# (77) CEO Fireside Chat With Jensen Huang and Pascal Daloz at Dassault Systèmes 3DEXPERIENCE World - YouTube
- Transcript ID: `3e6dr82pmdu`
- Source tier: `youtube_auto`
- Original language: `en_or_source_language_unlabeled`
- Translation status: `agent_translated_reviewed` (63/63 chunks)
- Original source: https://www.youtube.com/watch?v=3e6dR82pmdU
- Raw evidence: `data/transcripts/youtube/3e6dR82pmdU.json`
> 中文是派生翻译，英文是留存的源文本。中文未复核时，以英文和原始音视频为准。
## 00:00:08
**中文**
帕斯卡：大家好。你们叫 Solid Workers 吗？——Hard Workers。——Hard Workers。那么，欢迎杰森上台。杰森，谢谢你。你能来总是令人高兴。不知道大家是否意识到，我们有着长期合作关系；我想双方差不多在 30 年前就开始合作了。——是的，超过四分之一个世纪。——还记得怎样开始的吗？——我们是在上一次计算平台革命期间开始的，也就是个人计算机革命，当时……
**English / Original**
Pascal. Hey everybody. >> Uh, >> are you called solid workers? >> Hard workers. >> Hard workers. >> So, welcome on stage, Jason. Jason, >> thank you. >> It's always a pleasure to have you. I don't know if people realize but we have a long-standing relationship right I think we almost started the collaboration 30 years ago >> a yeah over a quarter century ago >> you remember how it started >> well it was we started during the last computing platform revolution in fact um the personal computer re revolution and
## 00:00:48
**中文**
原先的 Unix 工作站正迁移至基于 Windows 的工作站。让我们得以合作的技术以 OpenGL 为基础。我们发明了一项名为 CGFX 的技术，它是 CUDA 的前身。——是的。——如今 OpenGL 已演变为 RTX，实现完全路径追踪和基于物理的渲染；CGFX 当然也演变成了 CUDA。现在，我们在重塑计算平台之际再次携手。我们所做的一切都发生在数字世界中。40 年前，达索系统革新了“虚拟孪生”理念，即用计算机来呈现物理世界。
**English / Original**
what used to be uh Unix workstations uh was migrating to Windowsbased workstations And the technology that made it possible for us to collaborate was based on OpenGL. And we invented a technology called CGFX which is the the precursor of CUDA. >> Yeah. >> OpenGL became RTX today fully path trace and physically based and CGFX of course became CUDA and here we are working together again as we reinvent the computing platform. You know everything that we do is in the digital world. 40 years ago. So revolutionized the idea of virtual twins. The the idea of a virtual
## 00:01:30
**中文**
虚拟孪生当然是为了在计算机中呈现物理世界。如今，我们将在一种彻底革新的计算机——AI 计算机——中，以大得多的规模呈现物理世界。这真是一个无比精彩的时代。——没错，这是一段非凡的旅程。正如你所说，我们正在进入新篇章，也就是所谓的“生成式经济”：以加速计算和人工智能为虚拟孪生注入力量。
**English / Original**
twin, of course, is to represent the physical world in a computer. And so now we're going to represent the physical world at a much much larger scale in a completely revolutionized computer, an AI computer. And so this is a really really fantastic fantastic time. >> You're right. It's an economable journey. And as you say, now we are entering into a new chapter. We are now what we call in the generative economy where you know we are pouring the virtual twin with accelerated artificial intelligence.
## 00:02:02
**中文**
从你的角度看，Jensen，全球产业此刻正在发生什么？——我们正在彻底重塑整个计算技术栈。上一代的设计采用结构化表征：每一处几何形状、每一种材料，实际上所有东西都由我们明确指定。现在将转向生成式计算模型，而在生成式计算模型的世界里，整个计算技术栈都在被重新发明。由于 AI 是每个行业的基础，它将成为基础设施，就像水曾成为基础设施一样。
**English / Original**
From your perspective Jensen, what is happening in the global industry right now? Well, we're reinventing the computing stack altogether. And as you know, in the last generation, uh the representation of the designs were structured representations, meaning we specified every geometry, we specified every material, we specified literally everything. Now it's going to be a generative computing model. And in the world of generative generative computing models, the entire computing stack is being reinvented. And because AI is foundational to every single industry, it is going to become an infrastructure. Just as uh water was
## 00:02:45
**中文**
电力和互联网都是基础设施，如今人工智能也将成为基础设施。我们增长得如此之快，是因为每个行业都需要建设它，每个国家都会由它驱动，几乎每个社会都会拥有它。这是一轮新工业化的开端，对你们尤其利好，因为达索系统正是呈现一切待建事物的引擎。过去我们也许三分之一时间用于数字设计、三分之二用于物理环节；未来很可能发生根本变化。
**English / Original**
infrastructure, electricity is infrastructure, internet was infrastructure, now artificial intelligence will be infrastructure. We're growing so fast because every single industry needs to build it. Every single country will be powered by it. And um literally every society will have it. And so this is the beginning of a new industrialization which is really fantastic for you because as you know do uh is the engine of the representation of everything that you want to build and in the future in fact you know in the past I would say that we spent a third of our time in design and digital maybe twothirds of the time in physical it is very likely
## 00:03:29
**中文**
未来，我们会把 100% 的时间花在数字环境中。即使完成设计、仿真与验证，仍须把产品与软件集成。因此，达索系统中的一切——无论 CATIA、SIMULIA、BIOVIA，还是 DELMIA、ENOVIA 等其他品牌——都将构建在 NVIDIA 平台之上。
**English / Original**
in the future We're going to spend 100% of the time in digital and even after we're done designing it, simulating it, validating it, we have to integrate it with software. And so everything that's inside the do systems, whether it's Katilla or Simoleia or Biovilla or let's see what what are the other VAS? We got we got >> Delmia we got Inovia and and listen all of those VAS are going to be built on top of Nvidia.
