发言人1 10:08
Ladies and gentlemen, please welcome Chairman of taitra James Huang Gu Yu Shi gu xianheng Huan Ying Hu Mao xhuaa a zhifeng dong shizhan.
发言人1 10:30
Good morning everyone, welcome to Computex. I'm James Huang, the host of Computer tech. As the organizer of computers Hera is honored to host global technology leaders. Today is particularly special as we welcome first going to the computer text keynote stage for the very first time. Under the leadership of Chairman Liu, fcon has been undergoing a remarkable transformation, praising AI at the core of its growth strategy. From AI optimized manufacturing to smart electric vehicles, 8 data centers and age computing. Q When has redefined force-controlled not just as a manufacturer, but as a global enabler of intelligent manufacturer with engineering expertise and a clear vision for the future, he is driving force come to the forefront of the AI economy. But before Chairman Liu joins us, let's watch a quick video on first-come ISS AI vision.
发言人1 12:53
Ladies and gentlemen, please join me in warmly welcoming the chairman of firstcom, Mr Young Liu.
发言人1 13:10
Thank you, thank you James. Good morning everyone. It is a pleasure to be here with you today. You have seen so many factories with so many robots running in our current factories and these factories later I'm going to show it to you that it has changed. It's changed because of AI. So where's James? James is gone. Okay, anyway. I'd like to thank James for inviting me to be here to.
发言人1 14:01
For this, a very important event, the comp tax. To say something about fast counts endeavor in the AI. And after taking this mission I've been thinking about what should I deliver? So what can our customer, our audience will feel beneficial after hearing this speech? So I've been thinking about this. So after so many days of thinking, finally I thought I'm not going to talk to you about technologies because we have very technical experts. They are going to talk about technologies. I'm not going to share with you our business in AI because some of them, some of it are confidential. But why I decided to share with you is the train of thoughts, how fast can get to this point with our AI factories?
发言人1 15:16
So next, it actually all starts from here, about a year and a half ago when we had our hhtv 23. No, we sit in the meeting room in the hotel with Jensen and Jensen hand drawn this picture and look at this picture said, well, what does it mean? Okay, ever since these pictures, the spirit is there, but it's changed, evolved. So let's look at what we have done with a video that we prepared together with BCG. To show you what we have gone through in the last a year and a half, let's play the video.
发言人2 16:17
Fox, the leader in the manufacturing, highly automated and scaled worldwide. Now we're entering a new era powered by AI. We call it Genesis. With a rapid global expansion, expert production technicians are rising in demand, but seasoned talent is short in supply. Our AI agents capture expert knowledge in areas like defect resolution in equipment tuning, handling 80% of the work so humans can focus on the most complex 20%. These agents learn, adapt, and orchestrate production, improving cycle time over 10% while generating tokens that lay the foundation for a genic AI to scale expertise across factories.
发言人2 16:55
We're creating next generation robots that go beyond simple hands and eyes. Using the videos on the verse, we train robotic brains using millions of simulations, making them instantly effective in the real world. We call this physical AI a new form of intelligence that adapts and handles complex tasks with 0 human touch. Genesis is a powerful foundation combining innovative use cases, token repositories, and an enhanced tech stack to deploy AI at scale. Together with leading partners, we are transforming manufacturing. Foxconn Genesis The next era has already begun.
发言人1 17:33
Okay, so this is Genesis, and this is still a work in progress. So there are still things are evolving.
发言人1 17:48
Through this journey, we come up with many new ideas. And the very first new ideas is about the future factories. You will have one physical factories together with two other factories, which are very, very important. One we probably heard about. The digital twin, the Omniverse digital twin factories. So before you create your own, your physical factories, you can build your factory. On verse, practice it and optimize it through without the actual factory being built.
发言人1 18:40
In the past, we thought. There was Omniverse, that's already good enough. But you know, with the universe, it can already create data. And with that data, if we use it properly with the AI factory, then we may be able to create a base set of model that can be used when the actual factory is built. So we think in the future, the factory will be like this. With the physical factory will appear the last. The first one you will have will be the Omniverse Digital Twin factory. And then the AI factory will be the second one. And once you have the model build and the digital twin optimized to a certain level, then you can have your actual factory.
