In recent years, there have been many discussions about artificial intelligence. Many people have little knowledge of artificial intelligence, and media reports may not be comprehensive. So what is artificial intelligence? What is the nature of artificial intelligence? Possible application scenarios in the future? This article will take you one by one to familiarize yourself with these. What is artificial intelligence? What is artificial intelligence? To put it simply, it is to use the computer to realize the function of the human mind, that is, to realize the effect produced by the human mind through the computer. The problems to be dealt with by the artificial intelligence algorithm and the processed results are unpredictable. At present, the common reason for the common pattern recognition and robot technology in the society to be artificial intelligence is that the concept of artificial intelligence is unclear, so all advanced technologies are attributed to artificial intelligence. On the other hand, this will affect the development of artificial intelligence. For a long time, people used to refer to the system of computer processing as intelligent system. Therefore, when people see the word "artificial intelligence", they immediately associate with the intelligent system. In fact, this is a completely different concept. The intelligent system is a system implemented based on a deterministic algorithm. It is an algorithm that implements a certain objective function, and the processing result is deterministic. For example, the automatic control system, through the closed-loop PID adjustment, enables the mechanical position to reach the specified position as quickly as possible, so that the temperature reaches the specified index as soon as possible, etc. This algorithm is often a classic theory; in addition, there are many intelligent systems in pattern recognition. The classical classification algorithm, for example, using the Euclidean distance, can calculate which eigenvector data is closest to which vector data of several vector data. These are the basic algorithms for pattern recognition, and the pattern recognition system for importing these algorithms. It is a smart system. In the robot system, the walking of the robot and the movement of the arm need to be manually input into the robot system through the program, and the robot can walk and perform various arm movements according to the program of human input. The layman may think that the robot can do all kinds of actions as he likes, but it is not. For example, if a robot has an unpredictable obstacle on the road during walking, the robot will definitely be tripped; however, if the robotic system is equipped with an artificial intelligence algorithm, the robot can self-determine according to its own judgment. Ground around obstacles. Therefore, the difference between ordinary intelligent systems and artificial intelligence is that ordinary intelligent systems are classical algorithms, algorithms that only satisfy the objective function, and algorithms that solve the problem of predictability. Artificial intelligence is to imitate the problem of human brain processing. The method, or can objectively achieve the process that the human brain can achieve, the problems to be solved by artificial intelligence and the results of the treatment are often uncertain, or unpredictable in advance. Common artificial intelligence algorithm 1 Expert system based on knowledge base technology Objectively speaking, the most successful application of artificial intelligence should be the expert system of knowledge base technology. As early as 20 years ago, the author's research room developed an expert system of handwritten characters for how handwritten text looks beautiful. The rules of handwritten text are input into the knowledge base. This system can solve the problem of handwritten letters of the person in charge of the enterprise who is not fluent in handwriting. The processing of the expert system not only preserves the personalization of the written words of the person in charge of the enterprise, but also looks like Certain aesthetic effects. Compared with the backward hardware environment more than 20 years ago, today's research environment is much advanced. People can build a large social knowledge base system with large-scale network servers, and they can get unexpected application effects. For example, Japan recently established a voice knowledge base of 100,000 people on a web server, which can realize fully accessible automatic voice communication, and is currently used in the automatic translation system for foreign tourists when they are renting. The expert system based on knowledge base technology is to sum up the human brain-processed knowledge through the standard knowledge base, so that the expert system can achieve the processing functions that the human brain can achieve, so it can solve the problem that the traditional algorithm can't solve. . 2 Spatial mapping theory based on fuzzy mathematics* Advocates of spatial mapping theory have only discovered that this theory belongs to the theory of artificial intelligence. The principle of spatial mapping theory can be used to solve the pattern recognition problem of complex systems such as face recognition, image recognition or text recognition. Due to the complexity of the system, it is often difficult to find an algorithm that can be directly solved. Therefore, it cannot be processed directly by traditional algorithms like the traditional intelligent system. The spatial mapping theory based on fuzzy mathematics maps the problem of a complex system space to several A simple space problem, although only a limited problem can be solved in each simple space, but according to the combination theory, the combination of several simple system spaces can solve the problem of complex space; the most important place here is Mapping from complex space problems to simple system spaces is based on the processing of human brains, which is academically called human intervention. Because traditional mathematical methods are very rigid and difficult to intervene, fuzzy mathematics provides us with convenience. According to people's understanding of the processing