Chapter 317: Large Model of the Human Brain

While Lin Jia was meeting with senior executives of automobile companies in Haikou, Chen Yuanguang had already gone to Yanjing to make a detailed report.

There is a lot of talk among the Chinese community about open source technology.

The reason is simple. Both external forces and internal private enterprise forces hope that technology can be open source.

If we simply adopt a cooperative approach, then there are a lot of variables as to who goes first and who goes later, who can cooperate and who cannot.

Although Robin publicly stated that Baidu and Guangjia Aerospace are strategic partners, he actually has no idea whether they can gain the upper hand in this round of unmanned driving.

Even if he spoke with certainty during the interview.

Compared with the black box with great uncertainty, private enterprises naturally prefer open source technology, which means that the technology itself is also in their hands.

The bureaucrats who hold this view also have strong theoretical support. Internet technology has always been open source, and America's route and source code for artificial intelligence technology have never been hidden.

In addition, Chen Yuanguang believes that technology should be open source and we should listen to the opinions of professionals. Is there anyone more convincing than Chen Yuanguang, who leads the technology research?

This kind of view has always existed, but people are not brave enough to express it publicly, and at most they just say a few words within a small circle.

When everyone knew about Chen Yuanguang's public opinions to Yanjing, the discussion began to get louder.

They even dared to make statements through the South China Morning Post.

Both the South China Morning Post and the United Daily News can be seen as an unofficial channel for spreading messages, hoping to create public opinion from the outside to achieve their own goals internally. However, as East Asia's national strength increases, this method is becoming less and less effective.

Lin Jia was holding a meeting in Haikou and Chen Yuanguang was holding a meeting in Yanjing.

“Hello everyone, I think you should all be aware of the robot driver in Shenhai recently. There are various rumors online.

The robot is called Tiedan, and it is not the Terminator as foreigners say.

We have been communicating with relevant parties since the project was launched, from Tiedan testing on the test section to communicating with the Ministry of Industry and Information Technology, the Ministry of Transport and other departments to obtain their approval.

Including the subsequent formal operation as an online-hailing car on urban roads and the communication with relevant departments, Tiedan even obtained what should be the only robot driver's license in the world.

This has been fully communicated with the traffic management department.”

Chen Yuanguang said that he wanted to introduce the whole story first.

“Then I want to talk about the underlying technology. It’s not actually driverless, but more precisely, it’s a large model of the human brain.

We call it HBM internally.

Everyone is familiar with ChatGPT from previous years. This type of model is collectively called Large Language Model, or "large model" in Chinese.

To be more precise, it should be a large language model.

Input language, output language, although its output content has subsequently evolved from language text to images, tables and even videos.

The output of these is still data, but it is nothing more than a combination of structured data into unstructured data. The input of the big model is also data, which is the existing data on the Internet.
What we are doing with HBM this time is to input human brain waves and train a large model. It uses the machine body as a carrier to ultimately achieve an impact on the real world.

Taking Tiedan as an example, we collected more than 10 hours of brain waves from online ride-hailing drivers, and then fed this data to HBM. After self-training, it removed impurities from the data and began to output through the machine body.

First run on the test field, then on the test route, then in the restricted area, and finally run without restrictions.

The data used to train HBM is also data, but it is not text data, but the human brain.

You can see that data is abstracted from the adult brain and eventually input into the HBM.

Therefore, driverless cars are just one of its applications, just like large models outputting text is only the initial application, and soon they will start outputting images, videos, and tables.

Similarly, there will be many application scenarios for HBM in the future. Using it simply for autonomous driving is not cost-effective from a cost perspective.

Autonomous driving is just like ChatGPT’s earliest appearance, it’s just a prototype.”

Everyone present looked at each other in bewilderment. Even the regulatory authorities related to HBM had little understanding of the underlying technology and everyone thought it was just driverless technology.

Now it seems that this technology is much more valuable than the driverless technology they thought.

If it is just about autonomous driving, each company is not far behind L4.

It is entirely possible to compensate for this through hardware.

Optical satellite networking, which is being used by many new energy vehicle companies, makes up for the deficiencies in the algorithm at the hardware level, and the gap to L4 is already very small.

Therefore, open-sourcing this technology will not be a big loss for the Chinese, and it can also force America, Europe and Japan, which were not so keen on new energy, to join the new energy vehicle track.

Gasoline vehicles have natural defects in unmanned driving.

But if it is as Chen Yuanguang said, then everyone's views will have to be fundamentally reshaped.

“Yuan Guang, I am a complete layman in artificial intelligence technology. Although I have listened to many lectures given by experts like you, my understanding of the profession is still far behind yours.

I would like to ask, what are its application scenarios? Can you briefly explain it?

Another thing is what it can do in the military field.”

Chen Yuanguang said: "HBM can evolve.

Different types of work have different technical difficulties. For example, construction workers and textile workers are fully capable of doing simple mechanical repetitive work.

The driver's performance is slightly better, and it seems to be doing a good job now.

For jobs that require higher precision, such as electricians and fitters, it requires not only the evolution of algorithms but also the evolution of hardware, such as improving the accuracy of cameras responsible for vision and the accuracy of force sensors on fingers, etc.

Including the improvement of its brain computing power.

This will be an overall improvement.

For me, its biggest short-term use will be as construction workers on the moon, responsible for building a lunar base.

In the medium term, our space station will be dominated by robots, responsible for the repair of space mining equipment and the maintenance of the space station.

From a military perspective, I don't think it's cost-effective. Robots are expensive, inefficient in performing tasks, and lack robustness. In my opinion, they are far inferior to mechanical dogs and drones.

A few can be used for emergency rescue work, but the essence is still a cost issue. From a cost perspective, it is too expensive to replace drivers. " "What does robustness mean?"

