Rebirth 08: Rise from copycat phones

Chapter 584 New Cutting-Edge Technology: Intelligent Brain

Every autumn is a time when news about the high-tech industry is at its peak.

Major high-tech manufacturers tend to release a variety of new products during this period.

The traditional smartphone industry is a typical example. In August, Zhiyun released the S series, and September and October were the period when manufacturers such as Apple, Samsung, Vcool, Huawei, OPPO and Vivo, and Xiaomi successively released their flagship models.

In addition, Vcool Electronics released their first-generation smart electric vehicle, the Blue 2020. Fruit, Sixing, Vcool Electronics, and Huawei successively released VR virtual reality devices in October.

There are also many other high-tech products!

This led to the concept of the autumn peak season.

This situation is the result of a combination of factors... It is related to the process cycle and timeline of the upstream semiconductor industry... Taking Zhiyun Microelectronics as an example, in recent years, the technology verification has been completed by the end of the year, followed by small-scale production in the first half of the year, and large-scale mass production only begins in April or May.

This is also a new process chip from Zhiyun Group, and it won't be available for mass production until May or June... no earlier than that.

Similar situations exist with Apple and TSMC, whose business models are somewhat similar to those of the Zhuyun Group.

In addition, the factors that lead to a slow season in the first half of the year and a peak season in the second half are also related to domestic holidays.

Because there is a holiday every year during the Spring Festival in China... the entire domestic industrial chain will enter holiday mode before and after the Spring Festival. In addition, there will be a large number of employees moving around at the end of the year and the beginning of the new year... many people resign before the New Year or after the New Year.

Supply chain disruptions, coupled with employee turnover and other factors, meant that large-scale shipments were impossible in late February or March, right after the Lunar New Year... It was too late.

Many factors combined make autumn a very lively season every year, with all sorts of new things appearing one after another.

Similarly, this is also the period with the highest annual revenue and profit for Zhuyun Group!

In October, Xu Shenxue also received the operating report for the third quarter of this year... not a publicly released financial report, but an internal operating report.

"In the third quarter of this year, thanks to the positive impact of the launch of the S20 series phones and the back-to-school season in August and September, our smart terminal business segment performed very well, with total revenue reaching US$105 billion, an increase of 8.5 percent compared to the same period last year!"

"This data does not include software and internet services that rely on smart terminals, such as the revenue from the Yun APP store, which is calculated separately."

"Our smart terminal business usually contributes about one-third of our revenue in the third quarter each year. This year's revenue in the third quarter of autumn is at a normal level. Based on this data, the revenue of our smart terminal hardware business segment is expected to reach $310 billion this year."

Ji Chenghe was quite satisfied with the data and reported it with confidence!
"Among the other core business segments, the virtual device business maintained rapid growth, mainly benefiting from the increased production capacity of APO series graphics cards from the group's subsidiary, Zhiyun Microelectronics. Our virtual device shipments continued to increase, and we expect further growth in the fourth quarter!"

"According to our calculations, our virtual equipment business returned to profitability in the third quarter and has begun to enter a period of rapid profit growth!"

"After all, with the continuous large-scale shipment of virtual devices, our initial R&D costs, production line investment costs, and component development costs are being rapidly diluted. Once these huge initial costs are diluted, the gross profit margin of our virtual devices will be quite considerable in the later stages!"

"At the current pace, our overall virtual equipment business will enter a profitable phase this year, with estimated profits of around five billion US dollars. However, by next year, our virtual equipment business will experience explosive growth in profits, potentially bringing in at least two hundred billion US dollars in annual profits, becoming a new major profit growth point for our group!"

"In the robotics business, sales in the third quarter of this year increased significantly compared to last year, resulting in a substantial increase in overall revenue. Although the average selling price decreased, our gross profit margin did not decrease; on the contrary, it increased. Furthermore, many of the previous upfront costs have been gradually diluted. Therefore, the robotics business contributed at least $8 billion in profit to us in the third quarter, and we expect full-year profit to exceed $35 billion!"

"Our fastest-growing and most profitable business is still our computing chip business. In the third quarter, the shipments of our main computing chip, the APO series graphics cards, continued to increase, and the sales and revenue of our main APO5000 were very good!"

