After the layoffs, I became a tech godfather

Chapter 186: Mountain of Technology, Baiju Technology is serious!

The words "They have Xiaosha" silenced Intel's interim co-CEO David Zinsner.

In recent times, the two words "Xiaosha" have become the nightmare of technology companies across the United States. Wherever these two words appear, there will definitely be no good news.

Even after the failed assassination attempt led to the secret being leaked and all military communication satellites were destroyed, the entire America, including the White House, the Pentagon, and major confidential agencies, now dare not use modern communication equipment to transmit messages.

Now, to convey a confidential order, you have to drive all over the place to deliver it. Can you believe it?
Intel also received relevant urgent news that relevant American agencies need to redesign a set of internal non-public communication protocols that are completely different from all current communication protocols.

All the major companies are busy with this, but no one can guarantee whether it will work.

Mainly due to Xiaosha's ability...

"Xiaosha Xiaosha, it's Xiaosha again. How capable is Xiaosha? Will it ever end?" Jinsna finally couldn't help it and flew into a rage!
It's not just Zinsner. Huang Renxun, who is in NVIDIA's conference room at this moment, has a gloomy face that seems to bleed.

I originally thought that Baiju Technology was doing so many things, and each of them required computing power.

Although Xiaosha saves a lot of computing power both in training and deployment, every company deploys one, and each fine-tuned model requires personalized training.

Previously, only giant companies could train their own AI models, but now, very little computing power is needed to train and fine-tune them.

Although a single model requires less computing power, the total number of computing power cards required will definitely increase exponentially or even exponentially in a large-scale enterprise.

If there were no Moore's Law and Huawei Ascend, and if Xiaosha had not quickly improved the production capacity and efficiency of the production line, this would have been a potential customer of NVIDIA!
Even with Moore's Law and Ascend, Huang Renxun thinks Nvidia should be able to get a piece of the pie.

He had just asked a professional organization to evaluate that the total computing power required for what Baiju Technology wants to do is really too huge.

Even if we use Xiaosha to improve the wafer production line, it will definitely not be enough!

However, Huang Renxun didn't have time to be happy after seeing the data, as bad news came again - the three-dimensional metacognitive chip architecture of carbon-silicon fusion.

Originally, Huang Renxun was still praying that this thing was just a concept and would never be mass-produced in the near future. If his wish came true, Nvidia could still live comfortably until this thing was truly commercialized.

However, Hao Cheng did not even give him a moment to fantasize. As soon as Huang Renxun opened his eyes and put down his clasped hands, the press conference at the Linzhou Kingdee Sports Center Gymnasium gave him another heavy blow!
Hao Cheng's harmless smile looked so hideous in Huang Renxun's eyes that he gritted his teeth and wished he could eat Hao Cheng alive.

His bloodshot eyes were almost as big as quail eggs, and Hao Cheng's voice continued in his ears, as if the devil was whispering -

Hao Cheng said: "The full set of cores of the metacognitive architecture concept chip design is open to domestic companies. You can use this set of cores to design various chips, including mobile phones, computer CPUs, and various types of small chips.

“The ALU part can also be scaled up independently to form a high-computing power core for gaming cards and AI computing power cards for reasoning and training.

“As for the production issue, let me put it this way. Xiaosha has already provided a practical combination plan based on the current capabilities of our country’s chip production equipment and industrial facilities. It is expected that mass production will take place within 6 months!
"What is a combination plan? Give me an example..."

Hao Cheng briefly introduced the combination solution, and then gave an example of EUV lithography:
If we want to produce 3nm chips, we need to achieve mass production of EUV lithography machines.

EUV lithography machines cannot be mass-produced now. What is the problem? Xiaosha 2.0 has analyzed the problems after fine-tuning the lithography machine knowledge base and various production data. It found that the lens could not meet the requirements of the lithography machine, the lighting system was not stable enough, the control of the workpiece stage... and many other problems.

So, how to solve these problems?

For example, if the lens does not meet the requirements of the photolithography machine, Xiaosha 2.0 will conduct Xiaosha 2.0 (lens manufacturing fine-tuning model) after lens manufacturing and assembly, plus various data of the production process, and then find that there is room for improvement in assembly process and material purity.

