After the layoffs, I became a tech godfather

Chapter 243 Decoupling
Because the initial assessment of public enthusiasm was severely underestimated, some employees at Baiju Technology were working overtime all night for the first time and were already exhausted.

Even more troublesome than them are Intel and Qualcomm.

At this press conference, it was expected that Chinese domestic chip design companies, led by Baiju Technology, would be able to design and manufacture chips that are at or even surpass their level.

After all, we've already had so many vaccinations.

but--

"Tell me, what do you mean by 'You're only slightly better than a Core i9, how dare you stand on the same stage as Baiju Technology?' What do you mean by that? What's wrong with the Core i9? Is it really that bad?"

Intel's new CEO, Chen Liwu, took office just ten days ago, and Intel's expectation of him is to turn the tide.

But only ten days later, a huge mountain came crashing down on them.

When he got the specifications of the 3Z-9000 CPU, a huge weight was finally lifted off his shoulders, hitting him squarely on the head—his heart was no longer in suspense, because it was already dead.

Since the current 3Z-9000 does not use a metacognitive architecture but rather a traditional x86 architecture, its performance on paper is easy to calculate. It mainly involves frequency, core count, cache size, and power consumption!
Loongson didn't use any fancy terms, just one sentence: In the same size CPU, we've packed 32 dual-threaded cores, with a maximum frequency of 6.8GHz, and the power consumption at full power is only half that of the Intel Core i9-14900.

Chen Liwu silently calculated that, under the Windows system, the overall performance was more than 1.5 times that of the i9, based on numerical values ​​alone.

In the case of an unbounded system, the multi-threaded scheduling capability will be further enhanced by Xiaosha, potentially doubling in efficiency.

That's not the worst part. The worst part is that the first reaction from the Chinese market to this CPU was dissatisfaction!

"Yeah, what is this thing?!"

"What are the differences between it and Intel?"

"You've wasted Xiao Sha's time, all you've produced is this!"

……

Chen Liwu grew up in Singapore and naturally understands Chinese. He also often browses simplified Chinese websites.

He knew that when Loongson first made consumer-grade chips, even though their performance was only at the level of Intel's entry-level chips ten years ago, they still received tremendous support.

But now, in just a few months, everything has been turned around.

"White Horse Technology, the Hand of God?" What could Chen Liwu do? Turn the tide? He didn't think he had that kind of ability! If that's the case, why did he come back to Intel at this critical juncture?
"I believe our only chance lies in developing a second demand!" Chen Liwu said to the chairman and all members of the board of directors.
"Only by avoiding direct confrontation can we have a chance to survive. As for what constitutes a secondary demand and how to develop it, the marketing department needs to conduct research as soon as possible. And we currently have another advantage."

"Advantages?" Chairman Frank Yerry was still puzzled.

"Hmm," Chen Liwu said, "Nvidia will definitely be the first to die, because Moore's Threads is too ambitious and seized the opportunity given to you by Baiju Technology to quickly complete a technological leap."

"However, there are clearly no companies making CPUs yet. You could say that this is our opportunity."

What kind of opportunity is this? It sounds like the difference between a death sentence and a suspended death sentence, and this suspended death sentence isn't life imprisonment, but rather the kind where the death penalty is carried out after the suspended death sentence.

But then again, dying later is better than dying earlier.

Frank Yerry nodded slightly, acknowledging Chen Liwu's statement and the idea of ​​using market research to explore secondary demand. What else could they do? They might as well try anything, no matter how desperate!
……

After watching the press conference, in Beijing, the interim meeting of the High-Level Leading Group on Economic Issues continued, and the key word became two words: "decoupling".

He Qiang said, "This is not what we wanted, nor was it our original intention. But since it has come to this point, we only have one choice—to do it ourselves!"

The original plan, established before Hao Cheng's arrival, was to divide and conquer, taking them one by one.

However, it now appears that such meaning does not exist.

According to Hao Cheng, even if the whole world were to be brought into one camp, it would be of little help to us at this point.

