Chapter 291: Before The Joint Moon Landing
“In this regard, I have already communicated a round with high-level officials at Huawei. They give 4 billion every year, and my original intention was to use 2 billion of it to offset with their computing cards.
Now that 500 billion from China Investment will arrive, then Huawei’s 4 billion every year will all be offset with computing cards.”
To put it this way, Apollo Technology is facing ZC that is even more severe than Huawei’s.
Even Bilibili, where Apollo Technology advertises, is being forced back to Hong Kong for listing, Huawei is not to that extent.
Plus Apollo Technology has fully demonstrated its ambition and ability.
Having ambition alone is normal, everyone has ambition, but you need strength.
Apollo Technology is different, it can force Old Musk to flip sides two years early.
Originally Musk could still hang on under the Donkey Party’s banner until mid-2024, then switch camps.
It shows that Apollo Technology has dealt real damage even to Prince Musk.
For the White House’s Cold War living fossil, he would rather let Huawei go find advanced process foundries than allow Apollo Technology to buy Nvidia’s AI computing cards.
But fortunately there is Huawei, Huawei launched its own computing card back in 2019.
Now it’s 2022, ChatGPT hasn’t been released yet, major manufacturers can still buy castrated versions of Nvidia AI computing cards, Huawei’s Ascend is in a state close to being completely ignored.
Lin Ran is willing to offset half of the advertising expenses with Ascend, this is also one of the important reasons why Huawei agreed to the 20 billion price.
Domestic top technology companies are willing to use their Ascend computing cards, this is also beneficial for perfecting their computing card ecosystem.
As for why not cooperate with other manufacturers, even Huawei’s Ascend ecosystem, backed by them, is still a blank slate, other manufacturers can be imagined.
“Actually, Huawei’s computing cards still have quite a big gap compared to Nvidia, alas, but indeed there’s no way, our current situation is that we can’t buy Nvidia’s computing cards, not even the castrated version.”
Pony smiled bitterly, he thought of how his office computer still uses Linux operating system, the entire Apollo Technology has a dedicated desktop support team to ensure everyone can use Linux comfortably, and he knew what kind of predicament they are facing now.
Huawei has been pushing HarmonyOS to Apollo Technology, saying you will definitely find HarmonyOS more usable than Linux, as for rocket design simulation software, fluid mechanics computation software and other industrial software that cannot be used on HarmonyOS system, we will solve it for you.
However, Huawei is promoting this matter, but they just say they will solve it, and haven’t seen it solved yet temporarily.
Lin Ran also smiled bitterly: “Yes, the problem is, even if Nvidia would sell to us, I wouldn’t dare to use it, who knows what could happen.”
Lin Ran then said excitedly: “But fortunately, I have communicated with Huawei, for us, their chips are already sufficient.
Because deriving material properties from element characteristics, this type of model’s data volume is very sparse, among the three elements of data, computing power, and algorithms, the dependence on data and algorithms is far higher than on computing power.”
Pony also knows quite a bit about artificial intelligence, Tencent digs big shots from the AI field every year without count, even though ChatGPT hasn’t emerged yet at this time, he hopes to learn more information from Lin Ran to provide direction for subsequent work: “Lin Sheng, explain in detail.”
Lin Ran further explained: “This is because data in the materials science field is extremely extremely limited, whether data sharing or acquisition faces unprecedented obstacles.
Experimental data produced by different laboratories won’t enter the same pool unless published in papers, of course if they want to enter the same pool, there will be all sorts of worries.
Because you can hardly guarantee that data provided by all research institutions won’t pollute the database.
If someone fakes data, it will pollute the entire data source.
From what I know so far, similar research data is very scarce, the most data is less than 4000 samples.
Feature engineering is the key to AI model success, but its design is particularly complex in material property prediction.
Physical element properties, like atomic weight, electronegativity, and material structures, like lattice type, bond length, all need to be converted into numerical features for the model to learn.
Among them, feature selection directly affects model accuracy, wrong selection may lead to performance decline.