"In addition, the sales and profits of the APO4600S and APO4500M computing power graphics cards are also quite good!"

"In addition, revenue from a series of computing chips, including EYQ chips, EYEQ chips, PX chips, ZY chips, and LC chips, is also continuously increasing!"

"In the third quarter, our revenue from computing chips reached nearly $80 billion, and our gross profit was around $60 billion. This business is our most profitable core business, and its revenue and profit contribution has exceeded that of the smart terminal business!"

"Furthermore, our computing chip business segment has strongly supported the development of our robotics business, virtual equipment business, and sister companies such as Hailan Auto, Yihai Technology, and Weiku Electronics!"

"Our computing chip business has become the core business of our group, and its importance continues to grow!"

"As for Zhiyun Microelectronics, which supports the computing chip business, in addition to internal orders, its contract manufacturing revenue and profits are also rising continuously. In particular, our 14-nanometer and 7-nanometer process nodes have brought us a lot of contract manufacturing revenue!"

Today, Zhuyun Group is no longer the smart terminal manufacturer it once was... Although smart terminal business is still Zhuyun Group's core business, it is only one of its many components!

The same core businesses also include three major segments: semiconductor design and manufacturing, intelligent robots, and virtual devices.

The current Zhuyun Group relies on four pillars to operate... which has led to Zhuyun Group's exaggerated revenue and profits, as well as its even more exaggerated market value!
However, although Zhuyun Group earns a lot, it spends even more. Its daily capital expenditure in the semiconductor field is several tens of billions of dollars, and its various other R&D investments are also extremely large.

A significant portion of the money that Zhiyun Group earns each year is spent on driving continuous technological progress.

Moreover, there are all sorts of random technologies... and they cover a wide range of fields, not just limited to a few core businesses, but also exploring all kinds of cutting-edge technologies.

For example, the virtual business unit under Zhiyun Group attaches great importance to non-contact brain-computer interface technology, and began to invest heavily in it last year. The estimated R&D expenses for this year alone have exceeded five billion US dollars.

Quantum computing projects are another area where Zhuyun Group invests heavily. The annual research expenditure in the field of superconducting quantum computing is in the billions of dollars, and the R&D investment in photonic quantum computing exceeds two billion dollars.

Still in the field of computer science, Zhiyun Group also has a neuromorphic computer project. Ordinary people don't know much about neuromorphic computers, but it is a relatively cutting-edge new direction of computer research.

Simply put, this computer uses electronic chips to simulate the operating mode of a biological brain... It has many special features, such as being able to perform large-scale training with a small amount of training data, which means it has a powerful self-learning ability; it can perform low-precision fuzzy calculations to expand its task processing capabilities; and it has event-driven capabilities, meaning it only starts when there is a task and does not run when there is no task, in order to reduce power consumption.
And these also bring about a very important point, which is currently of great concern to scientists: extremely low power consumption!
Especially in the field of artificial intelligence, the power consumption of these chips can be controlled to one-thousandth or even lower compared to traditional computing chips.

The power consumption of a biological brain is very low.

The human brain has approximately 85 billion neurons, connected by 10 quadrillion synapses, and can perform 10 quadrillion operations per second. With such a massive number of neurons, its power consumption is only about 20 watts when handling huge computing tasks on a daily basis... This power consumption control is nothing short of miraculous in the field of computers!

This is a masterpiece created by nature's evolution over billions of years.

Scientists are currently trying to use electronic chip technology to create a neuromorphic chip that mimics the operating principles of a biological brain.

Note that this is not a biological brain; it is still an electronic chip... only its operating mode mimics that of a biological brain.

Zhiyun Group started its research and development in this field very early... After all, Zhiyun Group is involved in artificial intelligence, and it is interested in and researches any technology that can provide artificial intelligence capabilities.

Both quantum computers and neuromorphic computers are being researched.

However, the technological breakthrough in quantum computers was faster and earlier, and the YCL-51 superconducting quantum computer is already in use.