## 00:04:10
**中文**
四分之一个世纪前我们能预见这些吗？总之，设计、表征、全部仿真，乃至实际运营，未来都将由软件定义。从一双网球鞋到汽车，都将由软件定义；造车的机器人由软件定义；机器人协同造车的工厂由软件定义；汽车本身也由软件定义。因此，一切都将由软件定义。
**English / Original**
>> Did we know that a quarter century ago? And so so anyhow anyhow the the the the design this the representation all the simulation and and even the operations of it because everything will be software defined in the future. You know everything from a a pair of tennis shoes will be software defined in the future and so cars are software defined. Um the robots that build the cars are software defined. The the factories where the robots you know are orchestrated build the cars are software defined and the cars themselves are software defined. So everything will be software defined.
## 00:04:41
**中文**
一切都会在达索系统中得到呈现；我们将以虚拟孪生的方式设计并运营所有事物，首次真正实现你们的愿景。——是的。在继续之前，看看现场这群有点喧闹的观众。达索系统服务全球 4,500 万人、40 万家客户，以及超过 1,500 万名工程师和研究人员。这里可能汇聚了全球最大、或至少是最大的工程社群之一。
**English / Original**
Everything will be represented inside the sew and so we'll be designing everything um operating everything really as a virtual twin and realizing your vision for the first time. >> Yeah. You know before we go further if you look around the crowds you know at >> kind of a ruckus crowd. >> Yeah. At Das System, we work with 45 millions people around the world, 400,000 customers, more than 15 millions engineers, researchers. So I think we are you have here probably one of the if not the
## 00:05:23
**中文**
他们创造了我们日常身边超过一半的产品：机器人、无人机、飞机、汽车、医疗器械、药品、住宅、城市和工厂。这是一个了不起的社群，你不这么认为吗？我知道他们自己这么想。——当然，我们都是工程师。——你也是工程师吗？
**English / Original**
largest engineering community in the world. They do more than half of the products surrounding us every day. robots, you know, drones, planes, cars, medical devices, drugs, uh, home, city, factories. So, this is an amazing community. Don't you think so? I know they think so. >> Absolutely. Yeah, we're all engineers. >> You are an engineer?
## 00:05:57
**中文**
是的，当然，我现在仍是工程师，所以我属于这个社群。——没错。——如果让我重新来过，我会选择使用 SOLIDWORKS。——这正是世界模型重要的原因。对这个社群而言，成功不在于自动化；他们不是要把过去自动化，而是要发明未来。这也是我们宣布合作新篇章的原因：双方将共同把虚拟孪生工厂与 AI 工厂结合起来。
**English / Original**
>> Yeah. Yeah, sure. I'm still an engineer. So you you belong to this community. >> I am. >> If I was to start all over again, I chose to use Solid Works. >> You know, that's why the world model matters. In fact, because for this community, the success is not about automation. They don't want to automate the past. >> They want to invent the future. And this is the reason why we are announcing this new chapter in our partnership because together we are bringing the virtual twin factory with the AI factory.
## 00:06:38
**中文**
这对你们、对我们都是巨大机遇。为了证明其力量，Jensen 和我挑选了一些具体用例与大家分享，先从研究和工程谈起。——我记得，我们几乎每次合作都从计算平台开始。当个人计算机进入云端时，达索系统再次重塑了自己。
**English / Original**
This opportunity is really enormous for you guys and for us. So to prove the power of this right we have some concrete example. I think uh Jensen and I we have selected some use cases we want to share with you. Let's start with research and engineering first. >> Yeah. And so, you know, I remember that almost everything that we do together starts with the computing platform. And when when PCs went into the cloud, the so reinvented yourself again.
## 00:07:10
**中文**
如今从云扩展到 AI，我们又在重新发明自己。因此今天宣布的是一项大规模合作，是两家公司四分之一个世纪以来最大的合作。达索系统将集成 NVIDIA CUDA-X 加速库、用于物理 AI 和智能体 AI 的 NVIDIA AI，以及我们的数字孪生技术 NVIDIA Omniverse。这些库凝聚了我们四分之一个世纪的工作，现在要把它们融合起来。
**English / Original**
>> Uh when now we're extending from cloud to AI, we're reinventing again. And so this to today we're announcing a massive partnership. This is this is the the largest collaboration our two companies have ever had in over a quarter century. >> Y >> uh do the is going to integrate NVIDIA CUDA X acceleration libraries, Nvidia AI for physical AI and for agentic AI and NVIDIA omniverse our version of digital twin technologies. And so all of these libraries represent the bo our body of work over the quarter of a century. Now we're going to fuse these technologies
## 00:07:49
**中文**
把这些技术融入达索系统，使大家都能受益于加速计算和人工智能，以过去无法达到的百倍、千倍乃至很快达到的百万倍规模开展工作。过去需要预渲染或离线运行的仿真，如今将真正成为你们始终设想的实时虚拟孪生。
**English / Original**
into the so so that all of you will have to benefit of accelerated computing artificial intelligence and be able to work at a scale that's a hundred times a thousand times and very soon a million times greater than what you were able to do before. What used to be, you know, pre-rendered or what used to be offline simulations will now be literally the virtual twin vision that you've always always had along.
## 00:08:20
**中文**
一切都将实时完成：实时设计产品并在风洞中仿真；实时连接机器人，让它们在工厂里运作并制造产品。这些将在未来 5 到 10 年内发生，非常了不起。谈到这里，让我们从生命开始。生命也许是迄今最复杂的系统；想想生命世界中编码了多少知识。
**English / Original**
Everything will be done literally in real time. You know, we'll we'll design design uh products and simulate it in a wind tunnel in real time. uh we'll interconnect these robots and let them operate in a factory in real time and they'll be building your products literally in real time and all of this is going to be happening and you know in the next 5 10 years this going to be extraordinary >> speaking about this let's start with life I think life is the most complex uh system ever created right when you think about it how much knowledge is encoded In the living world with our
## 00:09:00
**中文**
借助虚拟孪生，我们既向生命学习，也理解生命，以便复制并规模化这些能力。这是可以实现的。——这是 NVIDIA AI 与 BIOVIA 的融合。——是的，BIOVIA。——我们稍后还会回来讨论。——这之所以可能，是因为我们拥有一种基础，也就是所谓的世界模型。
**English / Original**
virtual twins, you know, we are learning from life. We are also understanding it in order to replicate and to scale it. So this is possible. >> This is Nvidia AI integrating with bio. >> Yes, >> bio. Right. >> And we will come back to this. >> Yeah. >> But you know this is possible because I think we have this foundation. We call it the world model.
