Technology invades the modern world

Chapter 291 Before the Joint Moon Landing (Seeking Monthly Tickets!)

"I have already communicated with Huawei's senior management on this matter. They provide 40 billion yuan every year, and my original idea was to use their computing cards to offset 20 billion yuan of that amount."

Now that China Investment will receive 500 billion yuan, Huawei will use computing cards to offset all of its annual 40 billion yuan debt.

To put it simply, Apollo Technology is facing a more serious crisis than Huawei.

Bilibili, which advertised with Apollo Technology, was forced to return to Hong Kong for its IPO; Huawei, however, will not be treated the same way.

In addition, Apollo Technology has already fully demonstrated its ambition and capabilities.

It's normal to have ambition; everyone has ambition, but you need to have the ability.

Apollo Technology was different; it forced Lao Ma to defect two years earlier than expected.

Musk could have stayed under the Trump Party banner until mid-2024 before switching sides.

It's clear that Apollo Technology dealt real damage to Prince Ma.

For the White House's Cold War veterans, they would rather let Huawei find advanced manufacturing processes than allow Apollo Technologies to buy Nvidia's AI computing cards.

Fortunately, there is Huawei, which launched its own computing card in 2019.

It's 2022 now, ChatGPT hasn't been released yet, and major manufacturers can buy stripped-down versions of NVIDIA AI computing cards. Huawei's Ascend is practically ignored.

Lin Ran was willing to use half of the advertising expenses to offset the debt with Ascend, which was one of the important reasons why Huawei agreed to the 200 billion yuan price.

The fact that top domestic technology companies are willing to use their Ascend computing cards is also beneficial to the improvement of their computing card ecosystem.

As for why they don't cooperate with other manufacturers, the Ascend ecosystem, backed by Huawei, is still a blank slate, so it's easy to imagine what other manufacturers are doing.

"Actually, Huawei's computing cards are still quite far behind Nvidia's. Alas, there's really nothing we can do. We can't buy Nvidia computing cards right now, not even the crippled versions."

Pony smiled wryly. He recalled that his office computer still used the Linux operating system, and Apollo Technology had a dedicated desktop support team to ensure that everyone could use Linux comfortably. He then realized the predicament he was currently facing.

Huawei has been promoting HarmonyOS to Apollo Technology, saying that HarmonyOS is definitely better than Linux. As for industrial software such as rocket design simulation software and fluid dynamics calculation software that cannot be used on HarmonyOS, we will solve that for you.

However, Huawei is pushing for a solution, but we haven't seen a solution yet.

Lin Ran smiled wryly: "Yes, the problem is, even if Nvidia sells it to us, I wouldn't dare use it. Who knows what might happen?"

Lin Ran then added excitedly, "Fortunately, I've communicated with Huawei, and their chips are sufficient for us."

Because these models derive material properties from elemental characteristics, the data volume is very sparse. Of the three key elements—data, computing power, and algorithms—the dependence on data and algorithms far outweighs the dependence on computing power.

Pony also has a good understanding of artificial intelligence. Tencent recruits countless top talents from the AI ​​field every year. Even though ChatGPT has not yet emerged, he hopes to learn more from Lin Ran to provide direction for future work: "Mr. Lin, please tell me more."

Lin Ran further explained: "This is because data in the field of materials science is extremely limited, and data sharing and acquisition face unprecedented obstacles."

Experimental data from different laboratories will not be included in the same pool unless they are published in a paper. Of course, they will also have various concerns if they want to be included in the same pool.

Because it's difficult to guarantee that the data provided by all research institutions won't pollute the database.

If someone falsifies data, it will contaminate the entire data source.

From what I understand, similar research data is extremely scarce, with the most extensive data containing fewer than 4000 samples.

Feature engineering is key to the success of AI models, but its design is particularly complex in predicting material properties.

Physical element properties, such as atomic weight and electronegativity, and material structures, such as lattice type and bond length, must be converted into numerical features to provide the model for learning.

Feature selection directly affects model accuracy, and incorrect selection may lead to performance degradation.

Currently, the entire process still requires researchers to manually process feature values ​​and perform screening.

It relies heavily on the researcher's experience and intuition, and is very likely to miss important information.

