如何认识这一轮AI技术发展影响的深度和广度？如何判断这一轮AI技术变革的性质？
AI大模型是通用人工智能发展的重要里程碑。4月28日、5月5日中央会议关于人工智能发展提出三个第一次：第一次提出“通用人工智能”，第一次提出“产业智能化”，第一次提出“把握人工智能等新科技革命浪潮”。
这一轮AI所体现“革命性”特征，不是AI可以生成文字、可以生成图片，而是AI可以生成代码，构建人机交互新模式，与产品研发、工艺设计、生产作业、产品运营等制造环节、场景相结合，提升生产效率，形成新生产力，并引发产业竞争格局重构。
AI大模型事关人类生产工具的变革，事关一个国家制造业核心竞争力重塑，事关经济的长期繁荣和走向。
一、数实融合是全球制造业竞争格局重构的核心变量
数字化是巨变时代的分水岭，已成为企业、城市、国家之间竞争急剧分化的催化剂。制造业是数实融合最主要的产业部门，其融合的方式、广度和深度，能够直接影响甚至决定制造业的先进水平和全球竞争格局。 数字化转型网（www.szhzxw.cn）
1. 数实融合是美国制造业全球领先的根本原因
目前，许多人潜意识里非常认同数字时代德国制造业发展道路和模式，“唱衰”美国制造业。然而事实上，过去十年，美国是全球制造业发展的“样板间”。
无论是制造业的规模、增速还是竞争力，美国均持续领先德国和日本，并且差距不断拉大：2011到2021年，美国制造业规模相当于德国的1.5倍扩大到2.4倍，相当于日本的2.5倍扩大到3.1倍。美国制造业仍然以比日本和德国更快的速度发展。
过去10年，美国制造业领先地位的巩固和确立，是“软件定义硬件”的结果，是以“云+AI”为代表的数字技术深度、全面融入实体经济的结果，是向新型数字基础设施进行迁徙的结果。
2. 数字原生企业涌现是美国制造业升级的重要标志
数字原生企业的涌现是产业升级和经济繁荣的重要标志。
德国、日本与美国制造业差距的扩大，在微观上主要表现为制造业领域缺乏一批有竞争力的数字原生企业。 数字化转型网（www.szhzxw.cn）
德国“工业4.0”目标并没有实现，与预期差距较大，中小企业数字化进展缓慢，研究表明只有21%的中小企业在生产中使用了数字技术，也没有成长出一批数字时代有竞争力的中小企业。
日本的情况与德国类似，日本经历了“失去的二十年”，同样没有培养出一批数字原生企业。
数字技术驱动美国新企业、新产品不断涌现，以特斯拉、SpaceX、Rivian、OpenAI、Snowflake、Palantir等为代表数字原生企业，不仅成为全球的产业引领者，而且持续构建产品创新的新模式。
特斯特引领全球汽车工业电动化、智能化，正如瓦伦丁在《从丰田主义到特斯拉主义》所描述的，电动车领域的数字原生企业正在重新定义汽车，在个性化、以用户为中心的万物互联、万物智能时代，新物种为应对不确定性而生，以高频创新为驱动，基于进化组织的迭代思维，从MVP小步快跑加速商业化。
特斯拉作为一个数字原生企业，具有四个典型特征：软件定义、高频创新、客户运营商、进化型组织。
2010年美国DARPA推出自适应运载器制造（AVM）计划，提出“重新发明制造业”，通过彻底变革和重塑装备制造业，将武器装备研制周期缩短到现在的五分之一。
复杂制造业产品研制生产要像半导体行业一样，其产品设计、仿真、试验、工艺、制造等活动，全部都在数字空间完成，待产品迭代成熟后再进入工厂一次制造完成，缩短研制周期、降低研制成本。这一战略已经开花结果。 数字化转型网（www.szhzxw.cn）
2023年11月11日，美国新一代隐形战略轰炸机首飞，这是全球第一款数字轰炸机-B21，从一开始就采用数字化设计，基于云计算的开发、部署和测试数字孪生，带来更好的维护性和更长的生命周期，以及更低的基础设施成本,是最近30年来美军研发速度最快的机型，还可以像特斯拉汽车不断下载新软件一样，能够不断升级产品功能，战斗力升级将依赖于敏捷软件迭代。
这一趋势的本质是：“云+AI”已经不仅是一个商业基础设施，更是一个创新基础设施，是新企业、新产品孕育孵化的摇篮。
3. AI大模型是重塑全球制造业竞争格局的新起点
AI大模型正加速第三次“数实融合”浪潮全面到来，智能化是其主要特征。AI大模型将影响制造业发展格局，AI大模型将会融入制造业的研发设计、生产工艺、质量管理、运营控制、营销服务、组织协同和经营管理的方方面面。
在研发设计领域，AI革新传统的科研范式。
在生物医药领域，2022年，DeepMind开发的AlphaFold2模型几乎预测了所有的蛋白质结构。如今AI模型不仅能“预测”、还可以“生成”蛋白质，为未来的药物生产研发创造新的可能。例如由Salesforce Research公司开发的ProGen系统成功从零生成全新的蛋白质。 数字化转型网（www.szhzxw.cn）
RNA（核糖核酸）病毒研究计算效率低、不精准，中山大学基于Transformer架构的“LucaProt”深度学习模型，训练了大型蛋白质语言模型，将病毒发现周期从过去2-3个月时间缩短为一周，发现了数万种依赖传统人工比对方法无法识别的新型病毒，将全球RNA病毒多样性扩充了近30倍。这将会将缩短疫苗研制周期、降低研制成本。