Currently the entire process still needs to rely on researchers to manually handle feature values and do screening.
Extremely dependent on researchers’ experience and intuition, very likely to miss important information.
Nature’s sub-journal last year came up with a MODNet learning framework, which is a machine learning framework for material property prediction.
(“Material property prediction with limited datasets via joint learning with feature selection and MODNet” published on June 3, 2021 in Nature sub-journal NPJ)
They found that for predicting material vibrational entropy, reverse bond length and p valence electrons are key features, but manually identifying these features requires deep domain knowledge.
Extracting these data requires scientific laborers with sufficiently rich experience, while also ensuring data precision and reducing errors, the entire process is very tedious.
Because what we need to do is far more complex than what they did, we need to make a bigger, more complex model, summarizing and collecting feature data will definitely be very slow.
After all, this matter cannot be like cyberspace data, where you can ensure data accuracy through feature value elimination and various methods, its data, in computer terms, appears structured on the surface, but the core is extremely unstructured.
Therefore according to my estimate, at least for the first five years, Huawei’s computing cards will be sufficient.
As for after five years, Huawei’s computing cards will also advance with the times, plus we ourselves will cooperate with Huawei to advance their computing card progress.”
After listening, Pony could roughly sort out his thoughts, not saying he fully understood, after all, expecting a fifty-year-old to understand vibrational entropy, reverse bond length, and p valence electrons is still too much for Pony.
But the point Lin Ran wanted to express, he understood.
Pony said: “Lin Sheng, I have no intention of opposing cooperation with Huawei, similarly, I am very clear about the situation we face. Of course there are Cambricon, Alibaba, Baidu and other manufacturers with their own computing cards, but on one hand their computing cards manufacturing needs to rely on TSMC, on the other hand in terms of ecosystem, Huawei has gone the farthest, from a long-term perspective, they have the greatest determination and ability in long-term ecosystem building.
I am just sighing that the current situation we face is difficult.
Lin Sheng, I have a question, shouldn’t we collaborate with some university chemistry departments, physics departments and such on horizontal projects? Let them help us perfect our data pool?”
At this time on the market not only Huawei has computing cards, the several that Pony mentioned are all promoting them, but computing cards are not just about looking at hardware, the accompanying software ecosystem is equally important.
Why is Nvidia so dominant, doesn’t AMD also make AI chips? Why among America enterprises, AMD’s computing cards can’t threaten Nvidia? Nvidia’s moat lies in the ecosystem named CUDA built around the computing card over many years.
Similarly, Huawei has the determination to build HarmonyOS, in the computing card field, they are the best choice.
Plus everyone is a thorn in America’s side, huddling together for warmth is perfectly normal.
Lin Ran said: “Of course, I have thought about it, but not now.”
Apollo Moon Landing can even shear students’ wool, how could it not utilize China’s vast science and engineering students for building materials science artificial intelligence prediction models.
These are all high-quality pure natural laborers.
Better to do horizontal projects for Apollo Technology than for mentors, the former can at least really change the world.
“My idea is to wait until we finalize the entire data acquisition proposal, then expand outward starting with Shanghai Jiaotong University as a pilot.
To build this thing, it will definitely rely on domestic universities’ power, this is also our advantage.
Wasn’t it mentioned earlier that data is hard to acquire? Once there are rules, whether the standardization of acquisition or the elimination of dirty data, there will be methods to follow.
That is the time for domestic universities to participate on a large scale.
To put it this way, this set of artificial intelligence industrial software in the materials field will be our biggest moat.
Pony, think about it this way, if the metaverse can really become reality one day, virtual reality can really give people experiences like reality, the current physics engines that can only build animation effects definitely won’t be enough.
And the core of our industrial software will be the foundational basis for the future metaverse.
But this is a bit too far.
Short-term, I mean short-term as this century, our moat is materials field artificial intelligence industrial software, I call it Fuxi platform, the second is a series of self-developed industrial software around the aerospace field, we will build our own ecosystem on the basis of existing open source, the third is data, you also saw, now realizing moon landing is as simple as drinking soup for us.