……

Then the cycle continues step by step to complete the upgrade and improvement of the entire raw material and finally complete the assembly of the EUV lithography machine.

"Oh my god, it can actually be operated like this? That's too much! Doesn't that mean Xiaosha can improve everything?"

"Although this is true in theory, it requires too much computing power, and each step requires fine-tuning a professional model. The premise of fine-tuning the model is that you need to have a considerable amount of production data, which also needs to be accumulated."

"But it's not just that. Although Xiaosha's ability to fine-tune models is powerful, it also has limitations. It is still better at repetitive and regular work. Creative work also requires people, especially top professionals, to do it."

"But industrial production itself is repetitive work, so improving production equipment is definitely Xiaosha's strong point."

"Strong point? It's been two weeks since the strong point, and EUV is still not done?"

"Do you have any misunderstanding about time? If Xiaosha can't finish something that Huwei has not done in 10 or 20 years, then he is not strong if he can't finish it in two weeks? Do you know how many parts and systems an EUV lithography machine has?"

"I don't know. Do you know?"

"There are nearly 100,000 precision parts, and each part has its own components. It is a very complex integrated system. It is not as simple as you think!"

……

While the argument was going on, Hao Cheng spoke again: "Since we are taking the EUV lithography machine as an example, let me give you some information about the relevant progress:

"Two weeks ago, Xiaosha began to evaluate the problems faced by EUV lithography machines in mass production. From the overall to the details, we fine-tuned more than 10,000 related field models.

"Just now, the overall assessment has been completed and various plans have been implemented. Suppliers at all levels are working overtime to improve the parts production process based on the results of the model and with the help of the model.

“It takes about two months until the final assembly.

"Some people may ask, will it succeed? I can't say for sure, but I can tell you that it has a high probability! For EUV lithography machines, Xiaosha gave a probability of 99.999%.

"At every step of the model improvement production, Xiaosha's related fine-tuning models will be supervised, and the next step will not be taken if the standards are not met. In this process, Xiaosha will also continuously collect data to improve the production process of each step and each basic component."

As soon as Baiju Technology makes a move, the entire industry trembles.

"Guangquan, do you know what this means?" Kang Hui has become a little numb today. At a press conference of Baiju Technology, almost everything that came out was epoch-making.

"First of all, we can confirm that Xiaosha's ability is not creation, but a summary of 'experience'." Zhu Guangquan certainly didn't understand this, but Xiaosha's teleprompter had already told him:
"In other words, we can debug the lithography machine without Xiaosha, but it will be a little slower. We will start the machine today, test-produce a batch of chips, and then analyze all kinds of data from the production line. Then we will adjust the production line based on the data and enter the next debugging cycle.

"If we don't use Xiaosha, this round of debugging will take a week. Then, after 50 rounds of debugging, which is a year, some problems were solved.

"So with Xiaosha, what does this step become? After turning on the machine and getting the data today, Xiaosha will directly calculate and predict the next result. Then, based on the predicted result, it will calculate and adjust again..." In one night, Xiaosha can simulate 50 rounds. This simulation may be different from the production environment, and this can be used to calibrate the parameters in the next real production environment.

"This whole process can save more than 95% of time. So, Xiaosha's two months is equivalent to months, or three to four years.

"Of course, the actual process may not be that fast, because the speed of the production line cannot keep up with Xiaosha, and it will definitely lag behind. As for the 99.999% success rate, I think Xiaosha is still conservative, it must be at least 9 nines!"

Zhu Guangquan's explanation made everyone understand what this process was like.

"If that's the case, wouldn't it be invincible?!"

“The history of industrial development is about to be rewritten!”

“@Asmar Fukeli, what do you think?”

"Is there such a person? You are just talking nonsense here!"

"Hey, I'm happy to do it, what's it to you? ASML, that idiot, swore a few years ago that even if we were given complete drawings, we would never be able to build an EUV lithography machine.

"After seeing us solving the problem step by step, they changed their tune and said that we were reinventing the wheel and wasting resources by building lithography machines, and they wanted to condemn us for malicious research. I just want to see what this bastard has to say now!"

"What else can I say? There will be two lines in the future. China will have one line and one system, while Western countries such as Europe and the United States plus Japan, South Korea, Australia and a bunch of other countries will have one line and one system. Let's just play our own games!"