在生产制造环节，AI大模型可以直接服务智能汽车、机器人、芯片、服装等产品的研发创新，例如工程师可通过大模型自动生成代码指令，完成机器人功能的开发与调试，甚至还能为机器人创造一些全新的功能。
在设备运维环节，AI大模型大幅增强了传统垂直模型的能力。AI大模型具备了理解能力，电力行业无人机在山区电力设备上采集信息后，传统垂直小模型给出的判断：“销子不规范”，而大模型能够基于多模态发展出图像认知能力，给出的结论是：“高速公路附近上空，红色涂装的绝缘子左侧连接杆塔金件上，有10个螺栓，其中3个存在销子不规范，包括1个脱销、1个未插紧、1个损毁，已生成异常说明，建议尽快现场确认发起检修。” 数字化转型网（www.szhzxw.cn）
二、AI大模型赋能制造业的四个基本趋势
在“软件定义一切”的时代，AI大模型作为新的生产力工具，必将从内容领域（文生文、文生图等）深度扩张到生产实体领域，在制造业的各个环节中引发新的效率革命，加速制造业走向智能化。
1. AI驱动软件升级是大模型赋能制造业的主要途径
工业软件是制造业数字化转型的灵魂和关键。
AI大模型如何支持赋能制造业，有多种方式和途径，可预期的重要方式是：AI将重构软件开发模式、交互方式、使用流程和商业模式，无论是研发类、管理类、生产类还是后服务类工业软件，都将用大模型重新升级一遍，越是复杂的软件系统，未来改造的空间越大。
基于代码大模型打造的新一代AI编码平台产品，具备强大的代码理解与生成能力，支持代码补全、测试单元生成、代码解释、代码查错等核心场景。随着MaaS（模型即服务）的崛起，以模型为中心的开发范式将降低工业软件开发门槛，提高开发效率。 数字化转型网（www.szhzxw.cn）
根据CSDN在2023年初的评估，GPT4的软件编程能力相当于国内月薪3万元的软件工程师能力，相当于谷歌年薪18万美元L3级工程师。美国一个软件岗位招聘做了一个测试：一个只有4年编程经验的工程师借助AI工具，其软件开发效率相当于19年编程经验的5倍。
在工业软件开发层面，AI大模型正在革新软件开发范式。AI将与人类共同协作开发，倍数级提升软件研发的效率，例如服务于一线研发人员的内容生成工具（文档、编码、测试、发布、运维），可以大幅提升生产力。同时“代码大模型”的研究和应用，正在引发AI编码的革命。
AI成为芯片设计新工具，AI与EDA的双向奔赴，将开启芯片设计的下一场革命，Synopsys和Cadence等传统芯片设计公司也在积极拥抱AI设计。
英伟达Hopper架构H100拥有13000个AI设计电路，用AI设计GPU比传统EDA减少25%芯片面积，功耗更低。谷歌开始使用强化学习（RL）技术设计自己的TPUAI加速器布局。
在工业软件性能层面，AI大模型会推动软件智能升级。例如在研发设计场景中，Back2CAD基于ChatGPT等推出CADGPT™，支持智能推荐、文档生成、代码生产等功能，能够有效辅助产品的研发设计。
2. 弥合数据流断点是AI大模型赋能制造业的重要价值
每一次人机交互技术的突破，都将带来一次产业重构。AI大模型带来了新的“人机交互”革命，未来自然语言将能操控一切，深刻改变人们使用搜索引擎、消费购物、生产制造等的方式，并深刻影响未来的产业竞争格局。 数字化转型网（www.szhzxw.cn）
制造业数字化的核心是，以数据的自动流动化解复杂系统的不确定性，将正确的数据、以正确的方式、在正确的时间传递给正确的人和机器，提高资源配置效率。
但企业实际的运营状态是：多个环节中存在数据流的断点，需要工程师开发各种工艺软件和流程软件。AI大模型为改变这一现状找到了新路。
这条新路是，基于AI大模型的自然语言交互能力，为制造业企业内部、产业上下游之间的实时、泛在的连接提供了软件开发、交互的新方式，降低了工艺和流程的软件开发门槛、提高了效率，弥合了企业数据流动过程中的无数个断点。
例如国内机器人公司，借助通义大模型开发机器人行业模型，基于自然语言，可以实现人和机器的互动。如机器人收到了人的指令后，可以进行理解、推理和分析，并自动生成软件代码，组织协调不同智能体完成不同场景下的任务。 数字化转型网（www.szhzxw.cn）
这一功能大大降低了工艺开发人员的门槛，提高了开发效率和质量。从全局来看，不仅能避免出现数据断点，减少人工干预带来的影响，从而提高产品的稳定性和可靠性，促进了数据在多个环节的自动流动，提高了整个系统的智能化水平。
进入数字时代，以往高度一体化、集中化的制造业体系，逐渐走向生产分散化和组织灵活化。
AI大模型+智能协同办公平台，有助于打通制造业的一个个数据断流节点，推动数据在研发、生产、配送、服务等环节高效流动，从而提升制造企业内部、甚至产业上下游之间的协同效率，推动制造业走向“智能协同生产”。
“融合”是半个世纪以来技术演进的基本规律，信息技术（IT）、通信技术（CI）、控制技术（OT）和以云计算、AI为代表的DT技术加速融合。 数字化转型网（www.szhzxw.cn）
展望未来10年，AI大模型将会赋能每个智能终端、智能单元和智能系统，AI大模型驱动的智能在云边端实时协同成为基本趋势，被AI大模型赋能的智能体将无所不在，设备、产线、工厂、企业中的智能体将无所不在，数据流的核心价值将从描述信息走向决策流和控制流。