## 00:09:23
**中文**
世界模型扎根于生物学、物理学和材料科学。关键问题是：要为生命构建世界模型，需要怎样的计算？——首先、也是最重要的，是理解生命的语言。在物理设计中，设计始于想象，再用由人设计的几何和纹理等结构化信息来表达物体。但生命不同，它在人类之前就已存在。因此我们必须学习 DNA 的语言。
**English / Original**
>> Yeah. the world model where it's grounded in biologies in physics material sciences. So the key question I have for you is what does it take to compute a world model for life of life? Well the most important thing the first thing that we have to do is understand the language of life. >> Yeah. And so you know if of course in the world of of physical design the design started with your imagination and you represented that physical object using structured information geometries and textures that were uh designed by you. However life is different. Life existed before us. And so we have to go learn the language of DNA, learn the
## 00:10:05
**中文**
我们要学习蛋白质和细胞的语言，理解它们如何相互作用及各自性质。第一阶段就是学习生命的含义，我们正在攻克它。第二阶段则是生成：一旦学会某种事物的意义，就能在语言之间翻译，例如在人类语言与生物学语言之间互译，并把生物学语言解释成人能理解的语言。更进一步，我们可以生成用于药物的新蛋白质，或生成可用于药物的新化学物质。
**English / Original**
language of proteins and learn the language of cells and understand how they interact and its properties. That first stage of learning the meaning of life is what we are in the process of tackling. The second part of course is generative. Once you could learn something, learn the meaning of something, we can translate it between languages. We can translate between human language and the language of biology between the language of biology and interpret it so that we can understand it in human language. Beyond that you can now translate and generate gen generate new proteins that could be used for a drug or generate new chemicals that could be used used for a
## 00:10:44
**中文**
我们还可以生成新材料，使其更坚固、更耐热、更轻、更易制造、使用寿命更长。这些性质如今都开始触手可及，因此未来十年它很可能成为影响最深远的工程领域之一。——完全正确，而且已经在发生。比如大家熟悉的 Bel 集团，他们生产著名的 Babybel；其使命很简单：为数百万消费者生产更健康的食品，同时减少资源消耗。
**English / Original**
drug and then of course generate new uh materials that could be stronger uh be more heat resistant, lighter, easier to manufacture, last longer. All of those properties um are now kind of within our grasp and this is one of the reasons why this is likely going to be one of the most impactful areas of engineering in the next decade. Exactly. And it is already happening. In fact, we have a case you know the bell group you know them they do the famous baby bell right and their mission is very simple. They want to basically produce healthier foods for million of consumer but at the same time they want to consume less
## 00:11:25
**中文**
他们希望减少用水，并逐步用非乳制蛋白补充、至少部分替代乳蛋白。因此他们正在发明所谓的食品科学。过去单个产品要进行数百次物理试验；如今他们可以自动生成，而屏幕上展示的正是这一过程。
**English / Original**
water and they want to progressively change or at least complement the dairy protein with the non-dairy proteins. So that's the reason why they are inventing what we call the food science. Before you know hundred of physical tests for one single product now they they generate automatically and this is what you can see on the screen.
## 00:11:50
**中文**
他们利用虚拟孪生自动生成蛋白质，因为其背后由生物世界模型驱动。结果不仅是创新更快，也是决策可认证；当人的生命掌握在手中时，不能随意试错。这正是我们所做的。现在转向另一个话题，也是你刚才开始谈到的内容。
**English / Original**
They generate automatically the protein from the virtual twins because it's against powered by the biological world model. So the result it's not only faster innovations, it's also certified decisions because you cannot uh play when you are you have the life of the people in your hands. That's what we do. Now let's move to something else and you started to speak about it.
## 00:12:16
**中文**
屏幕上的例子正在改变相关工程师的日常工作。如今，你先定义规格并运行仿真，生成式体验便会自动探索可能性空间，找出最优解。虚拟孪生实际上可探索近乎无限的可能性。问题是：我们能计算无穷吗？——不能计算无穷，但可以想象无穷；这也正是代理模型和仿真模型如此重要的原因。
**English / Original**
>> You know you have seen on screen this is changing in fact the daily life of the engineers in the spa. You know now you define the specs you run your simulations and automatically the generative experience is producing and exploring in fact the space of possibilities and finding the optimum solutions for you. Actually the virtual twin is exploring an infinite of possibilities. So the question I have for you could we compute infinity >> we can't compute infinity but we can imagine infinity which is the reason why these surrogate and emulation models the
## 00:12:58
**中文**
仿真与人工智能的融合极其强大。八年前，我把这一理念引入科学计算和仿真：未来不仅会使用基于原理的仿真——其中方程和物理定律都已充分理解、明确表达，但仿真耗时过长——还会用人工智能的生成式预测方法加以增强。
**English / Original**
fusion of simulation and artificial intelligence is so powerful. Eight years ago I introduced the idea to scientific computing and simulations. the idea that in the future not only will we use principled simulations where the the equations the laws of physics are well well understood and well represented however the simulation time takes way too long. Why don't we augment that with generative methods of predicting the future using artificial intelligence?