发言人1 19:58
And that's the vision that we see for the future factories. Next. What you have seen that with the BCG effort, we have gone through so many use cases. And these use cases, after accumulating many, many hours work with technicians. We have very important finding, which I would like to share with you.
发言人1 20:39
You this experiment or this study, we thought with gennai, we can maybe replace human. Every human. But very quickly, we realize, no, it will not.
发言人1 20:59
If you look at this chart, the wide curve is a human technician. How many times it will take the technician to try to reach to a certain level? And that's typical human. And with the blue and the green line, this is done with gennai's help. And you can tell with gennai very, very quickly. Now, with two to three times of trial, it can reach 80%. But after 80%, look at the curve, it flatten. Okay, so the human can do much better than the gennai after 80%. So we come up with this temporary conclusion that with gen AI, it can help for the 80% of the work. The rest of 20%, we still have to be done by the skillful workers or technicians.
发言人1 22:24
That's a very important finding I would like to share with you next.
发言人1 22:32
And because of the scale of fast cons's factories, we have central divisions and we also have many Bgs. And the Bgs, they have factories across the world. And how we can come up with an architecture that will be able to support this large scale and very complicated. Implementation and the model. And at the same time allows the model to evolve by itself. And it requires a very, very.
发言人1 23:19
The architecture becomes very, very important. We know it's still evolving, but we think with this initial architecture, we should be able to get to what we want so that the model, the base model created by the central department can then be utilized by the BG Bgs factory all over the world. And then they can continue to train with their factory actual data. And those data and the trend models can then feed back to the central model. And the model in the central can then share the result to all the factories now in the companies.
发言人1 24:20
So this is the architecture that we come up with. And we think, I just want to remind the audience that the architecture is very, very critical. It's not just the Iam itself and how are you going to deploy the result of the llum or the janai is becoming very, very critical.
发言人1 24:50
So next. While we are doing the smart, smart manufacturing, we are at the same time working on evon. You probably heard of our h.h. TD on High Tech Day. This is the annual event. Event in that event will show to our audience our new.
发言人1 25:23
Evol that we created in the last year and the very first product that we delivered, we shipped to the market is the Ebus. And with that Ebus running on the street in gaochang with a lot of sensors in the ebas on the ebas, and we collected a lot of data of the street. And very quickly, we realized that, oh, how are we going to utilize this data? We start looking for some way to share the data with the government and with the enterprises. And then we realized that a lot of. So icall the smart city applications, they all worked in silo.
发言人1 26:36
On our next page, we're going to talk about the smart CT platform to address that problem. But with smart Eev, you know, we are going to provide the smart evols to our potential customers. And this platform will include the onboard and all board. Applications and these platforms, we will open it up to our customers and also we plan to open it to make it a. A reference design? For our Mih members, okay, and we are almost there.
发言人1 27:38
And with smart EV, the most recent news is that we are made the announcement together with the Mitsubishi you Mitsubishi the very first. Car Oems utilizing our reference models with that reference model. For our customers, they can use our reference models. Maybe 80% of their work can be enumerated. They don't have to do 80% of the work, only they had to put 20% of their effort in making their new evols. And that's very, very common in the PC industry. A good reference design gave a good foundation that including very thorough test that required for the system and with that test that save our customers a lot of a lot of time. So that significantly improves the time to market and time to cost for our customers.
发言人1 28:59
Please stay tuned with this new evol.
发言人1 29:07
Next, we talked about the smart city while we were doing the Ebus and we realized that. Most of the smart city applications, be it smart transportation, smart hospital, or smart schools or smart buildings, they all work in silo. They don't work together. And we realize that we need a platform for all these smart city applications to build on top of it so that with this platform, it will connect to connect the government, the enterprises, and the citizens through this platform. And this platform will support. Not only the data sharing, but also the knowledge sharing.
发言人1 30:14
So with this smart city platforms, currently we are implementing it in Gaussian. And also we have a number of other cities in Taiwan currently working with us on the smart city. On the smart city project. We also have cities outside of Taiwan working with us.
