In fact, just like unboxing, Zhou Yan couldn't wait to take it apart and take a look.

However, the operation of this kind of program has very high requirements for the configuration of the computer, so Zhou Yan simply went back to the hotel room first and got back the computer he had configured before.

It's just that as soon as he returned to the room, Zhou Yan was not in a hurry to come out, and operated in the room.

Because the program was too big, Zhou Yan took out a mobile hard drive and copied it as a whole, and after some tinkering, he also opened the package sent by the female hacker.

Soon, the program was running.

It's still black, and nothing has changed in the dialog.

Zhou Yan first entered a basic algorithm requirement, requiring 01 to write the code of the monkey choosing the king, which is an algorithm routine used to verify whether beginners are getting started, and it is also a common test topic in schools.

And 01 lived up to his promise, and after thinking for a few seconds, he quickly gave the code.

The code sent by the female hacker itself is in the python running window, Zhou Yan copied and pasted the code, the window appeared, Zhou Yan entered several sets of values, and the result was that it worked perfectly.

Zhou Yan modified several of his requirements, such as changing the monkey to be connected from end to end, and then went to verify the value, and found that it was still correct.

There seems to be nothing wrong with this program.

Zhou Yan found more than a dozen interview questions from large factories with different difficulties, and he planned to test them one by one.

Starting with a simple one, Zhou Yan put Sandao in one breath and let 01 answer.

As expected, 01 gave the right answer.

Zhou Yan felt that it was the result of the female hacker adding the algorithm library, and this kind of verification could not test how powerful 01 was.

So Zhou Yan increased the difficulty and pasted two slightly difficult algorithm questions into it.

This time, 01 'thinking" took a little longer, but soon, the answer was still given.

Zhou Yan also verified the answer, which was correct.

So Zhou Yan was impatient to put it one by one, and he directly put the most difficult algorithm problem into it.

There are no readily available answers to this kind of question on the Internet, and he is also looking forward to a reply from 01.

The answer to the 01 response in the next second is still complete, but there is a problem and an error in the verification process.

Zhou Yan re-pasted 01's answer into the dialog box, and told 01 the error, asking him to debug according to the code he had sent in the past.

Three seconds later, 01 gave the answer.

However, the error is still reported during verification.

Zhou Yan became extremely patient at this moment, and pasted the error code generated by 01 himself into the question and answer box over and over again, asking him to correct his mistakes by himself.

After correcting it almost five times, 01 finally came up with the correct answer.

Zhou Yan knows that this is the process of the intelligent dialogue model to improve the answer, and it is also the correct way to play the intelligent model in the future.

Essentially, intelligent models function to meet human needs.

At first, you may have a vague idea of what your needs are, but as the conversation progresses, the intelligent model understands more and more what your needs are.

A large requirement will contain more than a dozen small requirements, which the questioner may not know at first, but with the deepening of communication with 01, both the questioner and 01 can understand the requirements more deeply.

This is actually a process of simulating human communication, when you ask a question, it is impossible to expect others to answer immediately, it is to clarify the needs of the other party, and then give further feedback, the more feedback the more correct, because the more difficult the request you make, the more difficult the other party needs to look for, the scope of the other party's search is getting smaller and smaller, more and more accurate.

After only communicating five times to answer this question, Zhou Yan was quite surprised.

So he changed his mind and chose to propose a big project, and let 01 first provide the steps needed to complete the big project.

And soon, 01 gave the steps, a total of five points.

Zhou Yan's request is: to conduct text statistical analysis of online texts.

01's reply is: 1. Import text. 2. Text preprocessing. 3. Text statistics. 4. Draw conclusions.

It's relatively simple, and Zhou Yan gave specific requirements.

So 01's reply became: 1. Import text. 2. Preprocess the text, import the stop word list, and use the word segmentation tool to preprocess the text. 3. Conduct word frequency statistics on the processed word segmentation text. 4. Conduct a statistical analysis of the word frequency after the word frequency statistics and draw conclusions.

It's a little more specific, but it's still a bit broad.

