Spark

In the next few days, Chen Nian's life basically returned to calm.

In addition to providing technical questions and answers about the new generation of small nuclear reactors to Lu Qingzhou of the Nuclear Power Research Institute from time to time, it is to organize technical discussions on the hull design of the 096 strategic nuclear submarine project.

In addition, the JL-3 project is also proceeding in an orderly manner under the auspices of CALT.

To say that the slowest progress is of course the Nantianmen project.

Of course, this project itself is different from other projects, and no one will limit it with too harsh speed requirements.

On the whole, several large-scale military projects in Huaxia are considered to be in the normal process, which once again triggered Chen Nian's insufficient pressure conditioned reflex point, making him involuntarily start to consider adding new civilian projects to the plan. Tech goals.

However, Chen Nian did not have a clear idea of ​​what to choose.

The suggestion from Xinghuo is to solve the problem of food and improved seeds first, but at this point in time, Mr. Yuan can be said to be in his heyday, and he and his team have almost occupied the resources in China's agricultural scientific research field. It is not cost-effective to forcibly change the route without new technology.

Therefore, this direction is temporarily shelved.

Chen Nian felt that it would be better to make efforts in the medical field and give some buffs to these older researchers, maybe they could achieve unexpected results.

However, this is not something that can be accomplished overnight.

After entering the field of near-future technology, the pace of research and development of each technology has been forced to slow down.

In contrast, the emergence of every technology is an earth-shattering and epoch-making product.

This puts higher demands on the determination and patience of the entire team. After all, everyone is a pioneer now.

There is no road ahead, and the footprints of the past are the roads of future generations.

Of course, compared to other ordinary scientific researchers, Chen Nian, a traveler who has seen the future, is much calmer in mentality. Although everyone has the belief that he must win, only Chen Nian knows that victory in the field of science and technology, In fact, it is not an expectation, but an inevitable result.

Looking back from the perspective of the future, everything becomes orderly.

It's a weird feeling, but it makes sense

A few days passed quickly, and the day Chen Nian and Wang Jiancheng agreed to meet soon came. Early on Wednesday morning, Chen Nian checked Wang Jiancheng's air ticket, asked Chen Guo to arrange the vehicle and itinerary, and waited at the airport early. Before 11 o'clock, I received Wang Jiancheng who flew over by an early flight.

Qiao Mai also came here together——Chen Nian wasn't too surprised by this. Although Wang Jiancheng didn't tell him, the logistics team had already notified Chen Nian.

As soon as they met, Wang Jiancheng rushed forward enthusiastically and grabbed Chen Nian's hand.

Senior, long time no see!

Li Xiang on the side looked at this scene dumbfounded, and joked:

Don't make it look like an interview with the leader, okay? Even if you come up for a hug, I don't think it's so inconsistent. How about you two resting on the plane?

Very good, very good! Hey, business class, very comfortable!

Wang Jiancheng rushed to answer.

Yo huh? You haven't spent that little money yet? Still have money to fly business class?

It can't be spent, it can't be spent at all. Now there are funds for any project, and I don't know where the money is spent.

Wang Jiancheng had a proud look on his face, and the three of them smiled tacitly without making a fuss.

In fact, no matter how high Chen Nian had on Qiao Mai and Wang Jiancheng, in a worldly sense, they were already unique successes.

While their peers are still struggling with postgraduate entrance examinations and work, they already have their own projects and even cooperate with those unattainable units.

This point alone can only be achieved by the top group of people.

Under such circumstances, spending money to fly business class and pursue some material enjoyment is really not a big deal.

——

In fact, Chen Nian would feel that something was wrong if they acted in a bitter and hard way.

After exchanging a few pleasantries, Li Xiang took a few people into the car that had been prepared earlier, and after bypassing several sensitive places that had been marked earlier, they finally settled down at the residence arranged by Chen Nian.

Let's make do with living here today—how many days do you plan to stay in Chengdu?

Chen Nian asked.

I guess it will take a week, but I have nothing to do, so I came here specially to see you. Sister Qiao Mai has something to do, right?

Yes, I'm not sure about the time. I plan to talk to Teacher Zhang Wen from the Mechanical Design Institute in a few days. I have already made an appointment with her, which is related to my project.

