Artificial Intelligence (AI). ^[21] Is a new round Scientific and technological revolution And an important driver of industrial change, ^[23] It is a new technical science that studies and develops theories, methods, technologies and application systems used to simulate, extend and expand human intelligence.

Artificial intelligence is an important part of the intelligence discipline, which attempts to understand the essence of intelligence and produce a new kind of interaction Human intelligence They react in a similar way Intelligent machine . Artificial intelligence is a very broad science, including robotics, speech recognition, Image recognition , Natural language processing , Expert system , machine learning, computer vision, etc.

The governance challenges posed by AI grand models cannot be ignored either. ^[41] Musk pointed out that the essence behind the mask of AI machine learning is still statistics. ^[35] To create a good innovation ecology, we need to do forward-looking research, and establish and improve laws, regulations, institutional systems, and ethics to ensure the healthy development of artificial intelligence. ^[41] Focusing on the future, while paying attention to risk prevention, we should also establish fault tolerance and error correction mechanisms at the same time, and strive to achieve a dynamic balance between norms and development. ^[41]

Chinese name: Artificial intelligence
Foreign name: Artificial Intelligence (AI)

subject: Computer science
Time of presentation: The year 1956
Place of presentation: DARTMOUTH Society

catalogue

Definition detail

broadcast

EDITOR

Artificial intelligence robot

About what is "intelligence", as it relates to such things consciousness (CONSCIOUSNESS), (SELF), thinking (MIND) (including UNCONSCIOUS_MIND) and other problems. It is generally accepted that the only intelligence known to man is that of man himself. But we all have such a limited understanding of our own intelligence, and of what constitutes human intelligence, that it is difficult to define what artificial intelligence is. The study of artificial intelligence often involves the study of human intelligence itself. Other about animals or other Artificial system Intelligence is also generally considered to be an artificial intelligence-related research topic.

Professor Nelson defines artificial intelligence this way: "Artificial intelligence is the science of knowledge - how it is represented and how it is acquired and used." And the other Massachusetts Institute of Technology Professor Winston said: "Artificial intelligence is the study of how to make computers do intelligent work that only humans can do in the past." These statements reflect the basic ideas and content of the discipline of artificial intelligence. That is, artificial intelligence is the study of the law of human intelligent activities, the construction of artificial systems with certain intelligence, and the study of how to make human intelligence computer To accomplish what was needed before intelligence To be competent for the work, that is, to study the basic theories, methods and technologies of how to apply computer hardware and software to simulate some intelligent human behaviors.

Since the 1970s, artificial intelligence has been known as one of the three leading technologies in the world. Space technology , Energy technology Artificial intelligence). It is also considered to be the three leading technologies of the 21st century. Genetic engineering , nanoscience , artificial intelligence). This is because it has achieved rapid development in the past 30 years, has been widely applied in many subject areas, and has achieved fruitful results, artificial intelligence has gradually become an independent branch, both in theory and practice has become a system of its own.

Artificial intelligence is a discipline that studies the use of computers to simulate certain thinking processes and intelligent behaviors of people (such as learning, reasoning, thinking, planning, etc.), mainly including the principles of computer realization of intelligence and the manufacture of computers similar to human brain intelligence, so that computers can realize higher-level applications. Ai will involve disciplines such as computer science, psychology, philosophy and linguistics. It's almost natural science and Social science The scope of all disciplines has gone far beyond the scope of computer science, artificial intelligence and Science of thinking Artificial intelligence is at the technical application level of the science of thinking, and it is an application branch of it. From the perspective of thinking, artificial intelligence is not limited to logical thinking, to consider image thinking, inspiration thinking in order to promote the breakthrough development of artificial intelligence, mathematics is often considered to be a variety of disciplines Basic science Mathematics has also entered the field of language and thinking, and the subject of artificial intelligence must also borrow mathematical tools. Maths Not only in standard logic, Fuzzy mathematics The scope plays a role, and mathematics enters the discipline of artificial intelligence, and they will promote each other and develop faster.