object, the fuzzy mathematics Membership function can be used to solve the problem from several angles, so as to solve the complex system problem. Because such an algorithm is in accordance with the way of human mind processing, and then through the fuzzy mathematics formula to achieve the effect of solving complex problems, it should belong to the theoretical category of artificial intelligence. More than 20 years ago, Japan used this theory to obtain a very high level of application in the automatic recognition of handwritten digits. For example, in the automatic identification of the difference between the number "9" and the number "4". At that time, the commonly used scanner had a resolution of only 100 dpi, and the scanned number "9" and the number "4" were easily confused: if the stroke below the number "4" vertical line is long, it is easy to be recognized as a number "9" "," in turn, the "9" handwritten in the lower part of the stroke may be recognized as the number "4". Therefore, the researchers used the Membership function in the space mapping theory to quantify the fuzzy values ​​of the number "9" and the number "4", so that very high-precision recognition results can be obtained. This technology is mainly used in the automatic high-speed reading system of supermarket voucher, and became a representative handwritten text recognition method in Japan at that time. This theory can also be used for other applications. For example, using this theory, the unmanned system of Japan's electrified rail transit can successfully handle the random problems that may be encountered during the automatic driving process according to the driver's experience through fuzzy inference, so that the train can be smoothly and automatically driven. Run under various conditions. 3 Deep learning theory based on neural network The deep learning theory of neural networks is based on the mechanism of information processing that directly mimics the neurons of the human brain, so it belongs to the category of artificial intelligence. People have placed great hopes on this algorithm, and believe that they can get unexpected application effects on the problem of pattern recognition of such complex systems. Unfortunately, the neural network algorithm has encountered difficulties in computational complexity and slow convergence of iteration from the beginning, and it is difficult to obtain practical applications. After 2000, neural network technology was sublimated into deep learning technology, and people once again set high expectations for neural network technology. However, breakthrough technological advancements necessary for large-scale applications have not yet been seen. 4 Probability Scale Self-Organization Theory* As early as more than 20 years ago, in order to be able to resist the neural network technology, some inventors proposed the self-organization theory of probability scale. So far, people realized that this theory also belongs to machine learning theory. The starting point of this theory is man-made thinking. If a scale of maximum probability value can be found, a self-organizing method can be used to obtain a solution that exceeds the maximum probability value of the traditional algorithm for randomly distributed data. Thus, an algorithm for probabilistic scale self-organization is generated. The programming of traditional algorithms is the beginning of the program, the middle process and even the whole process of the results are designed by the programmer in advance, which is predictive, and a remarkable feature of the probability scale self-organizing algorithm is the programmer's process of processing and The results are unpredictable. Until the advent of this theory, the processing results of all statistically related algorithms stayed before the processing of this algorithm. Conversely, the results of the algorithm can be processed by all statistically relevant algorithms. Breakthroughs, and the emergence of such algorithms, the various constants that are currently considered insurmountable statistics are not the best constants. Compared with the deep learning algorithm, the self-organizing goal of the probability scale self-organizing algorithm is clear and efficient, and each iteration must have a role. The computational complexity is linear, and the ordinary mobile phone App can be realized, which has great application prospects. Because this algorithm has a big breakthrough in theory and low computational complexity, it has always shown special application effects after its appearance. For example, on a text recognition OCR system, on a document file printed by a computer. In the absence of a baseline, when the file is deflected on the scanner, the algorithm can quickly calculate the angle at which the file is deflected, simply by arranging the text. In the application of face recognition, for example, finding a part of a face in a given image, the traditional method is to first give the color data of the face, and the program finds all the colors belonging to the face according to the method of the face. Pixel. However, the colors of images taken by the same person under different light conditions vary greatly. Moreover, people with different skin colors in the world have different colors in the same skin color. Therefore, the traditional definition of a color The algorithm for searching does not meet the needs of the actual application. The introduction of the probability scale self-organizing algorithm can accurately find the face part directly through only a few self-organizations, because no matter which skin color, regardless of the color of the captured image due to the difference of the shooting light, the entire image The color of the face part has the largest distribution density value, which means that the skin color of the face part has the highest probability value. The self-organizing algorithm through probability scale solves this problem very simply, and does not need to adopt a method that can be used to automatically migrate to the face in the process of probabilistic scale self-organization through any part of the image. The part, and finally the outline of the entire face. This is an unimaginable recognition effect in the traditional pattern recognition algorithm, and such an algorithm that is superior to deep learning can be realized in an instant only through the mobile terminal. Strictly speaking, the probability scale self-organization theory should belong to machine learning theory. This theory is not well known. The reason is that it has not been disclosed as a know-how for more than 20 years. In 2014, patents were filed in the United States, Europe, Japan, and China, and patents have been obtained in the United States and Japan. 