Last year, Nature's sub-journal developed a learning framework called MODNet, which is a machine learning framework for predicting material properties.

(The paper "Material Property Prediction on Limited Datasets through Joint Learning of Feature Selection and MODNet" was published in Nature sub-journal NPJ on June 3, 2021.)
They found that antibonding length and p-valence electrons are key features when predicting the vibrational entropy of materials, but manually identifying these features requires deep domain knowledge.

Extracting this data requires experienced researchers, and ensuring data accuracy and minimizing errors is also crucial; the entire process is extremely complex.

Because what we're doing is far more complex than what they are doing—we're building a much larger and more complex model, and the summarization and collection of feature data will definitely be very slow.

After all, unlike data in cyberspace, which can be eliminated through feature removal and other methods to ensure data accuracy, this data, in computer terms, appears to be structured on the surface, but is actually very unstructured at its core.

Therefore, according to my estimate, Huawei's computing cards will be sufficient for at least the first five years.

As for five years from now, Huawei's computing cards will also keep pace with the times, and we will also cooperate with Huawei to advance their computing card development.

After listening, Pony was able to roughly grasp the ideas, though he couldn't say he fully understood them. After all, expecting a fifty-year-old to understand concepts like vibrational entropy, anti-bond length, and p-valence electrons would be asking too much of Pony.

But he understood what Lin Ran was trying to say.

Pony said, "Mr. Lin, I have no objection to cooperating with Huawei. Similarly, I am very clear about the situation we are facing. While there are manufacturers such as Cambricon, Alibaba, and Baidu that have their own computing cards, on the one hand, their computing card manufacturing needs to rely on TSMC, and on the other hand, in terms of ecosystem, Huawei has gone the furthest. In the long run, they have the greatest determination and ability to build an ecosystem."

I'm just lamenting the difficult situation we're currently facing.

"Mr. Lin, I have a question. Shouldn't we collaborate with chemistry and physics departments at some universities on collaborative research projects to help us improve our data pool?"

At this time, Huawei is not the only company on the market that has computing cards. All the companies that Pony mentioned are also promoting them. However, computing cards are not just about hardware; the software ecosystem that goes with the hardware is equally important.

Why is Nvidia so dominant? Doesn't AMD also make AI chips? Why is it that both are American companies, yet AMD's computing cards can't threaten Nvidia? Nvidia's moat lies in the CUDA ecosystem it has cultivated over the years around its computing cards.

Similarly, Huawei is determined to build HarmonyOS, and in the field of computing cards, they are the best choice.

Given that everyone is a thorn in America's side, it's only natural that they would band together for mutual support.

Lin Ran said, "Of course, I have thought about it, but not now."

If the Apollo moon landing could exploit students, how could it not take advantage of China's vast number of science and engineering students when it comes to building artificial intelligence prediction models for materials science?

These are all high-quality, all-natural laborers.

Instead of helping my supervisor with a research project, I'd rather do a research project for Apollo Technology; at least the latter can actually change the world.

"My idea is to wait until we have finalized the entire data collection plan, and then expand it outwards, using Shanghai Jiao Tong University as a pilot."

Building this thing will definitely rely on the strength of domestic universities, which is also our advantage.

Didn't we mention earlier that data collection is difficult? With rules in place, we have methods to follow, both in terms of the standardization of data collection and the removal of dirty data.

That will be the time when domestic universities will participate on a large scale.

To put it simply, this AI-powered industrial software for materials science will be our biggest competitive advantage.

Pony, think about it this way: if the metaverse could one day become a reality, and virtual reality could truly give people the same experience as reality, then the current physics engine, which can only build animation effects, would definitely not be enough.

Our industrial software kernel will be the foundation of the future metaverse.

However, this is a bit too far.

In the short term—and by short term I mean this century—our competitive advantage lies in three areas: first, artificial intelligence industrial software in the materials field, which I call the Fuxi platform; second, a series of self-developed industrial software for the aerospace field, which we will build our own ecosystem on the existing open-source foundation; and third, data. As you can see, we now aim to make lunar landing as easy as drinking soup.

A step ahead leads to further progress. Data on the lunar surface, interactions between the Earth and the Moon, the probability of meteorites in various regions of the Moon, and the lunar landscape will all serve as our competitive advantage.