## 00:13:28
**中文**
可以作这样一个类比：狗能在空中接住球，但它们并没有运行关于球反弹或弹性的物理仿真，而只是观察并预测球会到哪里，然后把它叼住。同样，AI 可以学会预测物理现象，准确判断材料如何碎裂、碰撞会产生什么结果；这些能力已经触手可及。我们有一项名为 PhysicsNeMo 的技术，本质上是一套感知物理规律的系统。
**English / Original**
It's a little bit like the analogy I would give. It's a little bit like um are dogs able to catch a ball out of the out of the air and yet they're not doing physics simulations of of balls bouncing or elastic nature of the ball. They're just literally watching us and predicting where it's going to go and they snatch it out of the air. And so the the idea that an AI could learn how to predict physics and learn how to predict very accurately how materials would crumble, what happens to a crash, those things, those capabilities are within grasp. We have a technology called physics Nemo. Physics Nemo is essentially a physicsaware
## 00:14:10
**中文**
它是一套 AI 模型仿真系统和 AI 框架，可创建由基于原理的仿真器训练、或与其协同工作的 AI 模型。因此它既扎根于物理定律，又能快一万倍地做出预测。如果一切本已实时运行，就能再以大一万倍的规模预测，而这还只是今天的水平。想象一下未来将走到哪里。仿真与模拟结合来辅助设计，将带来真正的革命。这里展示的正是客户 Lucid 的案例。
**English / Original**
AI model simulation system and AI framework that allows us to create these AI models that are either trained by principal simulators or work alongside principal simulators. So it's grounded in the laws of physics but able to predict 10,000 times faster. And now if everything is already running in real time then you can predict it 10,000 times greater scale. And that's just where we are right now. Imagine where we're going to be in the future. Uh the idea of simulation and emulation coming together to help you design is going to be really revolutionary. And again, this is exactly what you see here with a customer called Lucid. You know, we know
## 00:14:51
**中文**
Lucid 是全球最具创新力的汽车公司之一。他们把碰撞行为、空气动力学和整车性能提前嵌入汽车项目开发上游，因此工程师设计的不只是外形，也在设计行为，而我们还能对其进行认证。这正如你所说，是赋能设计师和工程师的愿景，同时也让业务人员能够为客户开发令人愉悦的体验。
**English / Original**
them. It's they are one of the most innovative car company in the world. And what do they do? In fact, they embed the crash behavior, the aerodynamics, the vehicle performance upstream early in the vehicle programs developments. So the engineers, they don't only design the shape, they design the behavior and we certify it. So this is exactly what you say. This is this John vision empowering designer and engineers and also unlocking the business people you know to develop the delightful experience for their customers.
## 00:15:26
**中文**
现在谈谈工厂。今天的工厂早已不只是物理资产，我想大家都同意：它同时由虚拟与现实组成。那么，物理 AI 究竟怎样用于工厂运营？——过去，人们先设计产品，再建设制造它的工厂。
**English / Original**
Now let's talk about factories. You start to touch a little bit this topic. The factories are not anymore today only a physical assets. I think we are all in agreement with this. It's made of virtual and real and same times. So let us know how physic AIs is really used to run the factories. Well the way that people used to think about designing products is they design the product and they build the factory.
## 00:15:54
**中文**
未来，你能设计和制造什么产品，很可能会受到所设计、建设的工厂的深刻影响。实际上现在已经如此：每座工厂显然都用 CAD 设计，但它还将在虚拟孪生中得到完整仿真和运营。在虚拟孪生中运行如此庞大的工厂极其复杂，因为工厂不是一个对象，而是由数百万个对象组成。我们还要仿真或模拟这些工厂将如何运行。
**English / Original**
In the future, it's very likely that the products that you are able to design and build will a lot be impacted by the factories you design and built. And so, it's very likely in the future, well, it will be it's in fact now that every single factory is designing CAD. That's obvious, but it will be simulated and operated completely inside a virtual twin. And operating a factory of these gigantic scales inside a virtual twin is extraordinarily complex. A factory is not just one object. It's millions of objects. And we want to also simulate or emulate how these factories
## 00:16:35
**中文**
我们要在真实世界投入运行前，安排好生产线、顺序和间距，组织其中的机器人，并运行机器人 AI，让这些 AI 机器人在工厂中搬运、装配、移动物品并保障安全。所有这些都会先在虚拟孪生中发生。因此，达索系统帮助人们设计、建造的对象未来将变得极为庞大：它们是对象系统、AI 系统和机器人系统，共同组成一座巨型工厂。
**English / Original**
will operate in the real world so that we could arrange the manufacturing lines properly, arrange it in the right sequence, um space it properly, organize the robots within it, run the robot AIs so that these AI robot robots could be operating inside the factory, manipulating things, assembling things, moving things, keeping things safe. All of this is going to happen inside a virtual twin. And so, you know, do the the products that do is going to help people build and design are going to become gigantic in the future. These are going to be systems of objects, systems
## 00:17:15
**中文**
这些都是高速计算机。——再举 Omron 的例子：他们并非只用虚拟孪生来可视化工厂，而是做得更多。他们设计的是所谓“软件定义工厂”。
**English / Original**
of AI, systems of robots, all coming together into a giant factory. This is exactly >> you're going to these are fast computers. What I'm saying? And again this is exactly what we do with Omron. You know you have seen on the screen they don't use a virtual twin only to visualize the factories. In fact they do much more. They engineer what we call a softwaredefed factory.
## 00:17:38
**中文**
没错。不同之处在于，他们从第一天起就设计工厂的自主化部分；不是等生产系统已经运行后再设法注入自主能力。因此，这些工厂会变得灵活得多、更具韧性，也更能自适应。
**English / Original**
>> That's right. >> And where where is the different is coming from is in fact they are designing the autonomous part day one. It's not something they come when the production system is already up and running and they try to infuse the autonomous part in it. So as a result those factories they become much more flexible, resilient you know and also adaptative.