Research value

broadcast

EDITOR

Robots with artificial intelligence

For example, the heavy scientific and engineering calculation was originally to be undertaken by the human brain, but now the computer can not only complete this calculation, but also can do it faster and more accurately than the human brain, so the contemporary people no longer regard this calculation as "complex tasks requiring human intelligence to complete", it can be seen that the definition of complex work is changing with the development of The Times and the progress of technology. The specific goals of the science of artificial intelligence have naturally evolved with The Times. On the one hand, it is constantly making new progress, and on the other hand, it is turning to more meaningful and more difficult goals.

Usually, the mathematical basis of "machine learning" is" Statistics "," Information theory "And" cybernetics ". Other non-mathematical subjects are also included. This type of "machine learning" relies heavily on "experience." Computers need to constantly gain knowledge from the experience of solving a class of problems, learn strategies, and when faced with similar problems, apply the empirical knowledge to solve problems and accumulate new experiences, just like ordinary people. We can call this type of learning "continuous learning". But in addition to learning from experience, humans also create, or "jump learning." This is called "inspiration" or "Epiphany" in some cases. The hardest thing for computers to learn has always been Epiphany. Or to put it more strictly, it is difficult for a computer to learn "qualitative change that does not depend on quantitative change" in terms of learning and "practice," to go directly from one "quality" to another, or from one "concept" to another. Because of this, the "practice" here is not the same as human practice. The process of human practice includes both experience and creation.

This is intelligentize A researcher's dream.

In 2013, S. WANG, a data researcher at the Digin Data Center, developed a new method of data analysis that led to a new way of studying the properties of functions. The authors found that new data analysis methods provide a way for computers to learn to "create." In essence, this approach provides a fairly efficient way to model human "creativity." This pathway is given by mathematics, an "ability" that ordinary people cannot have but that computers can have. From then on, computers are not only good at calculation, but also good at creation because they are good at calculation. Computer scientists should resolutely deprive the "creative" computer of too comprehensive operational capabilities, otherwise the computer will really one day "anti-hunt" human beings.

When looking back at the process of deduction and mathematics of the new method, the author expands his understanding of thinking and mathematics. Mathematics is concise, clear, reliable, strong pattern. In the history of the development of mathematics, the brilliance of the creativity of mathematical masters shines everywhere. This creativity is manifested in the form of various mathematical theorems or conclusions, and the greatest characteristic of mathematical theorems is that they are based on some basic concepts and axioms pattern A logical structure that contains rich information expressed in a linguistic way. It should be said that mathematics is the simplest and most straightforward reflection of (at least one kind of) creative model.

Development stage

broadcast

EDITOR

In the summer of 1956, a group of visionary young scientists led by McCassey, Minsky, Rochester and Shennon met together to study and discuss a series of related issues with machines to simulate intelligence, and first proposed the term "artificial intelligence", which marked the official birth of the emerging discipline of "artificial intelligence". IBM's "Deep Blue" computer beat the human world chess champion is a perfect performance of artificial intelligence technology.

Since the discipline of artificial intelligence was formally proposed in 1956, it has made great progress for more than 50 years and has become a broad interdisciplinary and frontier science. In general, the purpose of artificial intelligence is to make the computer a machine that can think like a human. If you want to make a machine that can think, you have to know what thinking is, and by extension what intelligence is. What kind of machine is intelligent? Scientists have made cars, trains, airplanes, radios, etc., which mimic the functions of our body organs, but can they mimic the functions of the human brain? We only know that what is inside our canopy is made up of billions of them Nerve cell We know so little about this thing that mimicking it is probably the hardest thing in the world.

When the computer appeared, humans began to really have a tool that can simulate the human mind, and in the following years, countless scientists worked hard for this goal. Now artificial intelligence is no longer the patent of a few scientists, almost all the university computer department in the world has someone in the study of this subject, learning computer students must also learn such a course, in everyone's unremitting efforts, now the computer seems to have become very smart. In May 1997, for example, IBM's DEEP BLUE computer beat chess grandmaster KASPAROV. You may not notice that in some places computers help people to do other tasks that used to belong to humans, and computers play their role for humans with their speed and accuracy. Artificial intelligence has always been the frontier of computer science, computers Programming language And other computer software exist because of advances in artificial intelligence.