However, research and development in neuromorphic computers has not stopped; on the contrary, the investment has increased year by year. It's not just Zhiyun Group that's involved; some companies, such as Intel, are also doing it. Intel started earlier and even launched its so-called first-generation neuromorphic chip a few years ago. However, they were a bit too hasty, and the neuromorphic chip they produced was too primitive and had no practical use, so they rushed to release it.

So naturally, it didn't cause any stir, let alone be used on a large scale.

The Zhiyun Group actually developed this level of neuromorphic chip a long time ago, but they felt the technology was too primitive and not very useful for the time being, so they didn't announce it.

Neuromorphic computing mimics how the biological brain works, but current research on the biological brain is still in a very primitive stage... It becomes extremely difficult to imitate something before we fully understand it.

The reason why neuromorphic chips have failed to gain widespread popularity is actually due to a core factor: a lack of applications... or rather, scientists have not yet found a technology that can efficiently leverage neuromorphic computing.

Unlike traditional GPUs, which have matured from deep learning to various artificial intelligence training and learning technologies, the GPU ecosystem led by Zhiyun Group is very comprehensive.

However, neuromorphic computing currently lacks corresponding application technologies...

This is also why the industry has been working on neuromorphic computing for many years, but hasn't made any real progress... Not only is the technology too primitive, but people also don't know how to use it!
However, today, Xu Shenxue received a report from the neuromorphic project team under Zhiyun Semiconductor, saying that they had developed the third-generation neuromorphic chip, and that the newly developed sixth-generation neuromorphic chip already had preliminary practical application value... especially in the field of artificial intelligence.

Upon receiving this news, Xu Shenxue went directly to the project team in the adjacent building and saw the sixth-generation neuromorphic chip immediately.

Fu Zhengyang from Zhiyun Semiconductor introduced some basic information about the project to Xu Shenxue: "Chairman Xu, the chip we are seeing now is the latest success of our neuromorphic computing project team. It is the product that the project team has iterated six times!"

"In terms of chip design, we have imitated the structure of a biological brain and designed a unique Smart Brain 6.0 architecture. We have also used our group's most advanced five-nanometer process for production. This architecture has 512 neuron processing cores, each core has 40,000 neurons, and a single chip has 20.48 million neurons!"

"From a purely hardware performance perspective, this represents a significant improvement over our previous fifth-generation technology. In the previous fifth-generation technology, a single chip had only 256 neuron processing cores, with each core supporting only 10,000 neurons, for a total of only 2.56 million neurons!"

"Through a new chip architecture and more advanced chip manufacturing processes, we have increased the number of neuron processing cores to 512. More importantly, we have increased the number of neurons supported by each neuron processing core from 10,000 to 40,000. This is a huge technological advancement!"

"Ultimately, we obtained this sixth-generation neuromorphic chip with 20.48 million neurons."

Xu Shenxue looked at the chip. It was estimated to be about 100 square millimeters in size, which was not very big... only slightly larger than the S1103 chip used in the Zhiyun S20 Max phone!
From the outside, it looks no different from an ordinary chip... but Xu Shenxue knew that its internal structure was very different from that of a traditional chip.

One important feature is that this chip has its own memory and does not need to rely on external memory to operate.

Fu Zhengyang continued, "The computing power of a single chip is still insufficient, so we designed a multi-chip architecture and used our advanced 3D packaging technology to package 120 chips together, thereby providing greater communication bandwidth between the chips. Ultimately, we obtained this larger chip: the Brain 6 chip!"

At that moment, Xu Shenxue saw a chip that was quite large... It looked like a big round plate.

Fu Zhengyang said, "Our engineers use a Smart Brain 6 chip, along with other auxiliary chips and a motherboard to form a basic server unit. Then, through a specially designed communication architecture, they connect eight server units to form a neuromorphic computing server."

"A single AI Brain 6 server can provide 19.6608 billion neurons!"

"As long as five smart brain servers are connected, we can obtain 98.3 billion neurons, which exceeds the 85 billion neurons in the human brain."

"If we build a server cluster using this Brain 6 server, we will be able to obtain a large-scale neuromorphic computing center with considerable computing power!"

"This kind of intelligent computing center can already effectively run many of our top-level artificial intelligence algorithms, as well as some more common commercial algorithms, such as Yun AI and GTAI!"