We will be the first to go to the moon, and the first to go to Mars, the first to leave our mark throughout the entire solar system.

Industrial software ecosystem, space-related data, and the Fuxi platform—these three are the means to achieve our goals. Therefore, regarding talent in fundamental disciplines such as computer science, physics, materials science, chemistry, and mathematics—as long as they are talented, we can cultivate them gradually. Our plans, even in the short term, are measured in centuries.

After pausing for a moment, Lin Ran grinned and said, "Pony, the foundation of this company that aims for the universe was laid by you and me in the first five years."

After listening, even though Pony was a seasoned veteran, he felt an unprecedented surge of ambition. He thought of a saying: "An old steed in its stall may still aspire to gallop a thousand miles."

Aren't I just like Old Ma right now? I've gone from being a young pony in the internet industry to the old pony I am today.

Pony finally understood what Lin Ran relied on to attract employees from NASA's golden age, such as Aldrin: grand goals and unparalleled achievements.

Lin Ran is now talking about grand goals; the manned lunar landing and the landing in the lunar south pole crater are unparalleled achievements.

The two together ignited a fire in Pony's heart. He felt that God had been very kind to him, and the shadows that miHoYo and ByteDance had cast over the Tencent empire seemed to have disappeared at this moment.

"Mr. Lin, don't worry, I will do my utmost and do everything I can," Pony said earnestly.

As one of the second batch of astronauts scheduled to land on the moon, Wei Xuhang's initial excitement had long since worn off.

He had already guessed that it would be his turn sooner or later when he was circling the moon in the command module waiting for Lin Ran and Aldrin to return to the command module.

However, the fact that he was going as early as October of the following year was still a bit unexpected.

The fact that the lunar landing method was not the Apollo moon landing, but an unprecedented segmented launch, landing at the lunar south pole and transferring lunar fuel, was even more unexpected for him.

When I first learned about it, I was overwhelmed with mixed feelings – excitement, elation, worry, and trepidation. As someone who had received astronaut training, my training included not only operations but also scientific knowledge. This knowledge clearly told me that the lunar south pole was a good place to build a base, but getting there would be extremely difficult, much more difficult than the Sea of ​​Tranquility where the Apollo moon landing took place.

I'm definitely worried, what if it's a one-way ticket?
However, there was only worry, not refusal. Wei Xuhang could not resist such a temptation, and Chinese astronaut Li Cong was even less able to resist it.

Li Cong and Li Guang trained with Lin Ran for fifteen days, but failed miserably in the exam. After their failure, they watched as Lin Ran became China's first person to land on the moon.

Putting aside the fact that Li Cong is a soldier and obeying orders is second nature to him, even if he were given the choice and told that the success rate was only 1%, Li Cong would still choose to carry out the mission without hesitation, and would even say "I guarantee to complete the mission" when accepting the mission.

So when Li Cong learned that Lin Ran had personally selected him to participate in this launch, he had only one thought in his mind: I've been chosen by the professor! The professor has such good taste! Holy crap, I'm going to the moon!
Even knowing that going to the moon would be a completely new and unprecedentedly dangerous undertaking, he never wavered in his resolve.

Inside the China Astronaut Research and Training Center in Yanjing Aerospace City, Wei Xuhang and Li Cong are receiving training.

The predecessor of this place was the High Altitude Physiology Research Group under Group 581 of the Chinese Academy of Sciences, established by Dean Qian and Zhao Jiuzhang.

It later became an astronaut training center with the motto "From here to space" and the code name Dawn.

The center has a dedicated high-tech simulation hall where a huge screen simulates the rugged edges of Shackleton Crater on the south pole of the moon, with jagged rocks and long shadows that look particularly dangerous under the low-angle sunlight.

Thanks to the successful Apollo missions to land on the lunar south pole, the simulation has been enhanced with new images and data, making it increasingly closer to reality.

Wei Xuhang sat in the cockpit of the simulated lunar lander, his hands gripping the control stick tightly.

Although his spacesuit was only for training, the heavy pressure gave him the illusion that he was already on the moon.

“Wei Xuhang, pay attention to your altitude and descent rate,” the trainer’s voice came through the headset. “The Antarctic terrain is complex, and any deviation could lead to failure.”

He stared at the screen; the radar showed a huge rock blocking the planned landing point.