发言人1 30:42
So once we had a smart city platform done, I think sometime it will be sometime in a year, we can show it to you that, hey, with this platform, how easy and how useful it is to share data and to share the knowledge between all these applications next. Therefore, you know, with the smart manufacturing and smart evol, smart city, all these ideas came about because of that little joy, okay? And we've been thinking, wow, maybe this is it.
发言人1 31:33
We need AI factories. AI factory to power to power these three smart platforms. That's how we put all this together and we make these three platforms and examples of how AI Factory can do what AI Factory can do. Okay, so we demonstrated this idea in our even at the GTC 24, no, 25, you see 25 about these ideas. So you've heard about 3 plus three initiatives in the past from fasts. Now we add another three now after the three plus 3 to make 3 plus 3 plus three small platform phones becomes fast guns. New initiatives next.
发言人1 32:44
With these three smart platforms, we need to find a model to support it. We've been looking for models that adequate to support this three small platforms.
发言人1 33:02
We keep looking and looking and realize that most of the foundational models, they are all very generalized, okay? They don't understand about manufacturing, especially, you know, that they are their own numbers. But the numbers you find different machines means different things. And you have to have some understanding of this data in order to understand what it means. And with that finding, we're also, you know, together with, oh, with evols also, it's not a. General purpose a.m. can handle the general purpose a.m. probably can answer in a way you want to go, but know you if something going to break up and the signal sent from certain certain module means something will happen in the near future. And this knowledge and this token, they're all very, very different if you don't have special training or special way of processing these tokens, he will give you not just Holts, but illusions, wrong answers.
发言人1 34:51
So therefore, we better come up with our model to support these three applications. And therefore, the firstcom brand idea comes about and it's built on top of, of course, open source models.
发言人1 35:14
Currently we're using LMA 3, 4, and we sort of a created high quality, large scale pre training Cus and fast count Ka on data center will be used and currently is being used part of it to build this model and this first general generation model specialized in reasoning. And that reasoning is a little different from the general reasoning because of the applications. So the genetic workflow for very domain specific applications. And this is, you know, fox brain is going to develop to solve the problem I just described. And we plan to open it to make it open source to the community. And hopefully through your joining the open source communities, we can make this model more powerful, more useful.
发言人1 36:40
So with the Fox brain idea, we think, oh, now that we have the applications, the three smart platforms, we know that we are going to build a fox gram model to support it.
发言人1 36:59
Next. What do we build? Those things, those applications, those systems are we look around, we realize, oh Nvidia is not just chip companies. It has a full stack of software to support it. And we realized that with all the platforms that we talked about and the brain that we are going to create, the end media has a full stack to support us to do that. So we're going to build our platforms on top of the Nvidia AI software stack.
发言人1 38:03
Then the next one will be what? How well should we use? Right, but with this, definitely its immediate hardware. So we need hardware to run it. We need a tremendous compute power to support it. And that's why you heard the announcement yesterday, next.
发言人1 38:32
The AI factory will build. This, the very first NCP AI data center, and this AI data center is targeted to have 100 MW of power. And you know, the power, I don't want use the word shortage in Taiwan, very, very critical resource. So it takes us a couple of steps in order to build it up to 100 MW. So we will build like 20 MW to start with and then add another 40 in the next phase, then another 40, okay? Some of it will be in Gauss. The other ones will look around in Taiwan to look for you where we can get the power the quickest. Then we will build the data center there. So with all this AI related.
发言人1 39:54
The applications? The brand and the software stack and hardware and the computing power, all, you know, comes within the last 18 months, we have gone through all this journey and all this exercise.
发言人1 40:21
To come up with the current status together, we work with BCG and BCG and Nvidia and other partners to create the whole ecosystems. And hopefully with all these ecosystems, we're able to become the leader in the AI manufacturing at the least, and know if we're lucky, we can become the leader in the smart city platform. And very likely that will happen. If that happens, we estimated the compute power that will be needed for supporting the cities, the smart cities would be huge. So that's why about a year ago, when people asked us whether you see the slowdown of the of the compute power requirement, I told them, no, this is just the beginning and because we see the application is coming and we also see the evolvement of the models, it will only become bigger and bigger, you know? So don't worry, it will not slow down, at least in the foreseeable future.