无数个智能体在AI大模型的驱动下，实现决策智能与控制执行，走向自决策、自控制，人们将面对一个智能联合体的崛起。
3. 进入控制环节是AI大模型赋能制造业的关键标志
AI大模型进入制造业的核心价值不是在营销和管理等环节，而是要进入生产控制环节。
AI大模型的通用性、泛化性，以及基于“预训练+精调”的新开发范式，将从研发设计、生产工艺、运维质控、销售客服、组织协同等各个环节赋能制造业。
其中，我们认为进入生产环节最核心的控制系统，例如PLC、MES、SCADA等等，提升工艺生产流程的智能化，是AI大模型应用制造业的关键标志。
西门子和微软在今年4月宣布合作，基于GPT推动下一代自动化技术变革，合作开发PLC的代码生成工具，将AI大模型融入控制环节。
目前，在电力调度领域，AI大模型可以深入新型电力系统复杂调度控制核心业务环节，成为调度业务“专家助手”，可以为电力调度员提供电网调控策略，优化线路负载均衡，从而降低电网损耗等。
目前，企业正探索利用AI大模型能力，驱动工业软件SCADA智能化。
SCADA系统（数据采集与监视控制系统）可以应用于电力、冶金、石油、化工、燃气、铁路等领域的数据采集与监视控制以及过程控制等诸多领域。 数字化转型网（www.szhzxw.cn）
在SCADA场景下，典型做法是利用大模型在特定行业场景下的编程接口和生态库，产生工业逻辑代码（交互、建模、SQL开发），自动集成到工业软件中，基于结果闭环优化模型。
在汽车行业，近几十年来，汽车工业的转型，不仅是一场动力革命，也是一次控制革命。
传统汽车向智能汽车演进最大的技术变革在于汽车控制系统的创新，从传统汽车80多个ECU等电子控制单元，转向类似于智能手机的集中式架构（底层操作系统+芯片SOC+应用软件）。
今天，自动驾驶成为汽车工业转型的又一个重大方向。
目前大模型对自动驾驶的改变主要有两个方向：一是大模型作为赋能工具，辅助自动驾驶算法的训练和优化；二是大模型进入决策控制环节，作为“控制者”直接驾驶车辆。
2023年8月公开报道显示，特斯拉“端到端”AI自动驾驶系统FSD Beta V12首次公开亮相，完全依靠车载摄像头和神经网络来识别道路和交通情况，并做出相应的决策。
当然，目前AI大模型进入控制环节，实际的应用和落地过程仍然面临着许多问题，有待科研人员进一步探索解决。
4. 大小模型协同是AI大模型赋能制造业的重要趋势
AI大模型本身需要找到具体落地场景，离解决千行百业的实际场景问题，还有距离。从实际的产业发展看，一个重要的趋势是：通用与专用、开源与闭源、大模型与现存软硬件系统的协同配合，是产业落地的必经阶段，而且在这一阶段，大小模型高度协同的重要载体——AI智能体（AI Agent）将成为新的生产工具。 数字化转型网（www.szhzxw.cn）
AI Agent一般是指基于LLM、能够使用工具自主完成特定任务的智能体。AI Agent将LLM与其他模型、软件等外部工具协同，能够处理真实世界中的各种复杂任务。
2023年7月，阿里云推出了一款智能工具魔搭GPT（ModelScopeGPT），它能接收用户指令，通过“中枢模型”一键调用魔搭社区其他的AI模型，大小模型协同完成复杂任务，降低大模型使用门槛。
未来，AI Agent将主要由“感知系统+控制系统+执行系统”组成，不仅具有生成能力，还将同时具备任务理解、任务拆解、任务调度、执行规划、链条协同等能力。
其中LLM将主要承担指挥中心角色，类似人类“大脑”的角色，对接入AI Agent的数字化工具（比如SaaS软件、工业机器人、数字人等）进行统一智能调度管理，实时在生产、管理、服务等场景中，由不同组合的数字化工具协同完成具体场景中的实际问题。
三、打造“公共云+AI”体系化能力，推动智能制造迈向“新阶段”
今天的制造业转型升级，已经不再是单一技术的赋能，而是以“公共云+AI”为代表的技术体系的全方位赋能和支撑。
当前，必须把握好以AI大模型为代表的新一代人工智能技术发展的历史机遇，加速推动智能制造迈向“新阶段”。
1. 实施“公共云优先”战略，把公共云作为推动“制造业+AI大模型”融合创新的关键力量
公共云的大规模、高可用、低成本算力基础设施，成为产业智能化的关键基础。
特别是美国升级芯片管制后，公共云是缓解高端芯片瓶颈的最优路径，通过高效连接异构计算资源，突破单一性能芯片瓶颈，协同完成大规模智能计算任务，可以有效降低对海外高端芯片的依赖。
一是要将“公共云优先”战略作为制造业数字化转型相关政策规划的重要内容，明确中长期发展目标、重点任务和保障措施等； 数字化转型网（www.szhzxw.cn）
二是要避免芯片“挤兑”现象，警惕各地“小散多”一哄而上地建设算力中心，造成统一算力市场得“碎片化”，避免出现建得多、用不好、用不起的现象；
三是将数据中心利用效率作为数据中心建设考核指标，扭转数据中心建设“重建设、轻运营”“重投入、轻绩效”的模式。
2. 鼓励模型开源开放，支持科技平台企业做大做强模型开源社区，繁荣AI产业技术生态
AI的竞争既是一场技术战，也是一场商业战，核心是生态战，关键在于开源开放。开源开放可以降低研发成本和应用门槛，是创新到商业闭环的“助推器”。