One step fast, every step fast, lunar surface data, Earth-Moon transit data, crater probability in various moon regions, lunar landforms and so on, these data will also be our moat.
The first to the moon, we will also be the first to Mars, the first to spread our footprints across the entire solar system.
Industrial software ecosystem, space-related data, Fuxi platform, these three are the means to achieve our goals.
So talent around basic disciplines like computer, physics, materials, chemistry, mathematics, as long as it’s talent, we can cultivate them slowly, our plan even in the short term is measured in centuries.”
After finishing, Lin Ran paused for a moment then grinned and said: “Pony, the foundation of this enterprise aiming for the universe in the first five years is laid by you and me together.”
After listening, Pony, even though battle-hardened and seen much, still surged with unprecedented heroic ambition inside, he thought of a phrase: an old steed in the stable, ambition for a thousand miles.
Isn’t he right now Old Musk? From a little pony in the internet field back then to today’s old horse.
Pony finally understood what Lin Ran uses to attract Buzz Aldrin and other NASA Golden Age employees, it’s grand goals and unmatched achievements.
What Lin Ran is saying now is grand goals, manned moon landing and Lunar South Pole Shackleton Crater landing are unmatched achievements.
Both ignited the flame in Pony’s heart, he felt heaven had treated him well, the shadows brought to Tencent Empire by miHoYo and ByteDance seemed to disappear at this moment.
“Lin Sheng, rest assured, I will exhaust my mind and energy.” Pony said seriously.
As the second batch astronaut planned to land on the moon, Wei Xuhang’s excitement had long passed.
While orbiting the moon in the command module waiting for Lin Ran and Buzz Aldrin to return to the command module, he had guessed it would eventually be his turn.
But going as early as October next year was still a bit unexpected.
Including the moon landing method not being Apollo Moon Landing, but unprecedented segmented launch, Lunar South Pole landing, lunar fuel transfer, this was even more unexpected.
At first after knowing, mixed feelings inside, excitement, agitation, worry, apprehension various emotions, as someone who received astronaut training, training content not only operation but also scientific knowledge, this knowledge clearly tells him Lunar South Pole is a great place to build a base, but going there is particularly difficult, much harder than Apollo Moon Landing’s Sea of Tranquility.
Definitely worry, what if it’s a one-way ticket?
But only worry, no refusal, Wei Xuhang couldn’t refuse such temptation, China’s astronaut Li Cong couldn’t refuse even more.
Li Cong and Li Guang have trained together with Lin Ran, fifteen days of training, failed the exam miserably, after failure watching Lin Ran become China’s first person to land on the moon.
Not talking about Li Cong as a soldier, obeying orders is instinct, even if choosing himself, told success probability only 1%, Li Cong would still unhesitatingly choose to execute the mission, and upon receiving the mission say “guaranteed to complete the mission”.
So after learning Lin Ran personally selected him for this launch, Li Cong’s heart had only one feeling, I was selected by the professor, the professor really has vision! Damn I’m going to the moon!
Even knowing going to the moon uses a brand new method with unprecedented danger, his heart had no slightest wavering.
Inside Yanjing Aerospace City’s China Astronaut Scientific Training Center, Wei Xuhang and Li Cong are receiving training here.
Its predecessor was the high-altitude physiology research group under Chinese Academy of Sciences 581 group built by Dean Qian and Zhao Jiuzhang.
Later became astronaut training center, motto “From here to space”, code name Dawn.
The center has a dedicated high technology simulation hall, huge screen simulating the rugged edge of Lunar South Pole Shackleton Crater, rocky, long shadows particularly dangerous under low-angle sunlight.
Due to Apollo Technology’s successful Lunar South Pole landing, the simulation effect has new pictures and data added, increasingly close to actual conditions.
Wei Xuhang sits in the simulated lunar lander’s cockpit, hands tightly gripping the control stick.