发言人1 42:03
This is another discovery that we had, why we're working on the smart city. And we look at the demographic data and we look at. The issues in countries. And we realized that when a country that will become developed, what will happen that if you look at the GDP paradigm or pyramid, the triangle that I'm talking about, when the country is becoming more and more prosperous, the pyramid or the pyramid or the triangle will shift up, leaving the button GDP empty. So after your country is becoming more and more developed, the GDP will become higher and higher than the low GDP. Work will have less and less number of your countrymen to do.
发言人1 43:29
So how do countries solve that problem? In the past, they do it in two ways. One is to outsource the low GDP work to the low GDP countries. We have seen that in the past. We have seen the us outsource their work to Taiwan and then Taiwan outsourced the work to China. And I remember when we all about work from Taiwan to China about 40 years ago, the GDP there was about less than 1000. And with their hard work, after 40 years, the GDP has grown over 10000. So, what happened? What's the impact to the society there?
发言人1 44:30
Most of people want to have a high paid job and the society is able to pay it. But what about the low GDP, low pay job? Nobody, we'll be interested. So no, they eventually they're going to do outsource the low GDP job to other low GDP countries. Eventually, essentially you will be running out of low GDP countries. That's the limit.
发言人1 45:06
Another way to solve that problem is to import people from the low GDP country, the so-called immigrant workers. And this has been done in Europe, in the us, in Japan, and now in Taiwan. But that created some social problems. But in the past, there's no other way to solve their problems.
发言人1 45:38
Now, with the Advent gennai, we see a great potential of that with janai plus robotics that will feel the void, the gap. And that's the opportunity I see. And that is a tremendous opportunity with J AI plus Robotics to fill the void or the gap when a country, they become more and more prosperous, The low GDP work will be done by the gennai plus the robotics. It's not the service, it's not entertainment. I think that's the real challenge for all developed countries. So that's a great opportunity for all of us, and I urge you and the country developed countries, the leaders of the developed countries to watch the development of this very, very carefully.
发言人1 47:00
I think that's the journey I would like to share with you. And as I said, we are now going to talk about the business. We're now going to talk about the technologies. Just a trend of thought, how we started and how we come up with the current stage. And as I said, this is going to evolve. And with working with partners like nvid and the technology leaders, we think our collaborator will work. We are able to accelerate their progress.
发言人1 47:49
Okay, so before I introduce, you know, this talk, this broken into two parts. The first part it I just I'm almost done with it. The second part will have our special guest joining us. But before I introduce, damn man, I would like to. Make a brief announcement that fast-scan will work with Tech Orange to create Taiwan, leading robust community to drive innovation in the industry growth because of that GDP paradigm shift syndrome. Okay, so let's just wait and see. Okay, thank you for listening.
发言人1 48:56
Is our guest here?
发言人1 49:02
Okay, okay, good, good. Wait a second, okay, let's play.
发言人1 50:20
Okay, ladies and gentlemen, let's welcome the superstar of the AI industry and the leader of King Taiwan. Go to Taiwan. I don't have to him anymore, right?
发言人2 50:38
I being introduced?
发言人1 50:42
No, Jensen, I know you have a very busy schedule. You keynote honest.
发言人2 50:47
you have nothing terrific yesterday.
发言人1 50:50
thank you. I remember the last time when we had our H Htd 23 together, we rolled in with a Beauty and the Beast SUV.
发言人2 51:05
Do you remember the Beast?
发言人1 51:10
And I have to give, let you know, a good news about it has been chosen by Mitsubishi to become their car. Wow, and that car will be running on the road. Roughly about a year from now. Okay, with the luck you bring. And in the second half of the keynote, what I would like to do is to ask you questions and let you answer, okay?
发言人2 51:48
So you know, it doesn't matter what you ask. My answer always the same.
发言人1 51:53
I hope I asked. Actually, I collect a lot of the questions from the industry, from people who care about you and compile them into three category of questions. Number one is business related. Do you want me to start with the business related stuff?
发言人2 52:17
If we start with business related, I will go so long and then we'll end with just business related. So let's start with business related. I'm afraid to hear about the other two.