## 00:18:02
**中文**
但还有另一类工厂，也就是你经常谈到的 AI 工厂。它们正在各地建设，极其复杂。建造并真正实现这类工厂需要什么？——我们显然正经历一场新的工业革命：一种能影响众多行业生产力的基础技术，正如能源、机械动力、电力以及互联网曾经做到的那样，如今人工智能也在发挥同样作用。
**English / Original**
But there is another kind of factories right and you you talk a lot about this the AI factories they are building everywhere they are extremely complex. What does it takes to make them or to build them and to make them a reality? Well, we're going through what what uh uh clearly is a new industrial revolution, a fundamental technology that impacts the productivity of many industries. That's why it's an industrial revolution. Just as energy did that, just as mechanical energy, power did that, just as electricity and of course the internet did all that, we're now seeing artificial intelligence
## 00:18:43
**中文**
为了实现它，我们需要将三个巨型产业工业化并真正扩大规模。第一是制造大量芯片，所以芯片工厂数量不断增加；其中还包括封装工厂，以制造各种半导体产品。第二是计算机工厂：芯片造好后进入另一座工厂，出来的是超级计算机；这些超级计算机再进入人工智能工厂。眼下，这三个过去彼此独立的产业都在高速增长。
**English / Original**
doing that. In order to make this possible, we need to industrialize and really scale three different giant industries. The first one, of course, is building a lot of chips, which is the reason why the number of chip factories are increasing. You're going to be you're involved in a whole bunch of chip factories. And so, chip factories and packaging factories just to make all these semiconductor products. The second is computer factories. Once the chips are done, it goes into into another factory. What comes out of that is a supercomput. Those supercomputers go into an artificial intelligence factories. Right now, as we're speaking, these three entirely what used to be three different industries are all
## 00:19:24
**中文**
它们快速增长，是为了建设智能基础设施并制造 AI。这些工厂极其复杂：一座吉瓦级 AI 工厂约需 500 亿美元，而全球正在建设几十吉瓦。这是规模巨大的基础设施建设，也是人类史上最大的工业基础设施建设。工厂内汇集的技术非同寻常，而且我们要确保第一次就能正常运行。因此，我们采用达索系统的 MBSE——基于模型的系统工程——产品。
**English / Original**
growing incredibly fast so that we could create the infrastructure for intelligence and manufacturing the the AIs. Well, these factories are incredibly complex. They're, you know, a gigawatt AI factory is about $50 billion dollars. And now we're building tens of gigawatts around the world. It's an enormous infrastructure buildout, the largest industrial infrastructure buildout in human history. And so the amount of technology that comes together inside these factories are extraordinary. And we want to make sure that they work the first time. And so the way we're doing it, we're using MBSC a do product mechanical what?
## 00:20:03
**中文**
基于模型的设计。——基于模型的设计与工程。我原本希望它能叫某个 VIA 品牌。——还没有。——好吧。投入这些系统的是装有大量超级计算机的巨型数据中心；所需能源达一吉瓦，属于史上最大的工厂，而且造价高昂。因此，我们在实际建设之前，先用 MBSE 设计、规划并仿真所有细节。
**English / Original**
>> Model based design. >> Modelbased design >> engineering. >> I wish it would have been something via >> not yet. Not yet. >> Okay. All right. So you know MBVIA that's it modelbased via. And then and then um uh so so what goes into them? What goes into these systems are giant data centers with lots of supercomputers. The amount of energy necessary of course a gigawatt largest factories ever and it costs so much money. So we design, we plan, we simulate everything in MBSSE before we build it.
## 00:20:44
**中文**
我们的预期是：甚至在动土前，就先在虚拟孪生中运行网络和超级计算机，这能节省大量时间和资金。此后，数据中心会持续由 AI 保持最佳状态：AI 调节性能、功率、温度和冷却。要成功做到这些就需要人工智能。其运营闭环意味着这些 AI 工厂的虚拟孪生将永久运行，不断训练和更新模型。由此也再次证明，虚拟孪生绝不只关乎 3D。
**English / Original**
And so uh our expectation is and we even run the network and run the supercomputers inside the virtual twin before we even break break ground. That allows us to save tons of time and tons of money. And over time, of course, this data center has an AI that keeps it optimal. It's AIs that modulate the performance, modulate the power, modulate the the temperature, modulate cooling. And in doing so, if you want to do so successfully, you need artificial intelligence. And so that operating loop, so we're going to have these digital these virtual twins of these AI factories running forever training and updating our models. So again, this is proving that the virtual twin is not only about 3D.
## 00:21:25
**中文**
正如你所说，它揭示架构、揭示底层系统，当然也大规模揭示知识。——它看起来很真实。——而且它就是真实的。它真实运行，并整合完整的工业物料清单：来自制造商和供应商的物料清单被汇总起来，所有内容都按真实物理关系集成。由此我们能得到非常清晰的清单，准确知道要采购什么、哪个零件与哪个零件集成，提前看出是否适配，并确切知道各类零件数量。
**English / Original**
>> It's as you say, it's about revealing the architecture, the revealing the system underneath and obviously revealing the knowledge at scale. >> And it looks real >> and it's real. >> Yeah, it looks real, right? It looks real. It operates really it integrates all you know industrial bill of material and so the bill of material comes in from manufacturers and suppliers. We integrate everything really physically and so we have a very clear list of bill of material. We know exactly what we're going to buy. We know which part is going to integrate with another. Uh we could see in advance whether something fits or doesn't fit. We know exactly how many parts. So the inventory bill of
## 00:22:04
**中文**
这台超级计算机、这座 AI 工厂的物料合计约为 500 亿美元。因此，一切以数字方式呈现，绝不能出错。——现在我可指望你把它卖给你的生态系统了，对吧？——当然。不过我们自己就是第一个客户。
**English / Original**
material of this supercomputer comes out. This AI factory comes out adds up to about $50 billion bill of material. And so um this is incredible. We have everything digitally and so you know no mistakes will be made. >> Now I'm counting on you to sell it to your ecosystem, right? >> Yeah. Absolutely. Well, you're we're the first customer.
## 00:22:22
**中文**
——你们是吗？——是的，我们是第一个客户。——接下来要讨论怎样让知识真正投入工作，这正是虚拟伙伴所做的。昨天我们现场演示了即将到来的能力。工程师花太多时间查找信息或做其他事情，而那些并不是工程工作。
**English / Original**
>> You are? >> Yeah, we're the first customer. >> You are the first customer. Now the the next topic I want to discuss with you is really how do we put the knowledge at work. So this is exactly what our virtual companion they do. We made a live demo yesterday just to showcase what is coming and you know engineers they spend too much time to look for informations or to do something else. This is not engineering.
