Chapter 214 Ambitious


Chapter 214 Ambitious

After a certain audience member shouted something, Huawei’s awesome voices echoed in the audience, and the previous crisis of trust caused by the accident six seconds later was wiped out.

“In addition to the computing power advantage just mentioned, the biggest advantage of Ascend 1000 is that our basic material uses a new topological semi-metal, which can maximize its use when building an AI cluster. Utility.

It has no limit, or we have not discovered the limit of Ascend 1000, nor the upper limit of the AI ​​cluster built on Ascend 1000.”

NVIDIA previously said. The DGXA100 system can only combine eight A100GPUs as one unit. It is unable to cope with the explosive growth of computing power in large generative artificial intelligence models.

Later, NVIDIA built a DGXGH200 to connect 256 H200 super chips and up to 144TB of shared memory into a unit through a special connection system, but it still has limits.

This is determined by optical fibers and silicon-based materials. Breaking through the limits requires breakthroughs in materials science.

The Shengteng 1000 has achieved such a breakthrough.

In the past, everyone could only use the H800, a special version produced by China. In the past two years, even the H800 has become unavailable.

Of course, Nvidia is also trying to make some breakthroughs based on existing technology. For example, it is teaming up with Amazon to build an AI supercomputer equipped with 16,000 H200 GPUs.

"Everyone is exploring the limits of AI computing power. We are selling Ascend 1000 to the outside world. We welcome all colleagues to join the Ascend family and explore its limits with us."

Everyone present here What the big guys think in their hearts is that using Ascension is just a temporary measure.

Sooner or later I will have to design and build things myself.

After the emergence of topological semi-metal and Huawei launching the Ark mobile phone chip, not only major mobile phone manufacturers started chip research and development, but also various Internet giants were also engaged in chip research and development.

It’s just that the Internet giants are investing in AI chips. Everyone is envious of NVIDIA and wants to cut a piece of cake from NVIDIA.

In the past, domestic manufacturers had no chance because your chip manufacturing capabilities could not keep up anyway. Therefore, those who tried to share the cake were Silicon Valley giants, led by the two brothers Microsoft and Google. The two brothers were the most eager to cut the cake from Nvidia.

Take Microsoft as an example. It is completely allinAI. It has stockpiled 1.8 million AI chips in 2024 alone, claiming to invest hundreds of billions of dollars in AI chips and AI chips in the next three years. AI supercomputer is under development.

In the past, domestic manufacturers could not get a share of the cake. Only Huawei could barely enjoy the dividends of Nvidia's advanced chips not entering China.

After the feasibility of topological semi-metal has been repeatedly verified by Huawei, major domestic Internet giants have become interested. This condition was not met before, but now the conditions are sufficient, and we also want to share the cake.

It’s not a loss just to meet their own internal needs, not to mention that Baidu, Goose Factory and Ali can all use AI to provide external services.

Ali has always invested in research and development in the chip field. They launched the Hanguang 800 in 2023 and have been making attempts. After Xiao Ma returned to Ali, at the end of last year, he even announced that he wanted to develop AI. Consider it the number one strategy.

The same is true for Baidu and Goose Factory, as well as ByteDance, various new energy car companies and other large and small companies. Basically, they all have their own thoughts.

Nvidia’s revenue of US$26 billion in a single quarter is enough to make any company excited.

After the Huawei press conference, everyone was contacting Huawei, because it is right for them to develop their own chips, but this is a matter of the future, and they still need to meet their own internal computing power needs.

“We are very happy to reach a cooperation with Goose Factory to try out super AI computers through Shengteng 1000, but we need to start from the second half of the year to ship outside.

Huawei’s Shengteng 1000 external production capacity in the first half of the year has been taken over by Guangjia Aerospace. ”

Everyone is speechless, what kind of press conference can you hold if you don’t have the money?

I am further curious, has Guangjia Aerospace completed the production capacity of Shengteng 1000? Chen Yuanguang is also going to enter the field of artificial intelligence. Is this not good news?

Everyone knows that Chen Yuanguang is not an easy opponent. Whether it is his own ability or strategic vision, the most terrifying thing about Chen Yuanguang is that he has made no mistakes.

Doctor, everyone is saying that the future is the era of artificial intelligence. All giants are spending money on artificial intelligence, and resources are flowing into this field intensively. Light Armor Aerospace is no exception.

We lack a person who knows enough about artificial intelligence and a leader in this field. I pride myself on knowing a lot about artificial intelligence, but my energy on artificial intelligence is limited after all.

If you are willing to work at Guangjia Aerospace, you will be the vice president in charge of the artificial intelligence sector. Our annual R&D expenses in this field will not be less than 5 billion RMB, and you will be directly responsible to me. , with no other management levels in between. "Chen Yuanguang said.

