In recent ten years, artificial intelligence, as an extremely fashionable word in the field of science and technology, has been an industry that investors and practitioners are scrambling to set foot in. Especially in the period of economic weakness, artificial intelligence has become a lifeline for software Internet-related industries and a criterion for judging the technical level of enterprises in the industry. Words must be called "artificial intelligence" has become a face problem for enterprises.

strangely, artificial intelligence, as a big trafficker in the software industry, has a low threshold for employment. Any software internet enterprise that needs a bright spot can claim that it has the ability of artificial intelligence by calling several library packages. The technical bottom line of an industry depends on the level of ordinary employees in the industry. Therefore, the current technical bottom line of artificial intelligence should only be enough to write codes such as "import".

In recent years, artificial intelligence projects have mushroomed online. As a component or even a bright spot of information projects, they are constantly online around us, whether it is government infrastructure, B-side platform or consumer applications. The launch of such artificial intelligence projects, if only a gimmick, is at best a waste of trial and error costs. If it's a real application, it's really scary.

why do we call these technologies artificial intelligence? It is because we hope that through technical means, our products will have some capabilities that only talents have. Since artificial intelligence comes from "people", then "what is the boundary of human ability and the application boundary of artificial intelligence?" It has become the core issue of the artificial intelligence industry.

At present, the main research fields of artificial intelligence are human sensory imitation, interaction and learning. But no matter which application direction, we must first know that artificial intelligence, as a computer science, is based on how to look at the world digitally. We know that whether it is computer vision, image recognition analysis, voice analysis, machine learning and other mainstream artificial intelligence topics, the premise of technology is to transform images, videos, voices and method models into binary information that computers can recognize. In other words, to understand the boundaries of artificial intelligence applications, we must know to look at the boundaries of the world in a purely digital way.

as we all know, human rational thinking ability mainly comes from analysis and comprehensive judgment. The former uses deduction to infer the inevitability of things. The latter can not give the inevitability of things by induction, but only the probability in the sense of probability.

We can deduce countless theorems from one axiom by deduction, and get the axiom by induction, which basically constitute all the weapons of modern scientific development. But so far, our logical thinking ability can't be used to make value judgments, and when we think about problems in real scenes, we often unconsciously combine the current scenes and integrate value factors to make judgments.

Therefore, only when we clearly realize that this judgment does not need to integrate any value factors can we hand over the job to artificial intelligence. At present, our understanding is often too casual.

For example, when we use cameras to identify violations and stop them, we must consider making a mechanism to release emergency vehicles. If the car is a seriously ill patient or a pregnant woman in labor. Blocking must be able to intervene and remove in time, otherwise the consequences will be unimaginable. Therefore, we find that the application can only be used in the scope of identification, and it is not suitable for arbitrarily increasing business in identification.

nowadays, intelligent voice assistant and intelligent customer service have become the necessary applications of various internet companies. The boundary of speech recognition is actually quite obvious. Basically, every user has experienced all kinds of disharmony and made many jokes during the interaction with the intelligent assistant. In the exploration of language recognition, the biggest problem is to understand the differences between everyday language and artificial language.

We all know that in daily language communication, people judge the meaning of each other's speech or words, including context, the speaker's expression, regional culture and even sub-cultural attributes, in addition to the language itself. For example, when a rogue says to a woman, "I will be good to you", the meaning of this sentence is obviously completely opposite to the situation. At this time, artificial intelligence is difficult to make any correct analysis.

Wittgenstein put forward the theory of "language game" in his later period. The rational understanding of language is a process of systematically explaining the received text according to the relationships among contents, viewpoints and scenes. Simply limited to the semantics of the language itself as the basis for interaction, but people are treated as machines.

Therefore, we should understand that artificial intelligence should be more confined to meaningful information retrieval and query in language recognition and interaction, and should not be overused in dialogue process, otherwise, the more communication, the more the lack of "intelligence".

as we all know, the core of machine learning technology is actually to imitate human induction of information. Compared with people, machine learning can input more dependent variables in the process of induction, thus finding more connections. Therefore, the main value of machine learning is to tune a given model.

We can obviously find that the given model is the premise for the machine to "learn" and is naturally the boundary of machine learning. From beginning to end, mankind's grasp of knowledge also comes from induction. When we see the sunrise and sunset of the sun, and the weather is cloudy or sunny, we can give the corresponding time and season models and arrange our lives accordingly. The most fundamental point in this process is that human beings can draw universal laws through the induction of phenomena, thus creating new causal relationships. The new causal connection becomes an axiom, which is further applied to the new deductive inference and produces a new knowledge system.

in this process, we created a new symbol order, not just tuning the original symbol model. Just as a machine is hit by an apple 1 million times, Newton's law cannot be obtained.

Nowadays, artificial intelligence can not only play Go, but even compose poems. Even playing Go has achieved the goal of beating the invincible hand of human beings, and the poems written are often like fake chaos. Even AlphaGo has become a banner of artificial intelligence.

We should not only ask whether these fields that only humans can participate in can be completely replaced by artificial intelligence in the future. In fact, it is very simple. If a field can develop endlessly by inputting rules, it can indeed be replaced by artificial intelligence. But is art the rule? Today, when we read poems, listen to music and appreciate paintings, we may think that the author's level is limited because of a fixed technique and standard. With a certain standard, the work is ok, which has been practiced by our ancestors for thousands of years.

You see the times, yourself and the past time in a painting. You are immersed in the beautiful feeling brought by the painting. No matter whether it has those so-called techniques or not, you will experience a kind of comfort. This kind of comfort does not come from reason, but from your will, and from the fundamental reasons why you are human. Art was originally the inspiration of human beings in history. Through the beauty of art, we felt the true meaning of the world in an instant, which is unclear and unclear. How can the machine evaluate it?

people are still limited. As Kant revealed, human beings always look at the world within the framework of human limitations. The existence of artificial intelligence allows us to use machines to infinitely approach our own framework. We don't know how our framework is built.

When the NPC in "westworld" suddenly gets the subjectivity, he/she starts to think and judge, then artificial intelligence seems to have reached the end of the discipline. And what this "sudden" is, even human beings themselves can't know. And this gap is the task of human beings and the limitation of human beings, which we still don't know, let alone machines.

artificial intelligence is artificial "intelligence", but in fact it is not artificial "ignorance".