发言人1 52:28
the other keeping him under control, okay? But it's not a question that you thought about the business. It's not about in media chips or system business, it's about. The team Taiwan The very first question, what is Te Taiwan from your point of view?
发言人2 52:51
Well, Taiwan is. Is where our most important partners are, we've been coming here for 30 years and somebody did account, we have 350 partners here.
发言人1 53:09
350.
发言人2 53:10
350, of course, of course, I don't know them all, but apparently either directly or indirectly, we work with some 350 companies and you know that the technology supply chain is incredibly deep .
发言人1 53:26
here and .
发言人2 53:28
it's much more complicated than people think every as I made a movie the other day that I showed yesterday and it just shows the breadth and the depth of technology that is here in Taiwan and so as a computing company this is the epicenter of the world's computing industry, this is where it starts and everything from the chips of course to the systems and now increasingly increasingly the pioneering work that is being done in the automation of robotic factories and building robots, you know very well one of the largest electronic manufacturing regions in the world .
发言人1 54:13
and not only .
发言人2 54:14
is it the largest it's the most advanced and the things that we built here are incredibly complicated you and I were talking about building GB 200 servers it has 1.2 million components it weighs almost 2 pounds completely liquid cooled each rack is a couple two, $3 million .
发言人1 54:32
all these numbers.
发言人2 54:36
Yeah, and so .
发言人1 54:39
they're not that hard number.
发言人2 54:45
And just so you know, I rounded them all. It's 1.8 tons. It's 1.12 million parts. Yeah, so I rounded it up for your benefit. And so anyways, the technology is incredibly complicated in order to build one of those great black wall racks, how many companies were involved and how much robotics technology is actually involved?
发言人2 55:08
So if you can look, just look at the work that we've done in the last two years since you and I, we're on stage together starting to talk about AI and robotics since then Foxconn is using Nvidia Metropolis for smart cities, Nvidia Metropolis for smart factories, nvidia Isaac Groot for robotics, nvidia drive for autonomous vehicles. Nvidia Omniverse as the digital twin operating system for all of that. And yesterday we announced that we're going to build an AI Factory for you to use for me to use and for Taiwan, the entire ecosystem to use. And so now Foxconn is going to be a world class, regional cloud provider AI cloud provider. And so look at the breadth of work that we do together now. It covers literally every, not to mention for manufacturing inspection agents with computer vision and vision understanding. And so everything from agents to robots, robots to robots with wheels, robots that are orchestrating robots to build robots, you know? And so this is.
发言人1 56:39
I have seen that this is robot building robot. Yeah, this is person.
发言人2 56:43
this is literally coinin, this is Foxconn, you know, one of the most advanced technology companies in the world. You just happen to apply technology to manufacturing. You know, we apply technology for technology's sake, you apply the technology to apply for manufacturing and so. We shouldn't be surprised, but everybody should be deeply impressed by the technology of Foxconn.
发言人1 57:10
Thank you very much.
发言人2 57:12
And so did you notice? I just want to demonstrate something. It didn't matter what he asked.
发言人1 57:20
No, I ask only one question and that question. Go ahead and .
发言人2 57:26
ask about my jacket.
发言人1 57:30
No, no, no, no, no.
发言人2 57:31
I just want to demonstrate something. I'm .
发言人1 57:34
not going to ask about Jack.
发言人2 57:36
No, I just want to demonstrate it doesn't matter what you ask.
发言人1 57:39
you want answer the same thing.
发言人2 57:42
Because from there, I was going to talk about Omniverse and robotics.
发言人1 57:46
Great salesperson. Anyway, my next question is, have you heard about Tema .
发言人2 57:54
from Who .
发言人1 57:56
T EMA? Is it association, Taiwan Electronic and Electrical Manufacturers Association?
发言人2 58:06
Did I do something wrong?
发言人1 58:07
No, no, there is something he doesn't know. I was just made the chairman of Tim Tim.
发言人2 58:13
well, congratulations. Hema, congratulations. Did you do something wrong?