5 Optimal combination theory based on fuzzy event probability* Combination theory is the basic theory of artificial intelligence, so the breakthrough of artificial intelligence theory must rely on the breakthrough of combination theory. Combinatorial theory solves the best combination problem through graph theory. It was originally invented by Professor Liu of Taiwan University of Florida. In the early 1980s, Professor Wang, a visiting scholar from the United States, proposed the best combination theory using "entropy". It proves that the best combination results can be obtained, which has attracted the attention of the world academic community. However, the optimal combination theory using "entropy" also has the problem of large computational complexity and slow convergence, and its application is limited. How to achieve the best combination of results with high efficiency, such as large-scale integrated circuits need to achieve the combination of the smallest area, the shortest wiring length, and even consider the electrical characteristics, which is a problem that traditional combination theory cannot solve. In the 1990s, there emerged an optimal combination theory called the representative combination theory, the probability of fuzzy events. This theory quantifies the connection relationship between the various modules of a complex integrated circuit by the measure of the probability of fuzzy events, by considering the connection relationship between the various units - the closer and closer the fuzzy relationship is to be arranged as much as possible, while Also consider the probability relationship between a unit and the units that are likely to be arranged near this unit, and integrate the insignificantly small probability information and the inconspicuous small fuzzy information to obtain stable, obvious and valuable. Information. This is the breakthrough point of the fuzzy event probability theory, so the optimal combination result can be directly calculated by the efficient and multi-purpose integrated circuit optimization needs. The basis of this theory is the definition of the fuzzy value of the connection relationship between units based on human subjectivity, and therefore belongs to the category of artificial intelligence theory. Google’s $600 million startup in the UK used the artificial intelligence deep learning algorithm to develop a chess program that defeated Korean players, a sensation in the world. When I heard this news, the author immediately guessed that the maker of the program must be a chess master. Sure enough, when the Japanese NHK sent reporters to the producers of the UK interview program, it was confirmed that the programmers were outstanding in playing chess. Well, this proves from the side that this program is probably not primarily a deep learning algorithm. As a combination of theory, researchers will know that more than 40 combination elements belong to the unsolvable NP problem of Turing machine. Of course, it is an NP problem for chess. However, if you add artificial experience, NP problem can be solved. The programmer of the startup company has made the experience of playing chess into the program, so the effect of defeating the player can be achieved. From this point of view, this result is not unusual. The application of artificial intelligence in the future 1 3D moving target recognition More than 20 years ago, in the author's research room, there was a three-dimensional moving object recognition research group. At that time, everyone involved in this research knew that the three-dimensional moving object recognition technology was applied to the military. After the outbreak of the Gulf War in 1991, there was media exposure. The US missile-equipped fighters used to fire missiles at a civilian train, but fortunately they did not hit the train. In the second Gulf War in 2003, some media reported that the US missile-carrying aircraft aimed at a civilian train to launch missiles and accurately destroyed the train and caused a large number of casualties. In the face of the trains of unarmed passengers, what is the purpose of the United States launching missiles twice in succession? Researchers engaged in 3D moving object recognition know that this is the technology that the United States is demonstrating its three-dimensional moving object recognition. Because GPS positioning technology can only track fixed targets, it must rely on three-dimensional moving object recognition technology for moving targets. In the first Gulf War in 1991, the three-dimensional moving object recognition technology used in the United States was to register the feature vector values ​​of the image contours in three directions of the three-dimensional object, and to capture the three-dimensional moving objects when recognizing the three-dimensional moving object. The feature vector value of the contour of the image at any angle of the object is compared with the registered feature vector value to obtain an approximate value. This algorithm will get better recognition results under normal conditions. However, in a real war environment, the surrounding artillery fire is thick and the smoke is pervasive, which interferes greatly with the image of the moving object that is ingested. In particular, the image recognition by the contour itself violates the principle of informatics. The one-dimensional method is used to identify the two-dimensional image. The amount of information is not enough. If there is a little interference in the contour part of the image, a completely different recognition result will appear. Therefore, in the first Gulf War, the results of the US test against moving targets did not achieve the expected results. In the second Gulf War in 2003, the United States introduced artificial intelligence algorithms into mobile object recognition technology, especially for the case of severe image interference caused by harsh environments, which can strike mobile targets very accurately. At present, the mapping of terrain and landforms by drones, automatic search for victims, etc. require identification products of three-dimensional moving objects with artificial intelligence, and such high-end technology industry has high commercial value. 