At present, the biggest problem facing artificial intelligence is that it cannot accurately understand human language.

In other words, the current artificial intelligence can only realize the collection and collation of data, and cannot conduct in-depth analysis of data.

In fact, the analysis and description of the content is the ultimate form of artificial intelligence, and the artificial intelligence robot on the market only gives artificial intelligence a specific knowledge base, which has a certain retrieval ability, but has no comprehension ability, and is only a collation of information and words, he can only recognize the words, but cannot understand the meaning of the words.

Humans have been researching artificial intelligence for decades, but there is still no real intelligence.

Artificial intelligence is recognized as a major trend in the future, and in fact, Dianjin Group has been investing in research in this area.

It's just that no great effort has been made, mainly because there are still very few places where artificial intelligence can really be used.

For example, the speech recognition system of mobile phones can only recognize specific languages, and a slightly more complex one cannot be recognized, let alone semantic analysis.

In fact, the so-called research on artificial intelligence is still in a conceptual state most of the time.

There are many theories about artificial intelligence, and there is no one correct theory that really works.

Because practice is the only test of truth, but most theories are too difficult to put into practice.

Of course, Ye Qiu did not have an exact concept of whether some theories of artificial intelligence could be realized, so he specially consulted artificial intelligence engineers.

In the end, it was concluded that in terms of imitation ability, in fact, many laboratories that study artificial intelligence can already do it, but this imitation ability, in fact, so far, still refers to the most basic call through the database, not that artificial intelligence can really think and imitate.

There are neural networks that simulate the human brain as the core program structure of artificial intelligence robots, and the effect is not bad, but the current progress is still slow.

Of course, Ye Qiu also knows that nothing can be achieved overnight, and the explosion of the concept of artificial intelligence, more and more technology companies and Internet companies are joining it, which is actually driven by interests.

Just after Ye Qiu disappeared for a long time, Dianjin Group suddenly announced that it would spend tens of billions of dollars to set up an artificial intelligence laboratory, recruit artificial intelligence engineers from all over the world, and also reached a cooperation with the artificial intelligence laboratory of the Chinese Academy of Science.

The main reason why many Internet companies will give priority to the implementation of artificial intelligence projects is that Internet companies have the advantage of big data.

For example, search engine companies, such as social platforms, such as technology companies, etc., have unique advantages in this regard.

The most basic requirement for artificial intelligence is to have big data as a backing.

In fact, like the current intelligent input method, it can also be regarded as a kind of artificial intelligence, vocabulary association, etc., in fact, it is through the retrieval of big data, through the memory of people's word habits and then recommended to users.

The first generation of artificial intelligence robots launched by the artificial intelligence laboratory of Dianjin Group is an interactive artificial intelligence robot.

To put it simply, the first phase of the artificial intelligence robot in the artificial intelligence laboratory of Dianjin Group is to be able to imitate people's chat habits and interact with humans in questions and answers.

For example, the first problem that AI needs to solve is how to pick out the answer from the questions and answers of countless people.

Because of a question, there will be different answers in different people's answers, and these answers are known to the artificial intelligence robot, but how to pick the appropriate answer from these answers, this mechanism is the key.

At present, all the so-called robot customer service cannot be called artificial intelligence at all.

Because those robot agents can only answer some pre-set questions, and the answers to the questions are also pre-set, which is not intelligent at all.

But there is also a question involved here, or a return to the original question, where AI can pick out and answer by searching for other people's answers.

But because the text itself has no emotion, if the AI can't understand the meaning of the text, then if you scold him, he may refute it, but he doesn't know that you are scolding him, because the answer he gives is given by someone else.

So the first thing to do is still semantic parsing.

As a top Internet giant in China and even in the world, Dianjin Group has established an artificial intelligence laboratory and invested heavily in it, which will naturally attract a lot of attention.

The first step of the AI Lab is to gather more than 100 linguists to discuss, of course, the study of language is not as simple as studying each language, it also includes how to make the computer like the human brain, accept these languages, understand these languages.

The second is logic, and if there is no excellent ability to analyze logic, then human beings are not human beings but crazy people.

The reason why madmen are called madmen is that they have lost the ability to analyze logically, and their behavior has deviated from basic logic.

There are two different approaches to artificial intelligence in computer practice.

One is to use traditional programming effects, regardless of whether the method used is the same as that used by humans or animals, which is called engineering.

In fact, this method is the most widely used and has been put into practice the most.

The advantage of this method is that it is highly targeted, and there have been successful cases in many fields, such as automated robots, they use this method, this artificial intelligence robot, he is implanted with a whole set of automated processes, but it does not have other functions.

For example, simple text recognition, as well as playing chess, some game AI, and so on.

But the same method is very limited, because if the game is simple, then the program that is formulated is also simple, and it can only be run under the established rules, to give a simple example.

For example, this artificial intelligence robot, you have formulated the rules and procedures for playing chess for him, but if you want to let it play Go, then your robot's program must be recompiled and redebugged, otherwise, it will be wrong.

The other method is simulation, which means that we not only want to see the effect, but also need to achieve it in the same or similar way as the method used by humans or animals.

For example, the current popular simulation of building human neural networks as the core of artificial intelligence robots, trying to make artificial intelligence robots practice according to the principles of human bodies.

Unfortunately, this method is too difficult to get started, in addition to being proficient in computer programming knowledge, but also proficient in biological thinking methods, so both at home and abroad, as long as researchers enter this industry, they will become the focus of support.

Although the current progress is still slow, at least, this direction is considered to be the mainstream of artificial intelligence research in the future.