一是做好AI开源开放生态的顶层设计，将AI开源开放生态建设纳入国家规划、抓好落地实施；
二是鼓励地方政府联合AI开源社区头部平台建设AI赋能中心，依托海量开源模型和模型即服务平台（MaaS平台）加速制造业数智化创新应用；
三是鼓励应用牵引，加快产业落地，支持制造企业加速应用基础大模型、研发应用行业模型和企业专属模型，通过“用模型”反哺技术创新。
3. 启动工业软件AI驱动升级工程，加快制造业全环节全链条智能化升级
作为智能制造的关键支撑，工业软件对推动制造业转型升级具有重要战略意义。AI时代，所有工业软件都值得用大模型重新升级一遍。 数字化转型网（www.szhzxw.cn）
一是要大力发展基于AI的工业软件，推动“工业软件+AI大模型”技术研发，增强工业软件在智能化时代的自主创新能力，积极推动工业软件标准研制工作；
二是要充分发挥工业软件相关联盟的沟通桥梁作用，发挥AI企业、工业软件企业、科研院所和制造业企业各自优势，构建合作共赢、具有核心竞争力的AI驱动的工业软件产业生态。
4. 聚焦制造业重点产业链，分环节分场景打造标杆，示范推动大模型在制造业的规模化应用
制造业重点产业链是加快建设现代化产业体系的重要支撑，要找准关键环节、集中优质资源，搭建以“算力+算法+数据”为核心的要素体系，提升制造业的数实融合程度，促进制造业产业链安全和智能化升级。
一是启动大模型支撑新型工业化示范工程，以AI大模型为抓手，推进AI全方位、深层次赋能新型工业化，加快探索新型工业化“新模式”；
二是在产业基础好、创新能力强的制造业产业带、优势开发区、产业园区的等，率先开展“制造业+AI大模型”融合创新发展示范工程； 数字化转型网（www.szhzxw.cn）
三是通过“创新平台+数字工厂”等模式，针对感知、控制、决策、执行等关键环节的短板弱项，分场景加强产学研用联合创新，打造创新应用标杆，推动大模型规模化应用。
翻译：
An Xiaopeng: The manufacturing industry is the main battlefield for the application of AI large models
How to understand the depth and breadth of the impact of this round of AI technology development? How to judge the nature of this round of AI technological change?
AI large model is an important milestone in the development of general artificial intelligence. On April 28 and May 5, the Central Meeting put forward three firsts on the development of artificial intelligence: the first proposed “general artificial intelligence”, the first proposed “industrial intelligence”, and the first proposed “grasping the wave of new scientific and technological revolution such as artificial intelligence”.
The “revolutionary” feature of this round of AI is not that AI can generate text and pictures, but that AI can generate code, build a new model of human-computer interaction, combine with manufacturing links and scenes such as product research and development, process design, production operations, and product operations, improve production efficiency, form new productivity, and trigger the restructuring of industrial competition. 数字化转型网（www.szhzxw.cn）
AI large model is related to the transformation of human production tools, the reshaping of the core competitiveness of a country’s manufacturing industry, and the long-term prosperity and direction of the economy.