“Sorry, this refers to the ability of the system to survive abnormal situations, which can be understood as stability.

In short, these precision instruments have poor stability on the battlefield," said Chen Yuanguang.

“Yuan Guang, I originally supported open source technology, but after listening to it, I feel that this technology has great potential.

My thoughts have been seriously shaken, and I think many of my colleagues here have similar thoughts.

I hope you can help us all clarify the benefits of open source technology. "

Chen Yuanguang smiled and said, "This is also the biggest purpose of coming here this time.

Many things cannot be explained clearly through video, face-to-face interview is the best way.

I would like to first talk about the fact that past research on artificial intelligence has been about making machines simulate humans as much as possible.

Computers have advantages in many aspects, the most typical of which should be signal transmission speed.

The signal transmission of human neurons is an electrochemical process with a speed of 100m/s, while the speed of electrical signal transmission in silicon-based chips is close to 70% of the speed of light, or 90 million meters per second. The speed of electrical signal transmission in topological semimetals is even more amazing, reaching close to % of the speed of light.

The error probability of human neurons in the process of signal transmission is one percent, the error of silicon-based chips is one in four billion, and the error of topological semimetal chips is even smaller.

Chips have very obvious advantages in information processing speed and accuracy.

Two years ago, Intel had a neuromorphic project called Hala point, which used 11.5 billion digital neurons to simulate the human brain.

Even with so many digital neurons, and even though silicon-based chips have natural advantages over human neurons, Intel's Hala point can still only handle computing problems and does not perform well in neuromorphic computing.

On the contrary, a project called Brainoware, which was carried out by Harvard University at about the same time, performed better in simulating the human brain.

The Harvard project combines human brain cells with silicon-based chips to build a new hardware they named Brainoware.

They first used human pluripotent stem cells to cultivate brain organoids, and then used part of the entire brainnoware with traditional computer hardware and part of it with the brain organoids.

They built a three-layer computing framework, divided into input layer, reservoir layer and output layer, in which brain organoids were used in the reservoir layer.

The organoid receives signals through an input layer, which converts them into electrical stimulation signals. The brain organoid acts as an adaptive database, mapping these signals to the output layer. In the output layer, the neural activity representing the reservoir state is recorded and decoded to provide readouts for applications such as classification, recognition, and prediction.

By evaluating the response to stimulation with varying pulse duration and voltage, the physical reservoir properties of Brainoware, including nonlinear dynamics, spatial information processing, and fading memory, were tested. The system was then applied to real-world tasks such as speech recognition and nonlinear chaotic equation prediction.

In the speech recognition task, Brainoware needs to identify the voice of a speaker in a pool of speakers. A total of 240 isolated Japanese vowel audio clips pronounced by eight different male speakers were used to train the system.

Ultimately they achieved the same results in less than 10% of the training time of traditional hardware.

Okay, that’s it for the two examples.

These two examples illustrate that foreign hardware currently has inherent drawbacks, and the HBM model is very poorly adapted to traditional silicon-based chips.

Of course, I have not yet developed brain organs that can be used commercially on a large scale to replace silicon-based chips.

But what I can tell you is that if foreign countries want to use the HBM model, they must buy the topological semi-metal chips produced by Dongda, which is equivalent to us blocking the hardware end.

Having said that, even if the technology is not open source, considering that the HBM model is to make progress, we need to cooperate with leading domestic technology companies. The more people involved, the greater the risk of technology leakage.

We might as well open source it directly and block the supply from the upstream hardware end.

Sign a technology open source agreement with all participating countries and organizations, and all technological advances around the HBM model need to be open source and cannot be used in the military field.

If you don't comply, then we have an agreement to rely on and can openly refuse to supply topological semimetal chips.

Simply put, technology will be leaked sooner or later, and we have absolute control over the hardware. By opening it to other countries, we can jointly promote the development of HBM technology around the world and help Dongda chip companies open the door to the global market.

Of course, I think there are still many things that can be traded, and we can talk about them slowly.”

The Brainoware mentioned above is an article published in the electronic journal of Nature in December 23.

The composite machine, which combines biological and mechanical elements, has extraordinary potential in the calculation of nonlinear equations and speech recognition.

Maybe in the future mechanical ascension will be a privilege for a minority group.

After Chen Yuanguang's detailed explanation, the voices supporting open source prevailed.

"I think Yuan Guang said it very well. This is a very valuable bargaining chip for us. We can give it away, but we have to get something valuable in return."

"I also support what Yuanguang just mentioned about our chip industry entering overseas markets. I think this is a strategy they have to adopt if they want to use HBM technology.

Just like we had to buy Nvidia graphics cards in the past because there was a lack of alternatives in the market.

We have to come back with some real benefits.”

"I agree with everyone's views just now. I have a question for you. You just mentioned the Harvard research on the device that combines human brain tissue and computer hardware. You mentioned that it has not been developed yet.

I would like to ask, should we take this technical route in the future?
Does this technical route have potential? "

Not only was he curious about this question, but many people present were curious about it.

The combination of machinery and flesh may have serious ethical impacts.

“This technology route is very promising, but the cost is too high. Whether it is the generation and maintenance of organoids, the power consumption of the entire device, or the efficiency management at the data level, there are big problems.

We can sponsor some pre-research projects through the National Natural Science Foundation, but there is no need to be a pioneer in this technology route." Chen Yuanguang said.

By the end of the meeting, everyone's opinions gradually converged.

“Yuanguang, if the HBM model can only use topological semi-metal chips, then I think it is entirely possible to discuss open source technology.

Next, we have arranged for a specific company to find experts to verify the authenticity and submit a detailed report.

This is not because I don’t trust you, but this matter is related to a very important direction of our work, and we must do things rigorously from a work perspective.”

(End of this chapter)