## 00:22:54
**中文**
看，现在只需几秒：从一张草图出发，自动生成内容，从 2D 转成 3D，而且得到的是完整参数化模型，仿真也已就绪。这确实正在彻底改变人们的工作和整个工作流。现场有一个共同问题：这会取代工程师吗？——回答之前，AI 带来的一项真正革命性能力，是在不同的信息形式之间转换。
**English / Original**
So now look at it in few seconds you know you we started from a sketch we produce automatically the we moved from 2D to 3D it's a full parametric model simulation already I think this is really revolutioning everything uh people are doing and it's changing completely the workflows now there is a questions in this room and the question is this one do you think this will replace the engineers Well, before I answer that, the the um one of the one of the really revolutionary things that has made possible for us because of AI is the ability to go from an interchange
## 00:23:35
**中文**
具体说，是在结构化与非结构化信息之间转换。照片、录音或视频本质上都是非结构化信息，而我们希望把它们转成结构化表征。要从 2D 图像变成 3D，需要人工智能。一旦转成 3D，信息就是精确、可控且可交换的，我们还能增强和改进它；它进入了结构化数据库。反过来，从结构化变成非结构化显然很容易。因此，我们如今既有了信息表征，也有了使用它的能力。
**English / Original**
between structured information and unstructured information. Unstructured information is essentially a photograph or may it could be a recording, could be a video and we want to take that unstructured information and now represent it in a structured way. We need artificial intelligence to go from 2D to 3D image to to to 3D. And and once you have it in 3D, that information is precise. It's controllable. It's interchangeable, right? We can enhance it. We could improve it. And so it is com it's now in our structured database. We can go from structured to unstructured obviously very easily. And so now we have the representation, the ability to use
## 00:24:16
**中文**
我们可以用智能体和 AI 管理信息，增强设计流程。某个零件可以手工精细设计，也可以导入一张图像再加以修改，等等。这些智能体将成为我们的伙伴，因为我们将充当它们的管理者和架构师。
**English / Original**
agents and use AI to manage our information so that we can augment um the design process. We could decide that this part I'm going to design specifically by hand here. I'm going to take an image. I'm going to import it and then I'm going to modify it. So on so forth. Now these agents these agents are going to be companions of ours because they're they're going to we're going to essentially be their manager, their architect.
## 00:24:41
**中文**
我们会是管理者、架构师、设计师和创造者，并由大量智能体或伙伴协助完成不同任务。许多人以为，设计师人数和所用软件工具数量会因此减少；事实恰恰相反。我确信，未来每位设计师、每位 SOLIDWORKS 设计师都会拥有一支伙伴团队；你会训练它们、教给它们不同技能。
**English / Original**
We're going to be the manager, the architect, the design, the the, you know, the creator, and we're going to have a lot of agents or companions help us perform different tasks. Whereas most people think that that the number of designers therefore will be less than the past and and uh the the number of software tools that you will use will be less than the past is exactly the opposite. It is very likely and and and I'm I'm certain this is going to happen that every designer, every Solid Works designer, every designer in the future will have a team of companions and you've trained these companions and you've taught them different skills and
## 00:25:23
**中文**
你还会帮助它们彼此协调、相互协作并与你合作，而它们全都会使用达索系统的工具。因此，这些工具的用户将从生物人类扩展到人类加 AI，工具使用量将爆炸式增长。这当然会成为软件行业的巨大机遇。
**English / Original**
you've helped them coordinate and work with each other and work with you and they're all going to be using Disso tools. So the number of users of the soul tools is going to go from biological to biological as well as AI based. And so the number of tool use will ex will explode. And so this is um the software industry of course is going to be great for the software industry.
## 00:25:45
**中文**
这对所有设计师都很棒，因为有许多伙伴协助工作。比如，你和伙伴们一起工作，到了该喝鸡尾酒的时候——毕竟世界上某处总已经是下午 4 点半——你便让团队继续探索：探索这个方向、优化那个目标，并给我三个方案；针对另一组目标，给我十个方案。等你回来，所有选择都已准备好，随后再由你亲自精细调整。
**English / Original**
It's going to be great for all the designers because you have so many companions to help you do things. And what what would be really fantastic is that you're you're working with your companions and then you know it's time to it's time for cocktail and and um because it's solidly 4:30 somewhere and and and so you kick off your team to go explore all these different areas and you know I want you to explore this, I want you to explore that, I want you to optimize for these areas and give me three designs. I want to optimize for these years. You give me 10 designs. And when you come back, you have all those choices and then you can go in and fine-tune it specifically yourself
## 00:26:22
**中文**
因为你拥有结构化数据和 3D 数据，我认为重新构想设计与创造力的机会将对所有人具有革命性。这里还有一个案例：NIAR，即位于威奇托、专注航空业研究、测试与认证的国家航空研究机构。我们知道，飞机认证是一场噩梦。
**English / Original**
because you have the structure data, the 3D data. And so I think that the uh the opportunities to reinvent how you think about design and creativity is going to be completely revolutionary for everybody. That's also something uh we can showcase. In fact, you take NAR. NAR is a national institute for the aviation industry. They are based in Wittita and they focus on research testing but also certifying and we know that certifying an airplane you know it's a nightmare.
## 00:26:54
**中文**
认证需要三到五年，并要满足一万多项要求。借助虚拟伙伴，法规要求可以自动摄取，不必再人工阅读数百万页材料；合规性还能得到持续验证。这意味着从事后检查转向“设计即合规”。合规不再只是成本，反而可能成为竞争优势。我的问题是：怎样从语言模型迈向世界模型？
**English / Original**
It take between three to five years it's more than 10,000 requirement you need to be fulfilled. Now I think with the virtual companions what could we do? The recolleations could be automatically ingested, right, without having to read millions of pages and the conformity is constantly verified which means now we are moving from if you want um it's by design it's a compliance by design right and there is no and it's not anymore a cost it could become a competitive edge in fact so the question I have for you is how does it take to move from a language model to a world model.