His spacesuit though only for training, the heavy pressure makes him feel as if already on the moon.
“Wei Xuhang, watch altitude and descent rate,” the trainer’s voice comes through headphones, “South Pole terrain is complex, any deviation may lead to failure.”
He stares at the screen, radar shows a huge rock blocking the predetermined landing point.
“Obstacle discovered, adjusting trajectory.” Wei Xuhang mutters softly, fingers quickly operate lateral propulsion system, guiding the lander away from danger zone.
Fuel indicator light flashes, reminding him time is tight.
“Altitude 200 meters… 100 meters…” Wei Xuhang reports softly, voice focused.
Terrain on screen gets closer, he finally finds a flat area, carefully reduces speed.
Simulator emits slight vibration, screen shows “landing success”.
Trainer walks over, pats his shoulder: “Well done, your reaction is fast, but next time discover obstacles earlier.”
Wei Xuhang nods: “Understood, I will be more careful.”
Nearby astronaut Li Cong walks out from another simulator, removes helmet, smiles and says: “Xuhang, I almost crashed into a virtual rock just now.”
His tone relaxed, trying to ease the training tension.
Trainer facing the two adds: “Although Apollo Technology has achieved automatic landing, it’s impossible for automatic landing to be that smooth every time, you need to be prepared to take over the console manually at any time!”
At the other end of the training center, on a huge indoor simulation field, artificial craters and moon soil covered ground recreate the desolate scene of Lunar South Pole.
Lights adjusted to low angle, casting long shadows, simulating South Pole’s extreme illumination conditions.
The two wearing simulated spacesuits, holding geological tools, move carefully.
The trainer in charge of this subject beforehand stands on the observation deck, pointing to a rock and instructing: “This is simulated basalt, possibly found near Shackleton Crater, your task is to collect samples, ensure not damaging structure.”
“In the South Pole, shadowed regions may hide water ice. You must learn to identify possible ice layer features, this is also your most important task this time, find the existence of water ice.” The trainer continued: “If water ice can’t be found repeatedly in Shackleton Crater, then we may unfortunately have to change to another crater to build the moon base.”
In the training center’s low gravity simulation area, they are suspended by suspension system, simulating moon’s 1/6 gravity environment.
Wei Xuhang and Li Cong stand on slope covered with moon soil, simulating crater edge terrain.
Coach stands aside, holding tablet computer, recording their performance.
There is also low illumination operation training
Night falls, training hall lights adjusted to extremely low, simulating permanently shadowed regions environment at South Pole.
Wei Xuhang and Li Cong hold flashlights, attempting to install a simulated seismograph in the darkness.
“Visibility too low,” Wei Xuhang frowns, adjusts flashlight angle, “have to rely on laser rangefinder to judge distance.”
He turns on the device, screen shows: “Distance to target point 4.8 meters.”
The two move carefully, avoiding simulated rocks.
Li Cong connects the seismograph’s power cord, movements cautious.
Suddenly, trainer’s voice sounds: “Simulated spacesuit malfunction, Wei Xuhang, your oxygen system has a leak.”
Wei Xuhang immediately enters role, pretends to check spacesuit.
Li Cong follows training process, takes out spare sealing tape, simulates repair movement.
His movements smooth: “Repaired, continue mission.”
After completing installation, the two step back, check instrument operating normally.
In Yanjing days, training subjects cover everything.
Much more tedious than back when it was just command module.
Not difficult, but many subjects.
Since Apollo Technology’s successful soft landing at South Pole, inside China Aerospace there is a voice that there’s no need to do manned moon landing ourselves again, cooperate with Apollo Technology to skip moon landing and directly start building moon base.
China Aerospace only knows joint moon landing saves resources and speeds progress, what they don’t know is behind it there is also Yanjing side’s conception about lunar nuclear balance, their authority not enough to know this.
They only know whether joint moon landing or both sides’ cooperation, there is a mysterious force strongly pushing.