"Besides the technological advancements in hardware, the most important thing is that Professor Lin Andong personally designed and developed the fuzzy low-precision computing architecture for the intelligent brain system, specifically targeting the neuromorphic computing model!"

At this moment, Professor Lin Andong, who was standing to the side, also spoke up: "The fuzzy low-precision computing architecture that our research institute has recently developed is an algorithm architecture that simulates the operating mode of the human brain and is specially matched with neuromorphic computing!"

"Our design goal is to focus on the field of low-precision computing and bring out the best in fuzzy computing. Therefore, our AI computing system will not and cannot be used alone for training or running artificial intelligence. Instead, it needs to be used in conjunction with our next-generation superconducting quantum computing system."

"In this way, we can simultaneously obtain the low-precision computing that the Brain 6 system excels at, and the high-precision computing of the superconducting quantum computer center, in the training and operation scenarios of artificial intelligence. By combining the two, we can obtain tremendous hardware performance support in the field of artificial intelligence!"

"This will result in a synergistic effect, where one plus one is far greater than two, leading to superior AI training performance and significantly reduced energy consumption!"

"We have calculated that if we combine the intelligent computing system and the superconducting computing system, we can reduce the computing power consumption by more than five times for the same training task!"

"More importantly, our Brain 6 computing system, with its unique low-precision computing mode, plays an irreplaceable and significant role in multiple fields of artificial intelligence training and operation!"

Neuromorphic computing is characterized by its low-precision computing and low power consumption, which is the direct reason why major companies are so enthusiastic about it.

The technological breakthroughs made by Zhiyun Group in the field of neuromorphic computing are no exception, specifically targeting these two characteristics.

Professor Lin Andong continued, "According to our calculations, by simultaneously employing quantum computing, neuromorphic computing, and GPU computing, we may be able to train an artificial intelligence system with groundbreaking performance!"

"Because this kind of artificial intelligence system will simultaneously possess the characteristics of human fuzzy logic and the high precision characteristics of traditional electronic computers!"

"Although it is still impossible to directly break through to the field of strong artificial intelligence, we can still improve the current low artificial intelligence by at least one or even two levels!"

"This will help us obtain a new generation of artificial intelligence models that are more efficient, more accurate, and have lower power consumption. If applied to the field of robotics, we may even be able to obtain a humanoid robot that is indistinguishable from a real one by ordinary people... Of course, this anticipated new algorithm model will still be in the field of low artificial intelligence, and the robot will naturally not be able to develop self-awareness or anything like that!"

As Xu Shenxue listened, he also flipped through the technical documents in his hand!
He wasn't familiar with the technical details, nor could he possibly know... These were the most advanced technological achievements from multiple departments within the Zhiyun Group, especially including Professor Lin Andong's latest research findings...

Professor Lin Andong's research results are basically like cryptic texts. Not to mention laymen like Xu Shenxue, even the other top mathematicians in the Artificial Intelligence Research Institute would not be able to understand them without Professor Lin Andong's personal explanation.

So he skipped over these technical details and went straight to the results...

After reviewing the documents for a few minutes, Xu Shenxue said, "Now that the technology has reached the stage where it can be put into practical operation, let's get a batch of these Smart Brain 6 computers and use them together with our YCL-51 superconducting quantum computer and our APO6000 graphics card to see how they perform!"

"Go ahead and do it with confidence. Don't worry about funding or manpower; the group will provide ample support!"

Xu Shenxue didn't understand the detailed technical aspects, but he could comprehend the enormous application prospects described by Professor Lin Andong and Fu Zhengyang...

Now that we've come up with this new cutting-edge technology, we have to put it to use!

Upon hearing Xu Shenxue's words, both Fu Zhengyang and Professor Lin Andong smiled. The two of them were working together on this project. Fu Zhengyang was leading a team to develop the hardware for the neuromorphic computing chip, while Professor Lin Andong was leading a team to develop the algorithm architecture for neuromorphic computing.

The collaboration between the two parties ultimately led to the development of the Brain 6 system.

Now that the project I've worked so hard on has finally yielded results and gained recognition, I'm very happy!
(End of this chapter)