"No, it's not that simple", Wang Yinan didn't believe it, the old monk waited here for a long time with a kind purpose, just to tell himself an old-fashioned story.

However, Wang Yinan's visit to Purple Mountain today is all temporary, let alone coming to Linggu Temple, if someone arranges it, this person will inevitably be too powerful, or arrange such an old monk in every temple in Jinling?

Wang Yinan shook his head and drove away this incredible thought in his mind, just kidding, only a madman would do this kind of thing.

After telling Wang Yinan the story of buying and returning the pearls, the old monk sat cross-legged on the futon again, without saying a word, looking like a monk, Wang Yinan asked him anything again, but he didn't say anything.

Okay, Wang Yinan bowed respectfully to the old monk, turned around and walked out, no matter what, this guy reminded himself that he has been a little self-inflated recently, and it is still quite helpful, he should keep a low profile at all times, and making a fortune in a muffled voice is Wang Yinan's consistent style of doing things.

When I walked out of the mountain gate of Linggu Temple, the twilight was deep, and the wind blew through the Xixi Mountain, which seemed to remind Wang Yinan that it was already late autumn and the harvest season was coming.

"Pearl", "box", Wang Yinan walked out, and there were words in his mouth.

At the same time, at the Europa Nuclear Institute CERN, after pondering for a long time, Witten wrote the names of Wang Yinan and Li Wenjing in a letter sent to Sweden.

Like him, there are several physicists and chemists in Britain, rice and the Sun country, and it turns out that it is already the Nobel Prize nomination season before they know it.

Of course, Wang Yinan will not know everything that happened on the Blue Star at this moment, on the streets of Jinling, he kept repeating like he had lost his mind, "What is a pearl, what is a box".

In the eyes of the vast majority of people, especially the vast majority of mathematicians, the "G indescribable theorem" is a major discovery in the field of axiomatic logic after Gödel's incomplete theorem.

Many of the methods used in the process of proving theorems are only used for calculus and establishing theorems, and do not appear in the final paper.

However, Wang Yinan does not really want to prove anything indescribable, for him, whether the theory is perfect or not is optional, not necessary, like Grothendieck.

"What is a pearl to me?" , Wang Yinan asked himself, "That damn theorem is obviously not, because for me, except for a false name that does not belong to me, it has no value at all."

"So, are these formalized methods, and mathematical tools, pearls?" ，

"No, no, I'm dizzy by the old monk", Wang Yinan walked on the secluded tree-lined path and gave himself a crisp slap casually.

A mother with a child on the side hurriedly pulled the child into her arms, "Uncle should have something wrong with his brain",

"Of course you can't learn, it's a bad behavior to beat yourself",

"Don't ask him, it must hurt."

Wang Yinan didn't care to clarify, "I don't have anything wrong with me at all!" ”

He was thinking, "For me, the question now is not which one is the pearl, but all the things that have been learned, what is valuable to me?" ”

"The measure of whether it is valuable is very simple, whether it is helpful for what I want to do."

The more Wang Yinan thought about it, the clearer he felt, as if he had grasped something amazing.

"The indescribable theorem is definitely not for me, so these mathematical tools, formal methods, and so on are left," Wang Yinan thought.

"What use do these things do to me?"

"It is certainly useless for existing superconductivity theories, but in the future, if the Gödel system or the next generation of Gödel discovers new superconducting materials, there are these formal methods and mathematical tools."

"It will be possible to come up with new theories to explain the results of neural networks more easily, instead of being blinded as before, and you need to consult multiple masters and rely on the wisdom of others to find a theory that can be understood by humans."

"Although the complexity is to a certain extent, it will certainly be indescribable, but with the complexity of existing neural networks, there is still a good chance of finding a theoretical explanation that can be understood by humans."

"It's not this, it's not this", although it found an ideal use, the frustration that shrouded Wang Yinan showed no signs of abating.

"No, no, I must have overlooked something", in addition to explaining the new results in the future, there must be some other uses, Wang Yinan thought bitterly.

Unconsciously, he walked from Linggu Temple to the gate of Sun Yat-sen Mausoleum, and unconsciously, walked to the most prosperous downtown of Jinling, Xinjiekou.

Jinling as a commercial city, and Xinjiekou is the commercial center of the city, even if it is not a weekend, when the lights are on, it is still full of people, walking in the crowd, no need to wear a big hat, sunglasses and scarves, and no one knows, no one cares if you are an ordinary person, or a big star.