发言人1 58:23
I hope, I hope no. But anyway, it's the largest association in Taiwan, more than 3000 members and accounts for over 50% of export value. A Taiwan aggregate number .
发言人2 58:44
of workers and Con represented 40% of. 37%.
发言人1 58:56
Because model, we are worldwide, the revenue counted, the entire one is only about half. So Tema is the association, I think most of the members. Most of your suppliers are members of Tema. Now, with this rapid changing AI industry, what advices do you have for your suppliers as a team, a member?
发言人2 59:34
Any advices for them? Yeah I do, it'll come to me in just a second, but I'm just going to have to start. I do, so let me make an observation.
发言人2 59:46
Let me make an observation. In fact, the vast majority of the industrial, the industries, the company's focus on industries has largely been left behind, as during the software revolution. And the reason for that is because you're obviously an industrial, because you understand mechatronics, you understand mechanicals, you understand chemicals, you understand the deep physics, the sciences part of industry. That industry was was formed out of the last industrial revolution. And Taiwan is excellent in those industries. But when software came was a brand new skill, and those skills were quite unusual and quite unnatural for these companies, number one.
发言人2 01:00:36
Number two, first observation, the second observation is that when AI started 10 years ago, it was very interesting, but not very useful, not very useful for the industrial companies. And the reason for that is because the AI understands English, but it doesn't understand physics, which is what you need. Everything that you do needs physical understanding, physical intelligence. You know, when I roll a ball underneath a car, my dog knows to either go underneath the car or go around the car, but the AI thinks the ball has disappeared out of this universe. We need, we need to have physical common sense Now we're developing both of those things.
发言人2 01:01:20
Now, why is AI so great? The reason for that is because you teach an AI, you don't have to program an AI software programming skills is no longer as vital as it was before, you have an opportunity to leap the software chasm and go directly to AI. In fact, I can prove it to you.
发言人2 01:01:43
Everybody is a software programmer Now, the definition of a software programmer is somebody who programs a computer to do something that's a software programmer, give me an example of somebody who can't tell ChatGPT to do something, and if you don't know how to even tell ChatGPT something, you say ChatGPT teach me how to teach you how to do something, isn't that right? And ChatGPT writes a very nice prompt and say, try this, and I say OK, isn't that right incredible? So this computer is now intelligent and intelligent systems are easier to use, and now you have an intelligent system that's easier to use it understands the laws of physics. We combine all of that into Omniverse. You see how I came back again and came all the way back into Omniverse, which you are using as a digital twin, isn't that right? Within the digital twin, within the digital twin, you could design your plants, you could operate your plants, you could simulate your plants, and plan everything before you operated greenfield or brownfield, isn't that right? And so all of a sudden, every company in Tema, every company in Tema is now a technology company.
发言人1 01:03:01
And it has to be, it has to be a technology company.
发言人2 01:03:05
It doesn't matter what you ask me.
发言人1 01:03:10
Okay, no, if I switch to .
发言人2 01:03:13
one personal .
发言人1 01:03:15
related questions.
发言人2 01:03:17
it's going to come right back to Tema.
发言人1 01:03:22
Okay, I think Fatima, I think they heard it and they will have to keep up with the AI and transform themselves to have more and more software capabilities, right?
发言人2 01:03:38
And in fact, look, every company needs to do what Foxconn did.
发言人2 01:03:42
When you and I first talked about AI, you didn't talk about AI and read about AI and kept thinking about AI. You just started working on AI. Just get to it. You know, of course, the technology two years ago is not as advanced as it is today, but you knew that. You knew that it was going to get better. You knew that technology was going to get better, all of course it does, it's getting better at a million times every 10 years, a million times every 10 years. It's like changing every day, and it is AI is changing every single day, but you engage it today with that expectation that these challenges that you're experiencing are going to get solved almost by itself, but you've got to get engaged right away.
发言人2 01:04:26
Everybody has to develop a system, We have a system available to you. You bring it to your company and Foxconn and I will build it for you. All the software is capable, available for you. We're already integrated into all of your software developers, whether it's Cadence or Siemens, the So, or on a desk or whoever it is, We're already working with all of your software developers. Omniverse is integrated into all of that. Nemo is integrated into all of that. And so you could start trying AI.