In December 2017, Artificial Intelligence was selected as" Top 10 Chinese media buzzwords in 2017 ". ^[1]

A spokesperson for the Second session of the 13th National People's Congress holds a press conference on March 4, 2019 Zhang Yesui It said that legislative projects closely related to artificial intelligence have been included in legislative planning ^[2] .

" Deep Learning Platform Development Report (2022) It is believed that with the maturity of the environment of technology, industry, policy and other parties, artificial intelligence has crossed the power reserve period of technical theory accumulation and tool platform construction, and has begun to enter the golden decade of industrial empowerment with scale application and value release as the goal. ^[10]

On September 25, 2021, in order to promote the healthy development of artificial intelligence, A new generation of AI ethics "Published.

April 2023, the United States Science Times Five of the leading technologies that are profoundly transforming healthcare: wearables and apps, artificial intelligence and machine learning, telemedicine, robotics, and 3D printing. ^[20]

In March 2024, the Vincennes video model Sora was introduced to wide attention. With the rapid development of artificial intelligence technology, its potential risks also appear, and the boundaries between true and false seem to become more blurred. ^[42]

In 2024, researchers at Google DeepMind and Stanford University introduced a tool based on a large language model, the Search Enhanced Fact Evaluator. Formerly known as Search-Augmented Factuality Evaluator (SAFE), it fact-checks long responses generated by chatbots. ^[48]

Scientific introduction

broadcast

EDITOR

Practical application

Machine vision, Fingerprint recognition , Face recognition , Retinal recognition , Iris recognition , Palmprint recognition , Expert system , Automatic planning Intelligent search, theorem proving, games, automatic programming, intelligent control, robotics, language and image understanding, genetic programming, etc.

Subject category

Artificial intelligence is a borderline discipline that belongs to Natural science And the social sciences.

Subject involved

Philosophy and cognitive science, Maths , neurophysiology , psychology , Computer science, Information theory , cybernetics Theory of uncertainty

Research category

Natural language processing, knowledge representation, intelligent search, reasoning, planning, Machine learning Knowledge acquisition, combinatorial scheduling problems, perception problems, pattern recognition, logic programming soft computing, inexact and uncertain management, artificial life, neural networks, complex systems, genetic algorithms

Consciousness and artificial intelligence

Artificial intelligence, by its nature, is the simulation of the information process of human thinking.

The simulation of human thinking can be carried out in two ways, one is the structural simulation, which mimics the structural mechanism of human brain and creates a "brain-like" one machine ; The second is functional simulation, which temporarily ignores the internal structure of the human brain and simulates from its functional process. The emergence of modern electronic computer is the simulation of the thinking function of human brain and the information process of human brain thinking.

Weak artificial intelligence is now constantly developing rapidly, especially after the 2008 economic crisis, the United States, Japan and Europe hope to achieve re-industrialization by robots, industrial robots are developing at a faster rate than ever before, and more led to the continuous breakthrough of weak artificial intelligence and related fields of industry, a lot of work that must be done by humans can now be achieved by robots.

While strong artificial intelligence is temporarily at a bottleneck, it still needs the efforts of scientists and humans.

Technical research

broadcast

EDITOR

The main material basis used to study artificial intelligence and the machine that can realize the artificial intelligence technology platform is the computer, and the development history of artificial intelligence is the computer Science and technology Is linked to the history of development. In addition to computer science, artificial intelligence is also involved Information theory , cybernetics , automate , bionics , biology , psychology , Mathematical logic , philology , Medical science and philosophy And many other disciplines. The main research contents of artificial intelligence discipline include: Knowledge representation Automated inference and search methods, machine learning and knowledge acquisition, knowledge processing systems, Natural language understanding , Computer vision , Intelligent robot , Automatic programming And so on.

Research method

There is no single principle or paradigm that guides AI research today. There is debate among researchers on many issues. One of the long unanswered questions is whether it should mind or nerve Aspects of simulated artificial intelligence? Or like bird biology for Aeronautical engineering Again, human biology is irrelevant to AI research, right? Can intelligent behavior be performed using simple principles (e.g logic or optimize ) to describe? Or do you have to solve a lot of completely unrelated problems?