Following the flow of people, Wang Yinan walked between several shopping malls, he didn't have a clear destination, muttering mathematical terms that no one could understand, and from time to time he looked at the trendy clothes and leather bags in the window.

Wang Yinan felt that there were some amazing things, just within reach, but he couldn't grasp them, and this feeling almost didn't make him crazy.

At this time, Wang Yinan's phone rang, and it was Chen Zihao calling, "Boss, when are you coming to inspect?"

"Our prototype system is almost developed",

"But killer applications are still hard to find, artificial intelligence is a good direction, face recognition, image recognition, etc., but the restrictions on hardware are still quite large."

"Yes, we're trying some supposedly low-cost neural network chips, but the results are not very reliable."

Wang Yinan promised Chen Zihao, "I'll come over as soon as I'm done with things here, and I'll definitely touch everyone this week."

Chen Zihao, who got the letter, hung up the phone with satisfaction, and Wang Yinan went to take the phone and stood in the crowd stupidly, motionless for a long time.

Unfortunately, it was the child and her mother on the side, "Mommy, look, my uncle who beat me just now was motionless, was someone using the immobilization method?"

"Don't talk nonsense, only in mythology there is a method of fixation",

"Oops, it's really wrong, it's been a minute, and I can't move with the phone", the mother also panicked and was about to call.

Wang Yinan finally came back to life, and with a "snap", he gave himself a mouth again, "Mommy, my uncle beat himself again, and he looks very happy, I want to try it too."

"I'm so stupid, I just know that there are no wolves on the mountain", Wang Yinan rushed out of the crowd and rushed to the hotel with words in his mouth.

This time, Wang Yinan really understood what kind of gold mine he had overlooked.

In ancient times, there was Wang Yangming cave enlightenment, after going out of the customs, it was all over the world, invincible, now there is Wang Yinan, just surnamed Wang, Xinjiekou enlightenment, after that, Jinling has a legend of Xinjiekou slapping.

What did Wang Yinan think of that made him so excited?

As you know, the practical artificial intelligence applications are basically based on neural networks, and neural network applications are basically built on dedicated hardware.

The application process is like this, firstly, the use of large-scale samples, the neural network is trained, of course, there can be many kinds of training, classification, recognition is the most common, but also the most mature.

After the training is completed, that is, after the neural network is basically stable, the trained neural network is used to complete the actual task, such as recognizing a specific image, number or speech text, etc.

In other words, both training and application need to be built on dedicated hardware, mainly GPUs, ASICs, or the latest AI chips.

The same is true of Migo's Go dog, the neural network generated by training is incomprehensible to humans, and the theorem discovered by Grothendieck further proves that this kind of neural network, after reaching a certain level of complexity, cannot be described by limited axiomatic logic and formal theory.

However, note that the theoretical proof of the description of finite steps is impossible does not mean that the formalization of neural networks is worthless!

Theoretically complete understanding and description is impossible, but does not mean that a simplified approximate description is impossible!

In other words, using the mathematical tools and formal methods learned from Grothendieck, plus Wang Yinan's own research and understanding, as long as there are basic principles about the composition of neural networks, such as BP type or Hopfield type, whether it is based on annealing or deep learning.

Wang Yinan can find some general methods and techniques to map the neural network block to a finite state automaton, so that the entire finite state automaton and a certain block of the neural network are equivalent in a certain accuracy range.

That is, for the same input, a neural network and an equivalent finite-state automaton will get essentially the same output.

As a friendly reminder, any finite-state automaton can be equivalent to a piece of program code.

To be clear, Wang Yinan can now find a series of universally applicable mathematical tools and methods to map any type of trained neural network to a basically equivalent finite-state automaton, or a piece of program code for a specific CPU system.

In other words, as long as he can get the detailed structure of the U.S. Go dog neural network, and the final state of killing all sides after the training is completed, and then use this series of mathematical tools, formal methods, with manual, or some automated programs, he can get a specific system, such as a series of code snippets of X86,

The code may be chaotic and disordered, but the original trained neural network will be equivalent to the basic functionality!

As long as the neural network of the entire Migo Go dog is comprehensively mapped, and all the X86 code finally obtained is integrated, you can make Ke Jie two sons on an ordinary desktop or even a notebook!

"Damn, this time it's sent", Wang Yinan was on the hotel elevator, trembling excitedly.