"Obstacle detected, adjust trajectory." Wei Xuhang muttered to himself, his fingers quickly manipulating the lateral thrusters to guide the lander away from the danger zone.

The fuel indicator light flashed, reminding him that time was of the essence.

“Height 200 meters… 100 meters…” Wei Xuhang read in a low voice, his voice focused.

As the terrain on the screen drew closer, he finally found a flat area and carefully reduced his speed.

The simulator vibrated slightly, and the screen displayed "Landing successful".

The trainer came over and patted him on the shoulder: "Good job, you reacted quickly, but next time you need to spot the obstacle earlier."

Wei Xuhang nodded and said, "Understood, I will be more careful."

Astronaut Li Cong, who was standing next to him, stepped out of another simulator, took off his helmet, and said with a smile, "Xuhang, I almost crashed into a virtual rock just now."

He spoke in a relaxed tone, trying to ease the tension of the training session.

The trainer added to the two, "Although Apollo technology has achieved automatic landing, it's impossible for every automatic landing to be so smooth. You need to be prepared to manually take over the control panel at any time!"

At the other end of the training center, a huge indoor simulation field features artificial craters and a lunar soil-covered surface that recreates the desolate landscape of the lunar south pole.

The lights were adjusted to a low angle, casting long shadows to simulate the extreme light conditions in Antarctica.

The two men, dressed in simulated spacesuits and holding geological tools, moved carefully.

Beforehand, the trainer in charge of the subject stood on the observation platform and pointed to a rock, instructing: "This is simulated basalt, which may have been found near Shackleton Crater. Your task is to collect samples, making sure not to damage the structure."

“In Antarctica, shadowed areas may conceal water ice. You need to learn to identify possible ice features; this is your most important task this time—to find the presence of water ice,” the trainer continued. “If we repeatedly fail to find water ice in Shackleton Crater, then we may unfortunately have to find another crater to build our lunar base.”

In the low-gravity simulation zone of the training center, they were suspended by a levitation system to simulate a lunar gravity environment of 1/6.

Wei Xuhang and Li Cong stood on a slope covered with lunar soil, simulating the terrain of the edge of a meteorite crater.

The coach stood to the side, holding a tablet computer, recording their performance.

There is also low-light operation training.

As night fell, the lights in the training hall were dimmed to simulate the environment of the permanent shadow zone in Antarctica.

Wei Xuhang and Li Cong, holding flashlights, attempted to install a simulated seismograph in the dark.

“The visibility is too low,” Wei Xuhang frowned, adjusting the angle of his flashlight. “We’ll have to rely on a laser rangefinder to judge the distance.”

He turned on the device, and the screen displayed: "4.8 meters from the target point."

The two moved carefully, avoiding the simulated rocks.

Li Cong carefully connected the power cord to the seismograph.

Suddenly, the trainer's voice rang out: "Simulating a spacesuit malfunction, Wei Xuhang, your oxygen system is leaking."

Wei Xuhang immediately got into character, pretending to check the spacesuit.

Following the training procedure, Li Cong took out a spare sealing tape and simulated the repair process.

He said smoothly, "Fixed, continue the mission."

After completing the installation, the two stepped back to check that the instrument was functioning properly.

During my time in Yenching, the training subjects were all-encompassing.

The training is much more complicated than it was when it was just a command module.

It's not that it's difficult, but rather that there are many subjects.

Since Apollo's successful soft landing in Antarctica, a voice has emerged within China's aerospace industry suggesting that there's no need for China to conduct its own manned lunar landing mission. Instead, China could cooperate with Apollo to skip the lunar landing altogether and start building a lunar base.

The Chinese space agency only knew that the joint lunar landing was about saving resources and speeding up the process. What they didn't know was that it also involved Yanjing's concept of lunar nuclear balance. They didn't have the authority to know about this.

All they knew was that whether it was the joint moon landing or the cooperation between the two sides, there was a mysterious force driving it forward.

　Collaborating with Huawei is a purely reality-driven decision. Unless readers want to read a tech-heavy novel where the protagonist handles everything, if we're talking about making real-world connections, this type of protagonist's company inevitably has to collaborate with Huawei. I saw some comments, so I'll explain.

　　
　
(End of this chapter)