Can intelligence be expressed using high-level symbols, such as words and ideas? Or do you need "subsymbol" processing? JOHN HAUGELAND developed the concept of GOFAI(good old-fashioned Artificial INTELLIGENCE) and also proposed that AI should be classified as SYNTHETIC INTELLIGENCE, a concept that has since been adopted by some non-GOFAI researchers.

Brain simulation

Main article: cybernetics and Computational neuroscience

In the 1940s and 1950s, many researchers explored neurology, Information theory and cybernetics The connection between. Some of the first intelligences were built using electronic networks, such as W. GREY WALTER's TURTLES and JOHNS HOPKINS BEAST. These researchers are still around all the time Princeton University Held technical association meetings with RATIO CLUB in the UK. By the 1960s, most people had abandoned this approach, although the principles were reintroduced in the 1980s.

Symbol processing

Main article: GOFAI

When digital computers were developed in the 1950s, researchers began to explore whether human intelligence could be reduced to symbol processing. The research mainly focuses on Carnegie Mellon University , Stanford University and Massachusetts Institute of Technology And each has its own research style. JOHN HAUGELAND calls these methods GOFAI(Good old-fashioned Artificial Intelligence). [33] In the 1960s, symbolic methods had great success in simulating higher level thinking on small proof programs. Methods based on cybernetics or neural networks take a back seat. [34] Researchers in the 1960s and 1970s were convinced that symbolic methods could eventually be successfully created Strong artificial intelligence And that's what they're aiming for.

Cognitive simulation Economist Herbert Simon and Alan Newell Studying human problem-solving abilities and trying to formalize them, while they lay the groundwork for the basic principles of artificial intelligence, such as Cognitive science , Operations research And business science. Their research team used psychology The results of the experiment develop programs that simulate human problem-solving methods. It's always been there Carnegie Mellon University It continued and reached its peak with SOAR in the 1980s. Based on logic not like Alan Newell and Herbert Simon , JOHN MCCARTHY The idea is that machines do not need to simulate human thought, but should try to find the essence of abstract reasoning and problem solving, whether people use the same algorithms or not. He is Stanford University The lab is dedicated to using formal logic to solve a variety of problems, including Knowledge representation , Intelligent planning and Machine learning Also working on the logical approach University of Edinburgh And led to the development of programming languages elsewhere in Europe PROLOG and Logic programming Science. "Anti-logic" Stanford researchers (e.g Marvin Minsky and Seymour Piper Found to solve Computer vision and Natural language processing The difficult problems that require specialized solutions - they argue that there are no simple and general principles (such as logic) capable of achieving all intelligent behavior. ROGER SCHANK described their "anti-logic" approach as "SCRUFFY". General knowledge base (As in DOUG LENAT's CYC "SCRUFFY"AI is an example because they have to manually code complex concepts one at a time. Based on the knowledge that large memory computers were available around 1970, researchers began with three different approaches knowledge Structure as application software. This "knowledge revolution" has resulted Expert system Developed and planned, it was the first successful form of artificial intelligence software. The "knowledge revolution" also made people realize that many simple AI software may require a lot of knowledge.

Subsymbol method

Symbolic AI stagnated in the '80s, many people thought Symbolic system It will never be possible to mimic all human cognitive processes, especially perception, robotics, machine learning, and pattern recognition. Many researchers have begun to focus on subsymbolic methods to solve specific AI problems.

Bottom-up, interface AGENT, embedded environment (robotics), behaviorism, researchers related to the field of new AI robotics, such as RODNEY BROOKS It negates symbolic artificial intelligence and focuses on basic engineering problems such as robot mobility and survival. Their work again focuses on the ideas of earlier cybernetics researchers, while proposing the use of control theory in artificial intelligence. This is consistent with the representational perception argument in the field of cognitive science: higher intelligence requires representations of individuals (such as movement, perception, and image). Computational intelligence was introduced again in the 1980s by DAVID RUMELHART and others Neural network and connectionism This and other subsymbolic methods, such as fuzzy control and evolutionary computing, belong to the study of computational intelligence.