发言人2 01:04:55
And it's amazing what you've produced in just the last couple of years. The demos you showed me, you know, your factories are basically complete digital twins, and they look so real, they look so real robots working inside it, incredible.
发言人1 01:05:10
So in the past, we have physical factories. Then we use Omniverse to create digital twins's, right? We realize we need AI factories to support both of them. That's right, to make it more and more powerful. And all these changes evolves so fast. And the reason why it's evolving so fast, it's all because, damn, man, you creating chips running so fast and advancing so fast every year. And my next question for you, do you see any limit in terms of a hardware speed for the AI?
发言人2 01:06:02
Remember Moore's law ended. Moore's law ended because we're at the limits of physics, and so the cost of the transistor is going up as it should, the energy efficiency of the transistors and the processors are not going up very fast, understandably. And so we have to try different skills and different techniques. And so we're developing multi-chip packages, 3D packages, one way to do that, the other way to do that, of course, is this incredible invention called nvlink, where we connect all of these chips together into one giant chip, You know, the size of the wafer is probably, you know, like that. And so all of these chips are working together as one. We use advanced mechanical systems and liquid cooling systems to compress the systems into liquid one giant at rack, but very importantly, software, our architecture and the algorithms developing on top of it allows us to reformulate the AI models reformulate how the work is distributed across these supercomputers.
发言人2 01:07:09
The networking, of course, is a big challenge and a big opportunity. Instead of passing information back and forth, we could do in network, in fabric reduction. And so, you know, we all send our answer together for us to combine all the answers. But why don't we just send it to the network? The network combines the answers and already sends it back to us so we don't have to receive it. So we reduce the amount of traffic, like practically in half. And so all of these clever ideas combined has allowed us to improve AI performance tremendously.
发言人2 01:07:45
You know, people say it's about, you know, I don't know, 2x every six months. I actually think it's probably working faster than 2x every six months at this point. You know, it's probably like 2x every three months. The combination of all these things, I call it full stack, but basically you're innovating chips, innovating systems, innovating at the data center, innovating the operating system of the data center, and innovating the model all at the same time.
发言人1 01:08:10
So from why you see so far, only the sky is the limit.
发言人2 01:08:17
Only the sky's the limit. And this is the really amazing thing, young. There's the computer industry that was created.
发言人2 01:08:27
And remember the first time I realized this was going to be an industry, the computer industry was going to be when Bill Gates, in one of his interviews, and he was so young, it was 1984, 1984, maybe 1985, I heard an interview where he said that Microsoft is not a maker of word processors or spreadsheets. It's a manufacturer of software. And at the time, I heard it's a manufacturer of software. What does that mean? Software is so easy to manufacture? You write one copy and the rest of it is duplicated, okay, and the rest of it is free, and so how would you create an industry where everything that you produce is free? Well, of course, he invented licensing. And so I didn't realize what he had in mind was a whole bunch of lawyers, but more importantly, what he meant was there's a methodology, there's a production methodology associated with software, and look at all the software they produced today, it was a visionary idea, and now it's a trillion dollar industry.
发言人2 01:09:29
Well, our industry, remember AI is a great technology, but AI is also an industry in itself, just as software and you've spreadsheets and WordPress, but collectively software is an industry in itself. AI is going to be an industry, however the AI is going to be produced by factories, and so now you just have to think, how large is the automation agents and robots industries that will support? And every one of those agents and robots will need factories to produce the tokens for them at all times. I think that we've unquestionably discovered, invented, created a new multi trillion dollar industry, which is the reason why everybody's excited about it.
发言人1 01:10:16
Excellent, excellent, Okay, so you know, the time is running short.
发言人2 01:10:22
I'm trying.
发言人1 01:10:23
I have some questions .
发言人2 01:10:25
I can to exhaust all the time.
发言人1 01:10:28
Thank you, thank you. Then I'll skip some of the technology related questions. I'm gonna go to the personal questions, I hope you will not be offended.
发言人2 01:10:39
Okay, let's take a technology question.
发言人1 01:10:44
In China.
发言人2 01:10:45
everybody knows all my favorite restaurants in Taiwan anyway.