In a cafe near Mianyang High-tech Park, Lu Qi finally met a legendary figure in the entire Chinese world after going through many layers of inspections.

Lu Qi took a sip of the American food on the table that was no different from the one produced at Shen Haili’s independent coffee shop. He thought that this would not be a coffee shop specially built for Wright, then nodded and said:

“Mr. Chen, Since I will take a flight of tens of thousands of miles from California with a few layovers to come back to see you, it shows that I have a very strong interest in working with you.

After so many years of ups and downs in life, my career is not perfect, but it can also be considered successful. ”

Lu Qi is also a legend. After graduating from his Ph.D., he served as an executive at Yahoo, Microsoft, and Baidu. When he was at Baidu, he once served as CEO and led Baidu’s allin artificial intelligence strategy. He absolutely hit the ceiling of workers. The self-proclaimed working emperor is much more valuable, but compared with Chen Yuanguang, his legendary character is not worth mentioning. After the assassination of Amerikan was heavily exposed by overseas media, Chen Yuanguang's legendary character became even more legendary. First floor.

“Compared with salary, position, and power, I am more interested in what I do. Guangjia Aerospace is obviously such a company with a grand vision and more What matters is having a clear blueprint for achieving your grand vision.

I have a few questions to ask. You will take the time to see me. You must know something about my resume. The companies I worked for in the past were all toC companies. The business of Guangjia Aerospace is obvious. It's not toC's business.

So this brings up a question, what is the purpose of Guangjia Aerospace entering the field of artificial intelligence? Like other technology giants, do you want to do TOC business and generative AI?

Or what other purpose?

I need to know what the purpose of Light Armor Aerospace is in the field of artificial intelligence. "

Chen Yuanguang said: "No, no, no, that span is too big, and it is not my style to do generative AI purely for the sake of chasing popularity. This kind of thing is too boring.

We are just different from the giants who focus on toC business. We plan to enter the field of artificial intelligence from now on, and the most important thing is to meet our internal needs.

That is, space mining. In the future, space mining will inevitably be a large-scale and high-volume artificial intelligence operation. This year, we acquired a total of 7 robotics companies, including Datima, and spent a total of 5 billion RMB.

If the federal government didn’t allow it, I would even want to buy Boston Dynamics from Google.

The goal is to combine robots and artificial intelligence to create artificial intelligence for space mining.

In the process of pursuing goals, there may be some by-products based on actual output, but these by-products must serve our core purpose. "

Lu Qi nodded to express his understanding: "I understand, there is still a big difference between this goal and my previous direction.

But I also did similar business when I was at Baidu. Baidu’s autonomous driving business is more of a tob business.

As far as this is concerned, to be honest it is too difficult, really very difficult.

Now everyone knows that the three most important points of artificial intelligence are data, computing power and algorithms. Data and algorithms complement each other, and algorithms can be optimized with large amounts of data training.

In terms of computing power, judging from the potential shown by topological semi-metal, China will be able to solve this constraint sooner or later.

Only data, the field you want to do is too restrictive, limited to space mining, which means that all the data you simulate on the earth is useless. Robots cannot use the earth in the weightless environment of space. Paradigms under gravity.

This means that no data can be collected, and data can only be collected after the robot is sent into space. The data samples collected will be very small, which will be very inadequate for algorithm training.

Data can be a big problem. "

Lu Qi immediately grasped the key to the goal proposed by Chen Yuanguang. He is worthy of being a big boss in the field of artificial intelligence. He has always been engaged in search engine and artificial intelligence related work.

Chen Yuanguang said : "That's right, that's why we plan to launch a space station in the next two years. After launching the space station, we will send robots into space to collect data.

This is indeed difficult, but it doesn’t mean you don’t have to do it. ”

Lu Qi nodded repeatedly after listening. The reason why he resigned from Baidu was ostensibly his initiative. In fact, his strategic adjustment affected Baidu's core interests, and a large number of internal objections forced him to leave.

Lu Qi couldn't help but think of his former employer, Baidu. The money from search engines was so profitable, and path dependence led to Baidu's complacency and difficulty in transformation.

“Yes, you can’t just do it just because it is difficult.

Mr. Chen, I am very happy to receive such an offer in the twilight years of my career. Martyrs are full of ambition in their twilight years. I still hope to be able to do it.

I can also leave a big mark on mankind's journey to space in the future.

I have one last question, and that is treatment. I don’t care much about salary. I only care about options. Are there options? "

Chen Yuanguang felt relieved after Lu Qi agreed, and then he couldn't help laughing when he heard this question: "Options are useless. ”

(End of this chapter)

Previous Details Next