行优化和使用，可以联合大量的合作伙伴进行研究，实现商业化转型。

华为一直以来扎根江苏，坚持在江苏为江苏，在南京为南京。华为南京研究所是我们全球最大的软件研究所，我们在南京设立了天工实验室，目的是打造一个智能制造的平台，使能全国的智能制造产业，持续为江苏服务，为江苏的“让制造更聪明”添砖加瓦。 CXO UNION-CXO联盟（cxounion.cn）

翻译：

Xu Wenwei: Deepen industrial digitization and accelerate the high-quality development of intelligent manufacturing

The 2023 World Intelligent Manufacturing Conference opens in Nanjing. Xu Wenwei, Huawei Fellow, Director of the Scientist Advisory Committee, and member of the CCF China Digital Economy 50 Forum, delivered a keynote speech titled “Deepening Industrial Digitization and Accelerating the high-quality Development of Intelligent Manufacturing”. He said that the large model reshaping the manufacturing industry will quickly enter the production system and empower the industry. Transport capacity, storage capacity and computing power are the infrastructure base towards the intelligent era. At present, China’s computing capacity and storage power still have a large room for improvement. We need to upgrade the technology supply system to achieve new leap-forward development. CXO UNION-CXO联盟（cxounion.cn）

Xu Wenwei, Huawei Fellow, Director of the Scientist Advisory Committee, and member of CCF China Digital Economy 50 Forum

The following is the full text of Xu Wenwei’s speech

Dear leaders and guests, hello!

The theme of today’s conference is “Make manufacturing smarter”, and “Make manufacturing smarter” will definitely involve artificial intelligence. At present, large models are mainly trained on public data sets. In the next 6-12 months, training will be conducted on industry professional data; In the next 12-24 months, train the industry core data. In this case, large models enable manufacturing to truly become smart manufacturing. CXO UNION-CXO联盟（cxounion.cn）

As you know, large models must involve computing power. According to statistics, computing power has a strong leverage effect on the economy, and every 1 yuan of computing power input will produce at least 10 yuan of output in the manufacturing industry. At present, China’s overall computing power has approached that of the United States, but its per capita computing power is only about one-fifth of that of the United States. At the same time, the storage ratio of data storage capacity/data computing power volume is only ranked 11th in the world, and there is still a lot of room for improvement, so we should not only invest in computing power, but also invest in storage power.

When the industry enters the era of intelligent manufacturing, our goal is to achieve intelligent manufacturing how fast and good. To this end, Huawei works with industry partners to enable a new type of industrialization, from product research and development, manufacturing, to sales, to build a whole process system of intelligent manufacturing. The industrial agent consists of three parts: one is the large AI industry model as the “brain”; The second is the advanced industrial network as a “nerve” through the whole body; The third is digital industrial equipment and industrial edge engines as “limbs”.

Intelligent industrial equipment as the “limbs” of industrial intelligence. At present, the industry is facing the separation of OT system and IT system, and the future of intelligent manufacturing must be converged computing, that is, the integration of general computing and intelligent computing, and the integration of IT system and OT system. We provide full scenario serialization solutions for large, medium and small computing power. Huawei is also a manufacturing company with more than 30 years of manufacturing experience and more than 200 production lines. We have summarized our experience and shared it with society and industry to create a one-stop AI quality inspection solution. Through collaboration with automation equipment, AI reasoning subsystem and AI training subsystem, we have realized the solution deployment of one-stop delivery and rapid line change. In steel, metallurgy, packaging and other fields to achieve large-scale applications. CXO UNION-CXO联盟（cxounion.cn）

Industrial network is the “nerve” of industrial agent. Without connectivity, data can’t come up, intelligence can’t come down. At present, the network connectivity rate of equipment is only about 30%. In the past five years, the number of wireless connections per workshop was only more than 70, and now it is more than 1,500, which has increased more than 20 times in five years. Only by strengthening the intelligence of network equipment can the fully connected system truly become an intelligent manufacturing. Taking Nanjing Huarui Intelligent logistics as an example, the traditional logistics are stacked mode, the advantage of this model is simple, but the capacity of the inventory is only 60%, with the future development, from 60% to more than 90% of the inventory rate, will bring the connection of shielding and safety and reliability issues, through the joint innovation with Huawei, Nanjing Huaride intelligent logistics solution to achieve inventory capacity of more than 90%.

In the future, we hope that the entire factory can be connected to IP, so that the equipment can speak Mandarin, and use open architecture to achieve connectivity. This needs to solve the problems of IP network delay, jitter and packet loss, so we have developed innovative technologies such as TSN (delay sensitive network) to ensure that the IP network is stable and reliable, and the time delay can be controlled. CXO UNION-CXO联盟（cxounion.cn）

The core of “Make manufacturing smarter” is a large model represented by artificial intelligence. Huawei Pangu Grand model is actually divided into three layers: Layer L0 is we provide Pangu basic grand model, including natural language processing, computer vision, scientific computing, etc. Based on this technology platform, we provide the L1 layer, that is, the industry data after training, in the mining, weather forecasting, medical and other fields to provide industry large models. The next layer is the L2 scenario large model, which can be further optimized and used for specific application scenarios, and can be researched with a large number of partners to achieve commercial transformation.

Huawei has always been rooted in Jiangsu, insisting on being Jiangsu in Jiangsu and Nanjing in Nanjing. Huawei Nanjing Research Institute is our largest software research institute in the world, we set up the Tiangong laboratory in Nanjing, the purpose is to create an intelligent manufacturing platform, so that the intelligent manufacturing industry in the country can continue to serve Jiangsu, and contribute to Jiangsu’s “make manufacturing smarter”. CXO UNION-CXO联盟（cxounion.cn）

由CXO UNION-CXO联盟（cxounion.cn）转载而成，来源于苏写有为公众号；编辑/翻译：CXO UNIONCXO联盟小U。

如需加入CXO UNION（CXO联盟）高管社群，请联系社群小伙伴哦~

免责声明: 本网站(http://www.cxounion.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。

本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。