First, the integration of data and reality is the core variable of the restructuring of the global manufacturing competition pattern
Digitalization is a watershed in an era of great change and has become a catalyst for the sharp divergence of competition among companies, cities, and countries. Manufacturing industry is the most important industrial sector of digital integration, and the way, breadth and depth of its integration can directly affect or even determine the advanced level of manufacturing industry and the global competition pattern.
The integration of data and reality is the fundamental reason why the United States leads the world in manufacturing
At present, many people subconsciously very much agree with the development path and model of German manufacturing industry in the digital era, and “sing down” the American manufacturing industry. In fact, over the past decade, the United States has been a model room for global manufacturing development. 数字化转型网（www.szhzxw.cn）
Whether it is the size, growth or competitiveness of the manufacturing industry, the United States continues to lead Germany and Japan, and the gap is widening: from 2011 to 2021, the size of the manufacturing industry in the United States is 1.5 times that of Germany to 2.4 times, and is 2.5 times that of Japan to 3.1 times. American manufacturing is still growing at a faster pace than Japan and Germany.
The consolidation and establishment of U.S. manufacturing leadership over the past decade is the result of “software-defined hardware,” the deep and comprehensive integration of digital technologies represented by “cloud +AI” into the real economy, and the migration to new digital infrastructure.
The emergence of digital native enterprises is an important symbol of the upgrading of American manufacturing
The emergence of digital native enterprises is an important symbol of industrial upgrading and economic prosperity.
The widening of the manufacturing gap between Germany, Japan and the United States is mainly manifested in the micro lack of a number of competitive digital native enterprises in the manufacturing field.
Germany’s “Industry 4.0” goal has not been achieved, and there is a large gap with expectations, and the digital progress of smes is slow, research shows that only 21% of smes use digital technology in production, and there is no growth of a number of competitive smes in the digital era.
The situation is similar in Japan, which experienced two “lost decades” and also failed to develop a number of digital-native enterprises.
Digital technology drives the emergence of new enterprises and new products in the United States, with Tesla, SpaceX, Rivian, OpenAI, Snowflake, Palantir and other digital native enterprises as representatives, not only become the global industry leader, but also continue to build a new model of product innovation. 数字化转型网（www.szhzxw.cn）
As Valentine described in “From Toyoda to Tesla”, digital native enterprises in the field of electric vehicles are redefining the car, in the personalized, user-centered Internet of everything, everything intelligent era, new species are born to deal with uncertainty, driven by high-frequency innovation. Iterative thinking based on evolutionary organization accelerates commercialization from MVP baby run.
As a digital native enterprise, Tesla has four typical characteristics: software defined, high-frequency innovation, customer operator, and evolutionary organization.
In 2010, DARPA launched the Adaptive Vehicle Manufacturing (AVM) program, proposing to “re-invent manufacturing”, and shorten the development cycle of weapons and equipment to one-fifth of the present by revolutionizing and reshaping the equipment manufacturing industry.
Complex manufacturing product development and production should be like the semiconductor industry, its product design, simulation, testing, process, manufacturing and other activities are all completed in the digital space, and then enter the factory once the product iteration is mature, shortening the development cycle and reducing the development cost. The strategy is already bearing fruit.
On November 11, 2023, the United States first flew a new generation of stealth strategic bomber, which is the world’s first digital bomber -B21, from the beginning using digital design, cloud-based development, deployment and testing of digital twins, bringing better maintenance and longer life cycle. As well as lower infrastructure costs, it is the fastest model developed by the US military in the last 30 years, and it can constantly upgrade product functions like Tesla cars continue to download new software, and the combat upgrade will rely on agile software iteration. 数字化转型网（www.szhzxw.cn）
The essence of this trend is that “cloud +AI” is not only a business infrastructure, but also an innovation infrastructure, which is the cradle of new enterprises and new products.
AI grand model is a new starting point for reshaping the global manufacturing competition pattern
The AI grand model is accelerating the third wave of “digital-real integration”, and intelligence is its main feature. AI big model will affect the development pattern of the manufacturing industry, and AI big model will be integrated into all aspects of the research and development design, production process, quality management, operation control, marketing services, organizational collaboration and management of the manufacturing industry.
In the field of research and development design, AI revolutionizes traditional research paradigms.
In the biomedical field, in 2022, the AlphaFold2 model developed by DeepMind predicted almost all protein structures. Today, AI models can not only “predict” but also “generate” proteins, creating new possibilities for future drug production and research. For example, the ProGen system, developed by Salesforce Research, successfully generates entirely new proteins from zero.
The computational efficiency of RNA virus research is low and inaccurate. The “LucaProt” deep learning model based on Transformer architecture of Sun Yat-sen University has trained a large protein language model, and the virus discovery cycle has been shortened from the past 2-3 months to one week. The discovery of tens of thousands of novel viruses that could not be identified by traditional manual comparison methods has increased the global RNA virus diversity by nearly 30 times. This will shorten the vaccine development cycle and reduce the cost of development.