But Zhou Yan can't wait,

He directly offered to let 01 generate the relevant code according to his own steps.

Soon, 01 generated a series of codes.

Zhou Yan downloaded the text of "Breaking the Sky", verified this code, and soon got good data, although it is not comprehensive, but it is obvious that this project can be carried out.

Zhou Yan also put forward requirements for the improvement of this code, including the visualization of word frequency, which is displayed with a cloud map, including text training on word frequency, so as to find out the relationship between words.

This time, Zhou Yan's request was to deal with two texts, two of which were Jin Yong's "The Legend of the Condor Heroes" and "The Legend of the Condor Heroes".

Soon 01 gave the corresponding code modification, and Zhou Yan copied and pasted it before testing it.

In fact, the demand is also very simple, that is, to find the relationship between the protagonists.

Yang Guo corresponds to Xiaolongnu, so what does Guo Jing correspond to, result: Huang Rong.

Zhou Yan found the corresponding related words of Yang Guo and Guo Jing respectively, and at the same time limited 01 to the characters, and designed a character word judgment code by himself.

This time, 01 finally made a little mistake, such as the character segmentation, which will turn the 4-word character into a 2-word output.

This is the problem in the details, Zhou Yan is not a novice in the algorithm now, he knows how to improve, but Zhou Yan is more looking forward to the results given by 01.

So he pasted this question to 01, and 01 also gave a judgment on the word frequency, if the two split words appear many times at the same time, then determine that the two split words are merged into the same word, and delete the first two split words in the dictionary, and process the data that has been segmented, merge the two split words, and if there is no good word segmentation, the split word will be used as the basis for word segmentation, and the word segmentation will be followed.

After pasting it in, Zhou Yan found that the effect was unexpectedly good.

It's just that there are some problems in the character discrimination after that.,Will add weapons or something.。

Zhou Yan asked 01 again, and 01 answered some code modifications, but after adding it, it was still the same, and it was impossible to do a detailed division.

Zhou Yan asked 01 if there was any good way, and 01 said that before processing, the characters to be dealt with would be preset, so that there would be no mistakes in judgment later.

That is, to solve the problem from the source, which is also a very mechanical way to deal with it.

But sometimes human processing is also very mechanical, and it is also cramming, such as gene sequencing.

So there's no way to ask too much for this, and 01 is actually quite good to do this.

Zhou Yan preset a few more small projects and tested 01.

Finally, Zhou Yan also came to a conclusion.

That is, 01 already has the ability to work more complex code.

The more professional Zhou Yan has limited strength and can't test it, but there is obviously no problem with this 01 to deal with Xiaobai.

After testing the code,

Zhou Yan also tested the completeness of 01's text, such as asking 01 to write a paragraph according to the requirements, asking him to include the characters and events that appeared, and presenting him 01's thinking process.

For the purpose of doing this, Zhou Yan also hopes to find the algorithm logic written by the female hacker.

And because there is no threshold for Zhou Yan, 01 also shows his ideas, including splitting the events, sorting them, and then expanding them.

Thus the so-called is obtained

Sequence of character behaviors.

Here's what it looks like:

What A did, what A did led to what B did.

What B said, A said.

What B did, what B did led to what C did.

What did a say.

What did B do.

What did a say...

That's about it.

These sentences are then combined and further polished through natural language processing, such as object complements, to obtain the final sentence.

Zhou Yan actually understands that the essential reason why the intelligent dialogue model is so popular later is that it does not present corresponding steps to the outside world.

It's equivalent to a black box effect, which gives people a sense of uncertainty.

It's very mysterious and powerful, and it's sought after by many people!

However, there is a lot of room for manipulation (fooling) in this way of processing.

If you are asking about encyclopedic knowledge, people have already prepared a text library, and what you ask and answer is essentially a function of a search engine.

If you are asking about algorithms, and similarly, if you can have a collection of all the algorithms in the world, then the intelligent model can directly give answers to some basic problems without even mobilizing its own computing power.

This is really not fundamentally different from looking up a dictionary.