"Each one plays his own game? Without the Chinese market, they can't play a game! Let me tell you, the current system in Europe and the United States is mainly to go high-end, with a small amount of work, to grab a large amount of profit.

"It's the same with chips. We are the ones exporting low-end and mid-range products. They don't have the scale to do it. If we conquer the high-end market, will they still have room to reap high profits?

"Profit margins have been squeezed, and R&D costs must be reduced. Subsequent iterations cannot be guaranteed at all. There is no other way but collapse!"

"But why can't we also reap huge profits? Why can we make the prices of an industry as cheap as we can just by making a breakthrough? Isn't it better to make money without doing anything, like Europe and the United States?"

"It's not that simple. The market originally belonged to others. We are latecomers and we have to grab it! We can only monopolize it if we destroy all of them and come up with something that no one else has."

At this moment, a large number of netizens instantly became experts and began to analyze our various strategies. One of the widely recognized conclusions is: Even if we don’t make too much money, we must bring down the Western industrial system. This is not to harm others and not benefit ourselves, but the contradiction is irreconcilable. In the end, only you can survive.

Netizens were busy chatting and enjoying this brief joy.

The media outlets were also busy.

Just two minutes after Hao Cheng finished introducing the relevant things, CCTV News's news flash came out: "Xiaosha solves more than 10,000 key bottleneck problems, domestic EUV lithography machine mass production speeds up, and the first unit is expected to roll off the production line in March."

Celebrity Daily: "Xiaosha Fine-tuning Model: Good news for all walks of life, helping to achieve another breakthrough in new quality productivity!"

Reference News: "Moving into a new era, the battlefield of intelligent revolution has begun"

……

There was jubilation in the country, but America was already wailing. The various affiliated companies of the New Liberal Party were completely unable to sit still. They didn't care whether Xiaosha would find out or not, and quickly held an online meeting.

Someone raised a thought-provoking question: "We are also developing AI, not to mention the level, but why can our AI only talk, draw pictures, and make videos, while Xiaosha is so versatile that it can do everything?

“Industrial design and chip factories have been improved. Now even the entire chain of EUV lithography machines can be fine-tuned! Why is this?”

Compared with Xiaosha, the previous AI seemed a bit unprofessional.

"Industrial AI is actually a pseudo-concept. Its essence is industrial automation. It is not like a large language model, where neural networks are allowed to extend and calculate everything. However, Baiju Technology should have broken this limitation."

"Can't we open it then?"

"I can tell you're an amateur. You don't even use AI dialogue much. Don't you know about AI hallucination? There are no restrictions on language models because there's no harm in making mistakes.

“For example, they can draw a portrait with six fingers, ask legal questions, make up laws that don’t exist, and assist in writing code to make up methods that don’t exist. If you check it once and make the changes, there will be no problem.

"But if you think about it in the industrial field, if there is a problem that is not discovered, it may lead to a major accident."

The essence of traditional AI is calculation, combination, and probability, not real thinking.

But it is hard to say whether the things put together by AI really exist and can be used.

"So, Xiaosha is completely different from our previous AI, right? We will never be able to catch up, right?"

This is a serious injury. I have always wanted to find out the principles behind Xiaosha's training, but there has been no progress so far. I don't dare to do or even think about anything extreme.

What will happen in the future? It's all uncertain!
……

Meanwhile in Linzhou, the press conference is still going on, and the last product, Mo Shu, is about to be unveiled.

Xiaosha’s own PPT summarized it this way:
Xiaosha 2.0: A tool used by ordinary people to help improve work efficiency.

Xiaosha 2.0 fine-tuning model: a personalized industry AI model used by professionals, allowing all industries to quickly improve to the world's most advanced level.

Mo Shu: The best assistant for scientific researchers to make further breakthroughs and progress, and the glue for the integration and improvement of cutting-edge disciplines.

"Ah, there's a big trick here!"

"With these layers of advancement, Mo Shu is the ultimate weapon!"

"Whether it's Xiaosha 2.0 or fine-tuning, it can't break through the highest level in the world today, but Moshu looks like it's a good thing that has gone a step further on the basis of the top level!"

"The mountain of technology, Baiju Technology is serious!"

(End of this chapter)