发言人1 01:10:48
No, it's not that .
发言人2 01:10:49
everybody knows where I'm going to be this afternoon.
发言人1 01:10:52
There's no doubt there's an old Chinese saying that every successful man stance, a terrific woman. I'm not going to ask you who that woman is, but I want to ask you to you.
发言人2 01:11:07
I hope it's the same 1 I have in mind.
发言人1 01:11:14
Do you want to say something to her openly? Openly .
发言人2 01:11:24
say I say the same thing to her every day. Love you, honey. And so anyways, I met her when I was 17 years old and she had just turned 19. And so I dated an older woman for most of my life.
发言人1 01:11:40
What part of the state you met her?
发言人2 01:11:44
Oregon, Oregon? Yeah, we were lab partners in electrical engineering.
发言人1 01:11:48
Oh, that partner.
发言人2 01:11:49
Okay, so imagine the odds. Imagine how smart you have to be to achieve this. Okay, so there are 250 students in electrical engineering fundamentals, there were only three girls also, so statistically impossible, not only that, it is very clear who's the youngest person in class, I was £128 going to college. I graduated at £132 or something like that. Okay, that was £128, so I was definitely, I looked like a kid in a room of adults and so I just went up to her and I noticed this beautiful, beautiful girl in the class, listen and I went up to her and I used my superpowers .
发言人1 01:12:36
superpower .
发言人2 01:12:37
because I knew just looking at the audience who everybody knows what my skills are. And so I just told her, you know, would you like to see my homework? Because those are my superpowers. Everybody has superpowers. And so I also told her that that I don't know what grades you have, probably very good. However, if you study with me every Sunday, I promise you you'll get straight A's .
发言人1 01:13:11
every Sunday.
发言人2 01:13:12
And so I had a date every Sunday.
发言人1 01:13:14
I see, I see, very smart, yes, yes, that tells us the reason why he is so successful starts from age 17, Okay, not just now. Okay, my last question for you is that. Everybody is curious about your jacket, but I'm not. Recently, your watch is becoming very popular.
发言人2 01:13:46
I know, I don't know, I don't know.
发言人1 01:13:49
where is your watch?
发言人2 01:13:50
I don't wear a watch.
发言人1 01:13:52
but is that fake news? Yeah.
发言人2 01:13:55
it's fake news.
发言人1 01:13:58
Oh really, Oh, that's a fake news, then I have to tell my wife, you know, that's a fake news. I'm not going to buy you a watch it, that's price. Okay.
发言人2 01:14:10
that's a fake news, that's it. No, no, that's fake news. I don't wear a watch. No, I don't care about the time right now is all that I don't care about, that's why I don't wear watch. I'm 100% right here, okay?
发言人1 01:14:27
He really confused me. Is that watch news a fake? A fake news?
发言人2 01:14:34
It's probably a fake watch.
发言人1 01:14:38
Okay, thank you, thank you very much. We're very happy to have you here and through the conversation, we know the other side of Jensen and hopefully it will help you to know Jensen more besides his chips and his success, his business.
发言人2 01:15:02
Well, my chips are pretty good. All but if I could say young, I am delighted by our partnership.
发言人2 01:15:12
Obviously, we're manufacturing the most advanced computers the world's ever seen and ever known. And during our meeting yesterday, one of your leaders said no one has ever manufactured systems this complicated. But very importantly, no one has ever manufactured supercomputers at this rate. Most of the time, you plan a supercomputer for three years and you build one. We're cranking it out every hour. And so the idea that we're doing this at the scale that we're talking about is extraordinary, but that's the work that we do together for my benefit.
发言人2 01:15:49
The thing that I'm really excited is that you then use this incredible AI technology to revolutionize, transform Foxconn, and also to make it available for all of the Taiwan researchers and students and other startups and all the industries. I think it's a great example that you're showing. It's great that you're revolutionizing your own company with technology. You were an amazing technology company already before, now you extraordinary technology company, and then everything that you do for Taiwan. So I'm very, very grateful.
发言人1 01:16:19
Thank you, thank you very much, James, you for thank everybody you bye.
沒有留言:
張貼留言