Ultimately, the issue of the relationship between machines and humans still needs to be addressed.

Essentially, there is no fundamental difference between 14 billion and 70 billion.

Therefore, it is pointless to try to win over anyone.

"Therefore, our first step is to conduct a general survey of the import and export products of our foreign trade enterprises, and then address the problems in a targeted manner."

"Little Sha, Little Sha!" He Qiang said, glancing at Hao Cheng, who then called out to Little Sha.

【current……】

A census? Nothing beats a quick survey!
Under the current business logic, almost no slightest disturbance can escape Xiao Sha's notice, especially since almost every company cannot do without Xiao Sha.

Xiao Sha summarized a whole bunch of things in one go and directly formed a report.

This includes which industries rely on imports for which raw materials, components, and entire products, and what domestic alternatives are currently available.

And how long will it take us to resolve this issue, and what kind of help Xiao Sha can provide in this regard?

Take the engines of China's domestically produced large aircraft as an example. COMAC's C919 is equipped with the LEAP-1C engine from CFM International, a joint venture between America and France.

However, the CJ-1000 developed by Aero Engine Corporation of China is currently preparing for verification flights.

In theory, such verification would take a very long time.

It usually takes three to five years, and if something goes wrong in the middle, it will take even longer.

So what to do? A world model!

Yes, the version answer is the world model!
"Oh, I've heard of that! Many companies have mentioned it!" As soon as Hao Cheng mentioned this concept, someone chimed in.

"Yes, this concept has been mentioned before," Hao Cheng said. "The reason is simple: we can't solve a core problem in the field of AI. For example, I want to train autonomous driving, but the amount of data is insufficient."

"What do we do? We create scenarios and simulate various situations according to the physical laws of the entire world. The definition of a world model is the same: what is a world model? It is a dynamic system model for studying problems!"
"That's the right way to think about it. But how do we do it? What kind of dynamic method is needed for a system dynamic model? To simulate the physical laws of the world and create the scenarios we need, is two-dimensional computing power enough? It's simply impossible!"
"I can say with certainty that before the upgrade to the new core, Xiao Sha was willing but unreliable in his efforts in this area."

Hao Cheng's words were forceful and resounding. While seemingly referring to Xiao Sha, they were actually a critique of all companies that used the so-called world model as a pretext.

Because this is fundamentally a technical concept that cannot be deeply explored. Just because you claim to be trained using a "world model" doesn't mean you are automatically stronger.

Advanced concepts do not equate to advanced technology.

If we're just talking about concepts, isn't the concept of the "100-year goal at the summit of technology" in the first background image of Baiju Technology's official website even more advanced?
This includes using AI to completely replace the human immune system to help the body identify viruses and abnormal cells, and then using nanorobots to target and eliminate them, so that humans will no longer be troubled by diseases.

Ultimately, whether advanced concepts can truly lead to technological leadership requires more than just having the concept itself.

As for world models, the key question is how closely you can imitate the physical world.

Can something created using ALU logic operations simulate 30% of the data?

"The figure of 30% is my estimate, because our previous model could get around 40%," Hao Cheng said.

"Okay! Then other companies are at most 20%!" This was the sentiment of almost everyone present. If Xiao Sha is only 40%, then for other companies, 20% would be considered high. Perhaps only Huawei could be a little higher, reaching 30%? That's not necessarily true!
This is not prejudice, but a basic understanding that everyone has already formed.

"However, according to our estimates, a three-dimensional metacognitive chip can achieve 60% at the outset, and after a period of training and simulation, it should be able to reach 95% of the level of the real world that we already know."

The figure is 95%, but Hao Cheng added a lot of qualifiers here.

why?
Because our understanding of the world and the universe is still very limited.

Humanity's own scientific system is constantly being updated and improved.

Some theories are simply generalizations based on phenomena, without touching on the essence of things.

At this point, expecting a model to uncover the essence is definitely wishful thinking.

However, based on existing theories, there is great potential to simulate a near-complete world model.