What Zhou Yan really wants to do is to let the machine understand the needs of human beings, so as to realize the leap from text to pictures, to sound, to video.

Truly realize the linkage between intelligent models and tools, and broaden the application scenarios of intelligent models.

Imagine what instructions are given, and the virtual reality machine will immediately show you the corresponding picture, and even your behavior can be judged, so as to interact with the picture equally, and the effect will be extremely explosive.

For example, astringent movies, real-life interaction, in the previous life, the island country had a business in this regard.

And artificial intelligence can definitely push this business to a new level.

There was no one around, so Zhou Yan asked 01 to generate an astringent description, requiring no less than 300 words.

So 01, after a short period of thought, quickly generated a large paragraph of text.

And the content is a little... Stimulate.

01 doesn't know how to hide or scale, so what is presented is the most primitive description on the Internet, but it is such a naked description and stacked text combination, which has a special style and charm in it.

Even if the processing of words is still a bit mechanical, the overall needs can obviously be satisfied.

Zhou Yan even read this text.

After calming down for a while, Zhou Yan opened 01's manual again and looked at it, which was obviously the modification notes and some additions made by the female hacker to 01, and Zhou Yan also discovered the reason why 01's logic became stronger.

The female hacker added a series of mathematical methods, such as discrete mathematics, recursion, and linear algebra theory, which made the text more logical and less error-prone.

In other words, it is to put the text within the framework of logic.

Speaking of which, the mathematical level of this female hacker is really not general, even if Zhou Yan has systematically studied mathematical theories, but in terms of application, it is really not as good as this female hacker.

Obviously, she is what the former algorithm teacher called "someone else's child", and she can learn and use mathematics and algorithms to live.

Generally, like this kind of talent, the planning institutes of 985 colleges and universities are basically rushing to get them.

Not to mention that when she was attacking the Wave Chase Home app at that time, this female hacker was able to go back and forth with Zhang Long and Gu Yuan, who were middle-aged programmers, and she could advance and retreat freely.

You must know that Gu Yuanzhong's team used to specialize in security firewalls, but when facing a female hacker, they still couldn't catch her tail, which is enough to explain the problem.

Zhou Yan moved the intention of solicitation.

He opened a chat with the hacker

window, sent a message over.

Qin Yunuo is operating a computer in the team office of Jianye University, which belongs to the entrepreneurial base of Zhejiang University at this time.

She likes to be quiet, which is why the team assigned her a separate office.

Mr. Zhou: [01 I read it, and the change is very good. 】

Qin Yunuo saw Zhou Yan's news and replied: [There are still some imperfections, but I also said before, this is not a one-person project, if you want to be more perfect, you need to form a team to train the intelligent model for a long time. 】

Mr. Zhou: [There is a 4G pilot base in Hangzhou, have you heard of it? 】

Qin Yunuo pursed her lips slightly, she had paid attention to the news of 4G, and she also knew that a national pilot was being carried out, but most of them were carried out in coastal cities, just like there was one in Hangzhou.

She replied: "I've heard of it, but I haven't understood it. 】

Mr. Zhou: [Where do you think the best application scenario for this intelligent model?] 】

It's up to her.

Qin Yunuo: [Mobile phone. 】

That's the right answer, and the portability of mobile phones can make the most of this question-and-answer model, and even if the database absorbs enough, it can provide a guide for any category.

If it can be refined, travel, education, etc. can be further diverged.

Mr. Zhou: [Shouldn't the application scenarios of the code be suitable for computers?] 】

Qin Yunuo replied directly: [It's more chicken, at present, it can't do the self-operation of large-scale algorithms at all, and can only complete or answer some small questions, for professional people, it is at most a debug and annotation tool. 】

Mr. Zhou: [If it is a mobile phone, what type of application do you think is suitable for carrying this product?] 】

Qin Yunuo furrowed her good-looking eyebrows slightly, she tilted her head, and did not answer Zhou Yan immediately.

So, he won't use her as a question-and-answer model.

(After reading it, remember to bookmark it for next time!)