2 Automatic stock exchange fund hedging and financial forecasting* The most valuable technology in society is predictive technology. If people can correctly predict the stock market, they will get huge wealth. However, the current algorithm for estimating stock market results can not meet people's requirements. For example, the United States has some predictive algorithms known as military secrets. When published, the authors found that these algorithms did not have the high level of technological advancement in the confidentiality period. Therefore, even a little technological advancement in the prediction method will be very important. In the era of artificial intelligence, predictive technology will show outstanding subversive effects. First, it shows its progress in the concept of optimal prediction. In the past, people were eager to get an optimal prediction. However, according to mathematical optimization theory, the optimal solution must be based on the optimization of a given boundary condition. The optimal predictive value of artificial intelligence is the optimal value obtained by the predictor's understanding of various factors of society, the degree of understanding of the predicted target, the intelligence of the individual, and the influence of the conditions as the boundary conditions. This optimal value must be beyond the human's own solution, which will push the prediction theory to the highest stage. Secondly, artificial intelligence can be a breakthrough in optimization prediction. One is to use the probability self-organization theory to subvert the traditional statistical prediction. The other is to use the spatial mapping theory of fuzzy mathematics to predict The understanding of the relationship between social factors and predictive objects is constructed into a social model through the Membership function, and the experience extracted from sociology, philosophy, history and even the study of the influence on the prediction effect is optimized in artificial intelligence. The forecasting system can be fixed, and it can play a role in optimizing the forecasting; the third is to establish a social expert system, to build a large social large expert database for forecasting objects; and the fourth is to use machine learning related Algorithms such as analysis and regression analysis. In a word, the artificial intelligence optimization prediction system uses all the algorithms, knowledge and information related to prediction. The artificial intelligence optimization system does not calculate these algorithms separately, but builds them into an optimal decision-making platform. The calculation results of different algorithms are merged, mutual verification, information sharing with each other, and through machine learning algorithms. Finally, the self-organizing operation is performed, and the pseudo-presence is obtained, thereby obtaining the calculation result beyond the statistics, and obtaining the predicted value of the maximum probability. Introduced the decision-making platform of artificial intelligence, and all the above-mentioned factors that can play a role in forecasting are integrated with various data like the human neuron system through a new ultra-deep learning algorithm that can directly perform unsupervised learning on data. And the judgment of the prediction result as the human brain, the automatic evaluation of the data that has occurred and the results of the various algorithms of the system, the automatic correction of various parameters of the system, and the balance of the effects of various factors, Achieve automatic knowledge updates and knowledge accumulation. These are all realized instantaneously on an automatic basis. On this platform, the forecast results include stock trading and fund hedging are automatically performed. On the other hand, the function of this system that needs to be artificially processed is that the system can continuously correct the values ​​of various factors according to the operator's understanding of the predictive factors during operation, or increase the information, increase the prediction factors, or re- Adjust the framework of the forecasting strategy, etc., so that the level of forecasting continues to increase. The reason why such a system can correctly predict is the wisdom of the human mind, but human beings can't match the high-speed processing and decisive decision-making. Such a system will definitely play an immeasurable role in automatic stock trading, fund hedging and financial forecasting. effect. 3 Auto driving The auto-driving system that introduces the theory of artificial intelligence is the most important application topic in the industry. In this application field, one is the pattern recognition system that introduces the artificial learning machine learning theory, which can automatically identify the road condition information online for the automatic driving system as the basis for the car operation; the second is to automatically introduce the artificial intelligence car. Run the system. Why does auto-driving require artificial intelligence? Take brake control as an example. First of all, it is impossible for a car to run at a speed. When it is necessary to stop at a certain position, there are many situations: a skilled driver sometimes stops at the required position without braking, and sometimes gently brakes the brakes. Will step on the brakes and so on. This kind of control problem is that all the traditional automatic control theories can't be solved at present. The fuzzy inference technology introduced by artificial intelligence can formulate the experience of the skilled driver through the Membership function, and then follow the fuzzy inference algorithm to achieve close proximity to the skilled driver. Automatic driving control. The only thing mentioned here is the brake control. The automatic driving on the actual road has more complicated control problems, so it is imperative to introduce artificial intelligence algorithms. 