## 00:27:32
**中文**
语言模型显然要理解语言的句法、词汇和结构，还要有品味——例如怎样组织段落更好——并设有护栏，知道哪些该谈、哪些应避免。世界模型则不是遵循品味和价值，而是必须服从物理定律、理解因果关系：推倒一块多米诺骨牌，相连或邻近的骨牌也会倒；它要理解先后关系、惯性、摩擦和重力。
**English / Original**
>> Well, the language model has to um uh obviously understand syntax and vocabulary and structure of language and and has taste. You know, what's a what's a better way to compose a paragraph and it has guard rails. What are the things it should talk about and things that it should avoid talking about? Um in the world of in a world model instead of taste and values, it has to obey the laws of physics. It has to understand causality that if if you tip over a domino, you know, all of the dominoes that are connected to it or nearby it will be tipped over. It has to understand what comes what comes after and what comes before. It understand inertia and friction, understands
## 00:28:13
**中文**
它还要理解接触等你在设计时自然理解的现象。我们必须把这种感知能力教给 AI；它不一定能由语言表达，世上全部语言也无法囊括。因此，我们要用物理定律、仿真和大量例子，让它学会物理规律。你还提到可制造性设计：如今它被整合进设计流程，而不是设计完成后再由另一个团队判断能否制造。
**English / Original**
gravity, of course. It understands um uh uh contact, you know, all of the things that that you understand as you're designing things. We have to teach the AI that sensibility. that's not necessarily captured in language and all the language in the world won't capture that and so we have to use laws of physics and simulation and a whole bunch of examples to teach it the laws of physics and then of course one of the things that you you mentioned uh design for manufacturability today is integrated into the design process instead of you coming up with the design and then another team decides whether it's manufacturable design for
## 00:28:54
**中文**
可制造性应当上移，也就是“左移”。实际上，我们希望把一切都左移。合规尤其困难，因为那是机器与社会、人与价值相遇之处。语言模型中的人类价值可以提前融入设计流程，让产品始终合规；世界模型让它始终服从物理定律；同时始终面向可制造性设计，并持续使用获批组件、供应商批准清单等。
**English / Original**
manufacturability is really upstreamed shift it left. We want to shift left basically everything. And so one of the things that are very hard as you mentioned compliance is hard because that's where machines meets society and humans. And so that language model where human values could be now integrated into or shifted left into the design process. And so you're constantly in compliance. You're constantly obeying the laws of physics because of the world model. You're constantly designing for manufacturability. You're constantly designing to use components that are approved with, you know, vendor approval list and whatever it is. So that by the
## 00:29:35
**中文**
这样，等设计完成时，它已经可以直接投入使用。希望大家现在更清楚这项合作意味着什么。达索系统与 NVIDIA 正共同交付“知识工厂”：以虚拟伙伴和加速 AI 计算为虚拟孪生赋能。这不仅是性能提升；正如你所说，它是加速，而更重要的是开启新的可能性。
**English / Original**
time that you're done with the design, it's good to go. Yeah. Now, I hope you have a better idea about what this partnership is about. And I think together with Nvidia, we are delivering the knowledge factory. and to power the virtual twins with our virtual companions with an accelerated AI computing. This is more than performance. I think it's as you say it's an acceleration but more importantly this is opening the new possibilities.
## 00:30:07
**中文**
无论公司规模大小、身处哪个行业，我们都帮助你认证决策，在错误选择变成昂贵代价前将其排除，并创造全新品类的解决方案——软件定义产品、软件定义工厂，以及广义的软件定义对象。更重要的是，还必须保护你们的知识。我们如何保护数百万软件用户的知识？
**English / Original**
Whatever the size of the company, whatever the industry you are in, I think we help you to certify your decision. We help you to eliminate the bad choices before they become expensive mistake. And we also help you to create new categories of solution. You call it the softwaredefined products, the softwaredefined factories, the softwaredefed objects at large. And more importantly, we need to protect your knowledge as well, right? So how do you this how do you think we can protect the knowledge of all the million of people using our software?
## 00:30:45
**中文**
在回答之前，我先说，今天宣布的合作确实非同寻常。未来要构建的许多事物，没有加速计算就根本不可能实现；没有实时仿真不可能实现；没有人工智能也不可能实现。
**English / Original**
Well, first of all, before I go there, I think that that um our partnership today and what we're announcing is ex is really genuinely extraordinary. Um the the type of things that that you will be building in the future are simply impossible without accelerated computing. It's impossible without real-time simulation. It's impossible with art artificial intelligence.
## 00:31:09
**中文**
与其只想怎样提高生产率，不如想想哪些事情从此成为可能。是的，生产率会提高，你会更高效，也能比过去更快完成工作，就像 PC、云和互联网曾为我们做到的那样。每次技术革命都让我们更具生产力。但这一次，当规模扩大百倍、千倍乃至百万倍时，你能做的是过去根本不可能之事。双方最初的合作始于计算机图形学。
**English / Original**
instead of thinking about what could be more productive. Yes, productivity is going to be enhanced and and you'll be you'll be a lot more productive and you'll be able to do things more quickly than than the past just as PCs did that for us, just as the cloud did that for us, just as the internet has done that for us. Every technology revolution had made us more profit, more pro more productive. However, this time the type of things that you'll be able to do when you think about the scale a 100 times, a thousand times, a million times, these are going to be things that are just simply impossible to do before. Now the our our partnership started with computer graphics and our computer
## 00:31:45
**中文**
如今计算机图形已实现完全光线追踪和基于物理的渲染，达到照片级真实并可实时运行。因此，双方合作的基础一直是 RTX 和计算机图形，但现在已扩展至 CUDA-X、AI 和 Omniverse。这些建立在加速计算和 NVIDIA GPU 之上的计算平台，将革新工具，继而革新你们的设计方式、能够设计的事物，最终革新公司的运营方式。
**English / Original**
graphics as you've seen uh become fully ray traced and physically based and it looks completely photorealistic and it's real time and and so the foundation of our partnership has always been RTX and computer graphics but we've now extended it to CUDA X we've extended it to AI and we've extended it to Omniverse. All of these computing platforms sitting on top of accelerated computing and vidious GPUs are going to revolutionize the tools and revolutionize therefore how you design and what you can design and ultimately how your companies operate.