For example, the verification flight of the aircraft engine that Xiao Sha just mentioned.

"It can be said that Xiao Sha only needs one week of verification to achieve the results of your ten years of verification!" Hao Cheng said with such certainty.

"Hiss!" As an AI, it's acceptable for everyone to ask Xiao Sha questions, since even if the answer is wrong, the damage won't be significant.

However, when it comes to flight verification, the inherent caution of the Chinese people still prevails to some extent. What if? After all, the simulation level is only 95%, not 99%, much less 100%, right?
"There's no 'what if'!" Hao Cheng shook his head: "Let alone 95%, even in 40% of cases, it can play a huge role in certain specific fields. For example, why is our chip mass production so fast?"
"It was because we strengthened the simulation and verification process of the world model. At that time, although its ability to simulate the world was only 40%, it could reach almost 90% when it came to the production line."

"The larger the world and the more variables there are, the more complex the model itself will be. We can simulate the world up to 95%, but in practice, we don't need to completely simulate the world; we only need to simulate a part of it."

"For example, when we conduct flight verification experiments, we need to control variables. Once the variables are controlled, the accuracy of the model will be improved infinitely. Let's take production line simulation as an example again. With Xiao Sha's current capabilities, the simulation of the production line can almost reach seven nines, that is, 99.99999% simulation accuracy."

"This is already at the level where simulation can be used as a trial production."

The group of people still seemed a little uneasy, after all, it was a large aircraft, and they had to be extremely cautious.

"There's another point," Hao Cheng said, "but the most crucial thing is that we have a backup plan. For chip manufacturing, he could say that simulation can replace tape-out and trial production, but that's definitely not the case for large aircraft."

He continued, "There's one more thing I think you've all misunderstood. It's not that just because it's been verified in the world model, it doesn't need to be verified in the real world! It still does!"

"Huh?" Someone on the scene hadn't figured out where they came from, wondering why they'd come full circle: "Aren't we back to where we started?"

"What do you mean by going back to the starting point?" He Qiang finally understood: "Although it still needs to be verified in the real world, the verification results in the world model can guarantee one thing, that is, everything you do in the real world can be successful the first time!"
"A world model can help you solve all the problems in the early stages. But what it actually solves is the issue of trust, and at least for now, we definitely need real-world validation."

Hao Cheng nodded slightly. Even if the world model could be almost 100% accurate, Hao Cheng couldn't say that the security verification would be 100% problem-free.

In fact, even real-world verification cannot guarantee anything. Boeing, as an aircraft manufacturer, has been proven to have problems more than once or twice.

It's fair to say that if we actually went through the world model once, we might be able to reduce a lot of problems.

However, this is the current understanding, and it is also the safest approach.

The world model is continuously adjusted during the verification process, eventually providing a final answer to the real world. Design and manufacture according to the answer, and after a long period of verification, all adjustment time is saved.

"You'll need to prepare a report for the National Development and Reform Commission and the Ministry of Industry and Information Technology later." He Qiang nodded heavily and said, "This involves more content related to industry, especially advanced industry."

"Economic issues are urgent. Stimulating demand requires policies, while addressing shortcomings requires technology. You have an unshirkable responsibility to get involved in this. I have just consulted with the relevant leaders, and we will meet here tomorrow morning."

Hao Cheng was completely numb, but what could he say at this point? What else could he do? Just do it!

"Well, I just took a quick look. We currently rely heavily on foreign industrial products. Now that we've solved the semiconductor problem, the issue of precision instruments shouldn't be too big."

"Then, for aircraft engines and similar products, we provide limited models for rapid processing. As for medical devices, Baiju Technology has already been communicating with major domestic manufacturers."

"What's left are all resource-related items, and Baiju Technology can't be of much help with that right now."

"Alright, stop worrying so much. Do you think we're just sitting around doing nothing?" He Qiang laughed and joked, "Precision instruments, large industrial products, and medical intermediates—if we can solve these issues in a year, then there won't be any problems!"

(End of this chapter)