3.4 ITC image transformation code* With the evolution of code technology, today's technology has evolved without the need to design code symbols in advance and form code graphics for stable recognition results. Under the artificial intelligence algorithm, natural paper grain, voiceprint, natural image and even body information can be directly converted into code. In recent years, the popular AR technology in the society can connect to a website through the information of a certain printed image taken by a mobile phone. Since this technology can download open source programs from the Internet, it is rapidly spreading. However, the AR technology is an image recognition algorithm. The recognition result is a file that takes up more than ten megabytes of memory, which is not conducive to network operations and the application of a large number of images. From another perspective, Google Glass and Image Retrieval need to take an image and then go online directly through this image or perform a web search. In this way, another technology - ITC (Image To Code) technology came into being. ITC uses the algorithm of spatial mapping, which can construct some features of the image into the feature vector of the image, and then organize it into a 1036 code through the algorithm of probability scale self-organizing, which can realize the shooting of any image through the mobile terminal. A code, that is, any image can be directly used as a QR code. This result allows any product identification to become a two-dimensional code without any treatment, so that all products in the world can be connected to the Internet overnight, without damaging the aesthetics of the product logo. Google Glass sees the idea that any image can be connected to the network. It can realize that any product image can be retrieved online by mobile phone, which can promote the development of online sales and play an important role in the development of VR products. Compared with the traditional AR, the ITC code capacity is only one hundred thousandth, which is convenient for mobile terminal identification. It has the characteristics of small server capacity and fast retrieval speed, and is suitable for international large-scale and large-capacity applications. Prospects for future artificial intelligence applications: Artificial intelligence technology is driving some industrial changes. For example, driverless technology has become an important trend and strategic commanding point in the future development of the automotive industry. In addition to technology companies such as Google, Uber and Tesla, the traditional automakers such as Mercedes-Benz, Audi and Toyota are also competing. Great investment in research and development. Waymo, the company's parent company's "alphabet" company, said this year that cars based on Google's autonomous driving technology have shifted from road testing to public test rides, and announced that the next phase of the goal is to provide the public with no one. Driving a taxi service. If the driverless car is really a large-scale commercial application, it will bring about great changes to the relevant industries. In the financial industry, the world's first ETF-type fund, which is entirely selected by artificial intelligence and invested in US stocks, was launched on the New York Stock Exchange in October this year. The fund is powered by EquBot in Silicon Valley and uses IBM's Watson artificial intelligence platform. EquBot CEO Hida Katua said in an interview with Xinhua News Agency that the artificial intelligence program they use automatically scans and analyzes more than 6,000 stocks every day, and independently selects stocks with rising potential and portfolios. Active management. This marks the beginning of the revolution in artificial intelligence to replace the human brain, which may have a strong impact on such knowledge-intensive industries in the future. Artificial intelligence has also penetrated into social life. In the medical field, researchers at Stanford University in Silicon Valley announced an algorithm for diagnosing skin cancer earlier this year. After training, the performance of the algorithm is comparable to that of a professional dermatologist. Artificial intelligence is also found in some digital devices around people. For example, Apple's newly released iPhoneX mobile phone focuses on the rapid face recognition function, and the related chips use artificial intelligence technologies such as biological neural networks. There is also Alexa's smart voice assistant launched by Amazon, and there is also strong artificial intelligence support behind it. The social media site Facebook has recently begun to use artificial intelligence technology to discover suicidal users and take the initiative to encourage them to communicate with friends and get out of the shadows. Finally, the rapid development of artificial intelligence in various fields has led to discussions about the future. Although the current artificial intelligence technology is only limited to its own field, it is not universal, but there is a view that artificial intelligence will develop to that step sooner or later, and should be prepared. Tesla CEO Musk believes that super artificial intelligence may emerge in the future, which may threaten human survival and human needs to cope with challenges. Our company specializes in the production and sales of all kinds of terminals, copper terminals, nose wire ears, cold pressed terminals, copper joints, but also according to customer requirements for customization and production, our raw materials are produced and sold by ourselves, we have their own raw materials processing plant, high purity T2 copper, quality and quantity, come to me to order it! Copper Connecting Terminals,Cable Lugs Insulated Cord End Terminals,Pvc Insulated Cord End Terminal,Cable Connector Insulated Cord End Terminal Taixing Longyi Terminals Co.,Ltd. , https://www.lycopperterminals.com