## 00:32:20
**中文**
这是第一点。另一个非凡变化是，未来几乎所有工作都会把 AI 纳入闭环。谈到 AI 时，人们会强调“人类在环”，这很重要；但别忘了，你的 AI 伙伴也在环中，它会记住你喜欢怎样做以及你的各种偏好。
**English / Original**
And so I think that that's that's number one. The the other part that's extraordinary of course is that in the future almost everything that we we do will have AI in the loop. You know, when people think about about AI, they have humans in the loop. And that's important. But remember, you also now have your companion, your AI in the loop. That AI is going to remember how you'd like to do your preferences.
## 00:32:51
**中文**
因此，AI 会把你的技能、偏好、习惯和领域专长编码下来，成为陪伴在你身边的伙伴。它不会放在云端，也不会公开，因为它承载着你的专业知识。以我的收件箱为例，它在很大程度上记录了我 33 年积累的知识和专长；这些内容并非人人可见，当然也不开源，其中包含我的大量判断力与知识。未来，我会拥有持续与我合作的伙伴。老实说，真希望 33 年前就有它。
**English / Original**
And that AI therefore will codify your skills, codify your preferences, codify your habits, codify the domain expertise that you have. And that is your companion. And that companion sits with you. It's not going to be in the cloud, not going to be public because it captures your expertise. You know, if you look at my inbox, in a lot of ways, it has captured 33 years of my knowledge of my expertise. It's not available for everybody. It's not open sourced, of course, and it's it captures a lot of my sensibility, a lot my knowledge. In the future, I will have companions that are going to continue to work with me. I wish I had it 33 years ago to be honest
## 00:33:33
**中文**
现在，你们每个人都将拥有把自身知识和判断力编码下来的伙伴。最后一个问题：为什么你认为这项合作独一无二？你已经提到一些理由，但也在与我们领域的其他公司合作。我们共同做的事情为何特别？
**English / Original**
and and now all of you will have companions that codifies your knowledge, codifies your sensibility. Last word about why do you think this partnership is unique? You already say a few words but you are also partnering with other companies especially in our space. Why do you think what we are doing together is something unique?
## 00:33:55
**中文**
你们在虚拟孪生世界中的位置，以及最初开创这一切的愿景，是独特的。CATIA 永远是 CATIA，SOLIDWORKS 永远是 SOLIDWORKS，SIMULIA 永远是 SIMULIA；其他品牌也会继续存在，你们还会创造新的品牌，而它们都会构建在 NVIDIA 平台上——这是我最喜欢的部分。但更关键的是，你们所服务的生态系统、在座所有对达索系统产品充满热情的人，以及建立在这些产品之上的所有公司。
**English / Original**
Well, the so your your place in the world of virtual twins, your vision that started it all. Katillaa will always be Katillaa. Solid works will always be solid works. Simoleia will always be Simoleia. You know, all the other EAS will always be, you know, and and you'll come up with other other EAS and they'll all be built on top of Nvidia. That's the part that I like the best. But but the the ecosystem the ecosystem that you serve, the ecosystem and all of you here that are so passionate about the the so products um and and all of your companies that are built on top of the
## 00:34:36
**中文**
如今，达索系统产品将由加速计算和 AI 加速并放大，这才是最令人兴奋之处。它恰逢全球重新工业化、重新发明自身，并启动人类史上最大的工业基础设施建设。投入将达到数万亿美元、数十万亿美元；有估计称未来十年接近 100 万亿美元，约为 85 万亿美元。
**English / Original**
so products are now going to be accelerated and amplified by accelerated computing and AI. And that's really what's really exciting here. And it happened at at precisely the time when the world is reindustrializing, >> reinventing, >> starting the largest industrial infrastructure buildout in human history. Trillions of dollars, tens of trillions of dollars. You know, estimates have it close to a hundred trillion 85 trillion dollars in the next 10 years.
## 00:35:09
**中文**
所有这些都需要设计、仿真、验证和原型制作。由于一切将由软件定义、由 AI 驱动，它们也都需要虚拟孪生。因此我认为，达索系统 40 年前提出的愿景正在成为现实。——它正在成为现实。
**English / Original**
All of that needs to be designed, simulated, validated, right? prototyped and and and of course because everything has is going to be softwaredefined and everything will be AIdriven all of that needs to have virtual twins and so I think that the the uh the vision that do had 40 years ago >> is coming true. >> It's coming true.
## 00:35:35
**中文**
是的，而且就在此刻成为现实。这项合作让愿景真正落地。我很高兴能与你合作，Pascal。双方四分之一个世纪的伙伴关系对我意义重大。CATIA 当年把 NVIDIA 带入工业工作站领域；直到今天，CATIA 和 SOLIDWORKS 对我们仍非常亲切，也对所有人极为重要。若没有在座各位及你们出色的工作，我们工程师和科学家追求的许多成果都没有机会问世。
**English / Original**
>> Yeah. And it's coming true right now. Um and this partnership brings it to life. And so I'm delighted to be to be uh uh partnering with you, Pascal. And uh you know our quarter of a century partnership means a lot to me. I you know it really Katillaa brought us brought Nvidia into the industrial workstation world and today we still Katia and Solid Works are you know to us uh very very personal and really important to all of us and without all of you and the amazing work that you do uh many of the things that our our you know engineering and scientists uh
## 00:36:16
**中文**
因此，我要感谢大家所做的一切不可思议的工作，感谢 Pascal 的卓越合作，也感谢我在达索系统的所有朋友。谢谢大家。——谢谢。——现在你属于这个社群了。——当然。——务必记住……——我绝对是一名 Solid Worker。
**English / Original**
pursue uh wouldn't have an opportunity to come to life and so I want want to thank all of you for for all the incredible things that you do and Pascal for the great partnership and all of my friends at D. So, thank you for everything. >> Thank you. >> Thank you. >> Now, you you belong to this community. >> Oh, yeah. >> So, make sure >> I am definitely a solid worker.
## 00:36:50
**中文**
一定要再回来。如果你不能回来，就派你的虚拟孪生来。——不，我会回来；我的虚拟孪生留在家里工作就好。好了，各位保重。——大家保重。
**English / Original**
Make sure you come back. And if you are not coming back, send your virtual twin. >> No, I'm coming back. My virtual twin gets to stay home and work. All right, take care, guys. >> Take care, guys.