If it's inconvenient later, I'll go out and stay in a hotel by myself.

No inconvenience.

Chen Nian waved his hand and replied.

Zhang Wen from the Mechanical Design Institute has really come into contact with it. When she first started the lift fan project, Zhang Wen participated in the two-stage counter-rotating fan project. At that time, she was already a well-known expert in the field of motors.

After several years of precipitation, her level has naturally improved steadily - Qiao Mai's research direction is electromechanical-hydraulic integration. If you want to get in touch with high-level theories, you must not avoid Zhang Wen.

However, with Qiao Mai's current grades and level, can he really make an appointment with Zhang Wen?

Chen Nian was somewhat skeptical.

What did you tell Teacher Zhang? I've dealt with her before, should I mention it for you?

No, no, it's a good appointment!

Qiao Mai quickly explained:

She is quite interested in the agile pressure-sensitive hydraulic transmission technology used in the mechanical exoskeleton we made before. I have chatted with her a few times, and the relationship is not bad, so I won't miss it.

That's all right.

Hearing this, Chen Nian felt relieved.

It seems that although Qiao Mai's research direction can't be said to be very groundbreaking, it has been recognized by industry experts - this point, I have underestimated her.

After talking about the approximate schedule, Li Xiang made plans for Qiao Mai and Wang Jiancheng to play.

Of course, in fact, there is nothing to plan too much. All the people who come to Chengdu, basically no matter where you come from, play the same things:

See pandas, go to Mount Qingcheng, visit Dujiangyan.

Others such as Du Fu's Thatched Cottage, Xiling Snow Mountain, and Daci Temple are all subsidiary projects, and whether to go or not depends purely on personal interest.

After all, Qiao Mai is a girl, and she is more interested in traveling. On the other hand, looking at Wang Jiancheng, the only thing he is interested in is the panda.

He finally listened to Li Xiang's introduction patiently, and when Chen Nian was free, he couldn't wait to take out his computer from his backpack, and dragged Chen Nian to the side.

Senior, your surprise!

Wang Jiancheng said mysteriously.

Computer? Program? What? AI? Question-and-answer type?

Chen Nian's mind turned several corners in an instant, and when he said the word question-and-answer type, he basically already had the answer in his heart.

And Wang Jiancheng's expression changed from showing off at the beginning, to surprise, and then to faint disappointment.

——

He really wanted to scare Chen Nian, but it was obvious that Chen Nian had predicted his prediction.

.It's a question-and-answer AI, but it's different from the current text-based question-and-answer AI. My Bai Ze doesn't rely on a preset database to answer questions. It can.

Retrieve Internet information, perform machine learning steps, filter answers based on real people's habits, and at the same time keep moving closer to natural language?

Chen Nian interrupted Wang Jiancheng again, and the latter was stunned for more than ten seconds before finally answering:

.Yes. Senior, how do you know??

Good guy.

A helpless smile appeared on Chen Nian's face.

how do i know

Isn't this thing the chatgpt of the childhood version?

Chen Nian did not answer immediately, but said to Wang Jiancheng:

How far can it be done? Let's try it out?

Wang Jiancheng finally regained some enthusiasm. He turned on the computer, connected to the Internet, and clicked to start a small icon on the desktop.

White Pond.

As the fan rotates, the software's simple and crude main interface also pops up.

An input box below, a display box above, that's all.

Senior, it's alright, just enter the question in the input box, and Bai Ze will give you the answer.

Is it okay to ask anything?

Anything is fine!

Chen Nian thought for a moment, then swiftly typed a line in the input box:

The water can be drunk when it is opened, but why can't it be drunk when the door is opened?

As soon as this question was typed out, Qiao Mai and Li Xiang, who came to watch curiously, were all dumbfounded.

...What kind of mentally handicapped problem is this!

Li Xiang said bluntly, without giving Chen Nian any face.

And Wang Jiancheng, the author of Bai Ze, was even more stunned. He watched the program he wrote by himself running for several minutes, and finally jumped out an article that was obviously crawled directly from a search engine. Why can't you drink it directly is the content of the health-preserving article.

No, senior, are you too bullying? This kind of question, let alone an AI, may not even be able to answer a real person?

Then what can't be answered, isn't this just a question of polysemy?

Chen Nian said in a relaxed tone.