In the manufacturing process, AI large models can directly serve the R & D and innovation of intelligent cars, robots, chips, clothing and other products, such as engineers can automatically generate code instructions through large models, complete the development and debugging of robot functions, and even create some new functions for robots. 数字化转型网（www.szhzxw.cn）
In the field of equipment operation and maintenance, AI large models have greatly enhanced the capabilities of traditional vertical models. The AI large model has the understanding ability. After the UAV in the power industry collects information on the power equipment in the mountain area, the traditional vertical small model gives the judgment: “the dowel is not standardized”, and the large model can develop the image cognition ability based on multiple modes. The conclusion is: “In the air near the highway, there are 10 bolts on the left side of the red coated insulator connecting the gold piece of the pole tower, and 3 of them have non-standard pins, including 1 out of stock, 1 not firmly inserted, and 1 damaged. Abnormal instructions have been generated, and it is recommended to confirm the site and launch maintenance as soon as possible.”
Second, the four basic trends of AI large model enabling manufacturing
In the era of “software defines everything”, AI large models, as a new productivity tool, will expand deeply from the content field (Wensheng, Wensheng chart, etc.) to the production entity field, trigger a new efficiency revolution in all aspects of manufacturing, and accelerate the manufacturing industry to intelligence.
Ai-driven software upgrade is the main way for large models to empower manufacturing
Industrial software is the soul and key of the digital transformation of manufacturing industry.
There are many ways and ways to support the enabling manufacturing industry, and the important way to be expected is that AI will reconstruct the software development mode, interaction mode, use process and business model, whether it is R & D, management, production or post-service industrial software, it will be re-upgraded with the large model, the more complex the software system, the greater the space for future transformation. 数字化转型网（www.szhzxw.cn）
A new generation of AI coding platform products based on the code model, with strong code understanding and generation capabilities, support code completion, test unit generation, code interpretation, code error checking and other core scenarios. With the rise of MaaS (Model as a Service), the model-centric development paradigm will lower the threshold of industrial software development and improve development efficiency.
According to CSDN’s assessment in early 2023, GPT4’s software programming ability is equivalent to the ability of a software engineer with a monthly salary of 30,000 yuan in China, which is equivalent to a L3 engineer with an annual salary of 180,000 US dollars in Google. A software job recruitment in the United States did a test: an engineer with only 4 years of programming experience with AI tools, its software development efficiency is equivalent to 19 years of programming experience 5 times.
At the level of industrial software development, AI grand models are revolutionizing the software development paradigm. AI will collaborate with humans to improve the efficiency of software development, such as content generation tools (documentation, coding, testing, publishing, operation and maintenance) for front-line developers, which can greatly improve productivity. At the same time, the research and application of “code large model” is triggering a revolution in AI coding.
AI has become a new tool for chip design, and the two-way flow of AI and EDA will open the next revolution in chip design, and traditional chip design companies such as Synopsys and Cadence are also actively embracing AI design. 数字化转型网（www.szhzxw.cn）
The Nvidia Hopper architecture H100 has 13,000 AI-designed circuits, and the GPU designed with AI reduces the chip area by 25% compared to traditional EDA, with lower power consumption. Google began designing its own TPUAI accelerator layout using reinforcement learning (RL) techniques.
At the level of industrial software performance, AI large models will promote the intelligent upgrade of software. For example, in the R & D design scenario, Back2CAD launched CADGPT™ based on ChatGPT, etc., supporting intelligent recommendation, document generation, code production and other functions, which can effectively assist the R & D design of products.
Bridging data flow breakpoints is an important value of AI large model enabling manufacturing
Every breakthrough in human-computer interaction technology will bring about an industrial restructuring. AI has brought a new “human-computer interaction” revolution, and natural language will be able to control everything in the future, profoundly changing the way people use search engines, consumer shopping, production and manufacturing, and profoundly affecting the future industrial competition pattern. 数字化转型网（www.szhzxw.cn）
The core of manufacturing digitization is to resolve the uncertainty of complex systems with the automatic flow of data, pass the right data, in the right way, at the right time to the right people and machines, and improve the efficiency of resource allocation.
However, the actual operation state of the enterprise is that there are data flow breakpoints in many links, and engineers need to develop various process software and process software. AI grand models have found a new way to change this status quo.
This new road is that the natural language interaction capability based on the AI large model provides a new way of software development and interaction for the real-time and ubiquitous connection within the manufacturing enterprise and between the upstream and downstream of the industry, reduces the software development threshold of the process and process, improves the efficiency, and Bridges countless breakpoints in the process of enterprise data flow.
For example, domestic robot companies develop robot industry models with the help of Tongyi large models, which can realize the interaction between humans and machines based on natural language. For example, after the robot receives human instructions, it can understand, reason and analyze, and automatically generate software code to organize and coordinate different agents to complete tasks in different scenarios. 数字化转型网（www.szhzxw.cn）
This feature greatly lowers the threshold for process developers and improves development efficiency and quality. From a global point of view, it can not only avoid data breakpoints and reduce the impact of manual intervention, so as to improve the stability and reliability of the product, promote the automatic flow of data in multiple links, and improve the intelligence level of the entire system.
Entering the digital age, the highly integrated and centralized manufacturing system in the past is gradually moving toward decentralized production and flexible organization.
The AI large model + intelligent collaborative office platform helps to open up a data cut-off node in the manufacturing industry, promote the efficient flow of data in research and development, production, distribution, service and other links, so as to improve the collaborative efficiency within manufacturing enterprises, and even between the upstream and downstream of the industry, and promote the manufacturing industry to “intelligent collaborative production”.
“Convergence” has been the basic law of technological evolution for half a century, and information technology (IT), communication technology (CI), control technology (OT) and DT technologies represented by cloud computing and AI have accelerated integration.
Looking forward to the next 10 years, AI large model will empower every intelligent terminal, intelligent unit and intelligent system, AI large model driven intelligence in the cloud side real-time collaboration has become a basic trend, AI large model enabled agents will be everywhere, equipment, production lines, factories, enterprises in the intelligent agent will be everywhere. The core value of data flow will move from descriptive information to decision flow and control flow.
Driven by the large AI model, countless agents achieve decision-making intelligence and control execution, and move toward self-decision-making and self-control, and people will face the rise of an intelligent consortium.
Entering the control link is a key sign of AI large model enabling manufacturing
The core value of AI large models entering the manufacturing industry is not in marketing and management, but in production control.
The versatility and generalization of AI large models, as well as the new development paradigm based on “pre-training + fine-tuning”, will empower the manufacturing industry from all aspects of research and development design, production process, operation and maintenance quality control, sales customer service, organization and collaboration. 数字化转型网（www.szhzxw.cn）
Among them, we believe that the most core control system into the production link, such as PLC, MES, SCADA, etc., to improve the intelligence of the process production process is a key sign of the AI large model application manufacturing industry.
Siemens and Microsoft announced in April this year, based on GPT to promote the next generation of automation technology change, cooperation in the development of PLC code generation tools, AI large model into the control link.
At present, in the field of power dispatching, AI large models can go deep into the core business links of complex dispatching control of new power systems, become an “expert assistant” for dispatching business, and provide power grid regulation strategies for power dispatchers, optimize line load balancing, and thus reduce power grid losses.
At present, enterprises are exploring the use of AI large model capabilities to drive industrial software SCADA intelligence.
SCADA system (Data acquisition and monitoring control system) can be used in power, metallurgy, petroleum, chemical, gas, railway and other fields of data acquisition and monitoring control and process control. 数字化转型网（www.szhzxw.cn）
In the SCADA scenario, the typical practice is to use the programming interfaces and ecological libraries of large models in specific industry scenarios to generate industrial logic code (interaction, modeling, SQL development), which is automatically integrated into the industrial software, and the closed-loop optimization model based on the results.
In the automotive industry, the transformation of the automotive industry in recent decades is not only a power revolution, but also a control revolution.
The biggest technological change in the evolution of traditional cars to smart cars lies in the innovation of automobile control systems, from electronic control units such as more than 80 ECUs in traditional cars to centralized architectures similar to smartphones (underlying operating system + chip SOC+ application software).
Today, autonomous driving has become another major direction in the transformation of the automotive industry.
At present, the change of large model to automatic driving mainly has two directions: First, large model as an enabling tool to assist the training and optimization of automatic driving algorithm; Second, the large model enters the decision-making control link and drives the vehicle directly as the “controller”.
According to public reports in August 2023, Tesla’s “end-to-end” AI autonomous driving system FSD Beta V12 made its public debut, relying entirely on on-board cameras and neural networks to identify road and traffic conditions and make decisions accordingly. 数字化转型网（www.szhzxw.cn）
Of course, at present, the AI large model enters the control link, and the actual application and landing process still faces many problems, which need to be further explored and solved by researchers.
Large and small model collaboration is an important trend of AI large model enabling manufacturing industry
The AI large model itself needs to find specific landing scenes, and there is still a distance from solving the actual scene problems of thousands of lines and industries. From the actual industrial development point of view, an important trend is: general and special, open source and closed source, large model and existing hardware and software system coordination, is the only stage of industrial landing, and in this stage, the size of the model highly collaborative important carrier – AI Agent (AI Agent) will become a new production tool.
AI Agent generally refers to an agent based on LLM that can autonomously complete specific tasks using tools. AI Agents collaborate with other models, software and other external tools to handle a variety of complex tasks in the real world.
In July 2023, Ali Cloud launched an intelligent tool Magic GPT (ModelScopeGPT), which can receive user instructions, call other AI models in the magic community through the “central model” with one click, and collaborate with large models to complete complex tasks, reducing the threshold of large models.
In the future, AI Agent will mainly consist of “perception system + control system + execution system”, which not only has the ability of generation, but also has the ability of task understanding, task dismantling, task scheduling, execution planning, chain coordination and so on.