In fact, he really didn't deliberately use this cerebral thrombosis problem to make things difficult for Wang Jiancheng.

He just wanted to know whether the so-called artificial intelligence created by Wang Jiancheng had a natural language module, and if so, to what extent.

Obviously, the current results are not very satisfactory.

It is estimated that there is still a long way to go to reach the level of chatgpt in the previous life.

After all, the astonishing AI back then, let alone this kind of question, even for the mentally retarded, it can answer more than 99% of the questions in a serious manner, and it can answer well.

——

Of course, the success of chatgpt was finally achieved with the support of super computing power and after a lot of training and feedback.

Using its standards to demand Wang Jiancheng's Bai Ze is also a bit difficult for others.

Thinking of this, Chen Nian asked:

So, to what extent can you do it now? What is the core principle?

Some simple questions can be answered in the same way as real people.

With that said, Wang Jiancheng took the keyboard and typed a very basic question in Chen Nian's opinion:

How's the weather today?

Soon, Bai Ze gave its answer.

The temperature in Chengdu today is pleasant, with the highest temperature reaching 26 degrees, but the temperature at night is as low as 11 degrees, so we need to keep warm.

As soon as this answer came out, both Qiao Mai and Li Xiang couldn't help being amazed.

In their view, a machine that can do this is quite good.

Wang Jiancheng smiled shyly, and then continued to type:

Chengdu, recommended attractions.

After a while, a barely fluent answer jumped out.

After this step, Chen Nian finally figured out the limit of this Bai Ze.

To be precise, it is not a castrated version of chatgpt, but an enhanced version of siri after using more advanced machine learning algorithms.

——

This is not bad, after all, there are still more than two years before the real siri is born.

So, in fact, this is a question-answering machine based mainly on search engines? You probably haven't set up a natural language processing algorithm at all, right? These seemingly 'natural' answers are all excerpted directly from the Internet?

There is an algorithm, but the function is very weak at present.

Wang Jiancheng replied without hesitation.

Its natural language algorithm logic is based on a set of 'repeatability' scores designed by myself, that is to say, it will give priority to selecting answers that already exist on the Internet, and perform certain permutations and combinations.

The calibration process is to continuously compare the database to screen out answers with higher scores—the scoring standard is the degree of repetition between answers and Internet information.

Understood, introduce limited supervised training of the reward model.

Chen Nian understood Wang Jiancheng's logic in an instant.

It has to be said that this is a very clever solution in the case of limited computing power. With this solution, an artificial intelligence can be extremely smart when dealing with simple problems.

Because in fact, it is like a remote control chat device. What is hidden behind it is not a code of 0 and 1, but a huge database of real people.

Chen Nian still remembers that when the concept of artificial intelligence was just emerging, there was a so-called AI with such an idea in China, but he always sneered at this kind of AI because he went in the wrong direction.

The meaning of this kind of thing is purely to cheat funds and subsidies.

Thinking of this, Chen Nian said:

You have a good idea for this thing, but the route is completely wrong.

I can't support you continuing to go in this direction, because the further you go, the more damage you will do to the development of real AI technology.

Why?

Wang Jiancheng asked puzzledly.

Because it's not really AI, it's just a smart search engine.

Smart search engines get it.

Wang Jiancheng is not stupid. After hearing Chen Nian's explanation, he immediately understood the crux of the problem, and also realized the meaning of the so-called the farther you go, the greater the harm.

So, he hurriedly continued to ask:

What if I change direction? Does the whole logic have to be redesigned?

Of course it needs to be redesigned. Moreover, the most useful part of this Bai Ze is probably the part that you put the least effort into before, that is, its natural language module, especially its reward model.

If you can't solve the problem of the reward model, the evolution of AI will always stay within the threshold of human supervision, and what you create can only be fake AI.

Understood.

Wang Jiancheng let out a long sigh of relief. Judging from his expression, it seems that he has made up his mind to start from scratch.

Chen Nian suddenly felt that there was always some indescribable joy in what this kid did.

He has been thinking about building aerial platforms and space carriers, but it seems that what he is best at is artificial intelligence.

I don't know, in more than ten years, will he become the father of Nantianmen, or the father of artificial intelligence?

Maybe, both?

Thinking of this, Chen Nian couldn't help but look forward to it.