Among them, LLM will mainly play the role of command center, similar to the role of human “brain”, to carry out unified intelligent scheduling and management of digital tools connected to AI Agent (such as SaaS software, industrial robots, digital people, etc.), and cooperate with different combinations of digital tools to complete actual problems in specific scenes in real time in production, management, service and other scenarios. 数字化转型网（www.szhzxw.cn）
Third, build the “public cloud +AI” system capability to promote intelligent manufacturing to a “new stage”
Today’s manufacturing transformation and upgrading is no longer the empowerment of a single technology, but the all-round empowerment and support of the technical system represented by “public cloud +AI”.
At present, we must grasp the historical opportunity of the development of a new generation of artificial intelligence technology represented by AI large models, and accelerate the promotion of intelligent manufacturing to a “new stage”.
Implement the “public cloud priority” strategy, and take public cloud as a key force to promote the integrated innovation of “manufacturing +AI grand model”
The large-scale, high-availability, low-cost computing infrastructure of the public cloud has become the key basis for industrial intelligence. 数字化转型网（www.szhzxw.cn）
Especially after the upgrade of chip control in the United States, public cloud is the best path to alleviate the bottleneck of high-end chips, by efficiently connecting heterogeneous computing resources, breaking through the bottleneck of a single performance chip, and cooperating to complete large-scale intelligent computing tasks, it can effectively reduce the dependence on overseas high-end chips.
First, the “public cloud priority” strategy should be regarded as an important part of the policy planning related to the digital transformation of the manufacturing industry, and the medium and long-term development goals, key tasks and safeguard measures should be clarified.
The second is to avoid the chip “run” phenomenon, vigilant around the “small scattered” rush to build computing power center, resulting in the unified computing power market “fragmentation”, to avoid the phenomenon of building more, poor use, and can not afford to use;
The third is to use data center utilization efficiency as a data center construction assessment index, reversing the data center construction “heavy construction, light operation”, “heavy investment, light performance” model.
Encourage open source models, support technology platform enterprises to expand and strengthen the open source model community, and prosper the AI industry technology ecology
AI competition is not only a technical war, but also a commercial war, the core is an ecological war, the key is open source. Open source can reduce research and development costs and application thresholds, and is a “booster” of innovation to commercial closed loop.
First, do a good job of the top-level design of AI open source and open ecology, incorporate AI open source and open ecological construction into national planning, and do a good job of implementation;
Second, local governments are encouraged to build AI empowerment centers in conjunction with the AI open source community head platform, and accelerate the application of digital intelligence innovation in the manufacturing industry by relying on massive open source models and model as a service platform (MaaS platform); 数字化转型网（www.szhzxw.cn）
The third is to encourage the application of traction, accelerate the landing of the industry, support manufacturing enterprises to accelerate the application of basic large models, research and development of application industry models and enterprise-specific models, and feed technological innovation through “using models”.
Start the industrial software AI-driven upgrade project to accelerate the intelligent upgrade of the whole link and chain of the manufacturing industry
As a key support for intelligent manufacturing, industrial software has important strategic significance in promoting the transformation and upgrading of the manufacturing industry. In the age of AI, all industrial software is worth re-upgrading with a large model.
First, we should vigorously develop AI-based industrial software, promote the research and development of “industrial software +AI large model” technology, enhance the independent innovation ability of industrial software in the intelligent era, and actively promote the development of industrial software standards; 数字化转型网（www.szhzxw.cn）
Second, it is necessary to give full play to the role of communication bridge of industrial software related alliances, give play to the respective advantages of AI enterprises, industrial software enterprises, research institutes and manufacturing enterprises, and build an AI-driven industrial software industry ecology with win-win cooperation and core competitiveness.
Focus on the key industrial chain of the manufacturing industry, create benchmarks by link and scene, and demonstrate and promote the large-scale application of large models in the manufacturing industry
The key industrial chain of the manufacturing industry is an important support for accelerating the construction of a modern industrial system. It is necessary to identify key links, concentrate high-quality resources, build a factor system with “computing power + algorithm + data” as the core, improve the degree of digital integration of the manufacturing industry, and promote the safety and intelligent upgrading of the manufacturing industry chain.
First, start large models to support new industrialization demonstration projects, take AI large models as the starting point, promote AI all-round and deep-level enabling new industrialization, and accelerate the exploration of new industrialization “new models”; 数字化转型网（www.szhzxw.cn）
The second is to take the lead in carrying out the “manufacturing +AI big model” integrated innovation and development demonstration project in manufacturing industrial belts, advantageous development zones and industrial parks with good industrial foundation and strong innovation ability;
Third, through the “innovation platform + digital factory” and other models, aiming at the weaknesses of key links such as perception, control, decision-making, and execution, we will strengthen the joint innovation of production, university and research in different scenarios, create a benchmark for innovation application, and promote the large-scale application of large models.
本文由数字化转型网（www.szhzxw.cn）转载而成，来源于阿里研究院 作者：安筱鹏；编辑/翻译：数字化转型网小汤圆。
免责声明: 本网站(http://www.szhzxw.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。