Statistical method

In the 1990s, artificial intelligence research developed complex mathematical tools to solve specific branch problems. These tools are the true scientific method, meaning that the results of these methods are measurable and verifiable, and are also the reason for the success of AI. A shared mathematical language also allows for collaboration in existing disciplines (such as mathematics, economics, or operations research). STUART J. RUSSELL and PETER NORVIG describe these advances as nothing less than "revolutions" and "NEATS successes." Some have criticized these technologies for being too focused on specific problems and not thinking about long-term strong AI goals.

Integrated approach

Intelligent AGENT paradigm An intelligent AGENT is a system that perceives its environment and takes actions to achieve goals. The simplest intelligent agents are those that can solve a specific problem. More complex agents include humans and human organizations (e.g corporation ). These paradigms allow researchers to work on individual problems and find useful and verifiable solutions without having to consider a single approach. An AGENT solving a particular problem can use any method available - some agents use symbolic and logical methods, some are subsymbolic neural networks or other novel methods. The paradigm also gives researcher Provide a common language to communicate with other areas - e.g Decision theory and economics (The concept of ABSTRACT AGENTS is also used). The intelligent AGENT paradigm was widely accepted in the 1990s. AGENT architecture and Cognitive architecture researchers have designed systems to handle the interactions between intelligent agents in multi-angent systems. A system that contains parts of symbols and subsymbols is called Hybrid intelligent system The research on this kind of system is artificial intelligence system integration. Hierarchical control systems provide a bridge between the subsymbolic AI of the reaction level and the traditional symbolic AI of the highest level, while easing the time for planning and modeling the world. RODNEY BROOKS 'SUBSUMPTION ARCHITECTURE was an early project for a hierarchical system.

Intelligent simulation

Simulation of the way machines see, hear, touch, feel and think: fingerprint Identify, Face recognition , retinal recognition, Iris recognition Palm print recognition, expert system, intelligent search, Theorem proving , Logical reasoning Game, information sensing and dialectical processing.

Subject category

Artificial intelligence is a borderline discipline that belongs to Natural science , Social science , Technical science Three-way interdisciplinary.

Subject involved

philosophy and Cognitive science , Maths , neurophysiology, psychology, Computer science , Information theory , cybernetics , uncertainty theory, bionics Social structure and Scientific outlook on development .

Research category

Language learning and processing, knowledge representation, intelligent search, reasoning Programming, machine learning, knowledge acquisition, combinatorial scheduling problems, perception problems, pattern recognition, logic programming, soft computing, management of inaccuracy and uncertainty, Artificial life , neural network, Complex system , Genetic algorithm The most critical problem in the way of human thinking is the shaping and promotion of the machine's independent and creative thinking ability.

Safety problem

Ai is still here study But some scholars believe that letting computers have Intelligence quotient It's dangerous. It could rebel against humans. This hidden danger has also happened in a number of movies, the main key is to allow the machine to have autonomous consciousness and continuation, if so machine Having an autonomous consciousness means that the machine has the same or similar sense of creativity, self-preservation, emotions and emotions as a human spontaneous Behavior. Therefore, the security and controllable problems of artificial intelligence should be solved simultaneously from the technical level. ^[24] As the technology matures, the form of regulation may gradually change, but AI must accept that the nature of artificial supervision cannot be changed. ^[25] Generative AI can lead to massive privacy or personal information leakage issues. ^[33]

Implementation method

There are two different ways that artificial intelligence can be implemented on computers. One is to use traditional programming technology To make the system appear intelligent, regardless of whether the method used is the same as that used by human or animal organisms. This APPROACH, called the ENGINEERING APPROACH, has yielded results in several areas, such as Character recognition Computer chess and so on. The other is the MODELING APPROACH, which not only looks at the effect, but also requires the implementation method to be the same or similar to the method used by humans or biological organisms. Both GENERIC ALGORITHM (GA) and ARTIFICIAL NEURAL NETWORK (ANN) belong to the latter type. Genetic algorithms simulate the genetic and evolutionary mechanisms of humans or organisms, while artificial neural networks simulate the human or animal brain Nerve cell The way of activity. In order to get the same intelligent effect, both methods can usually be used. Using the former method requires manual elaboration of the program logic, which is convenient if the game is simple. If the game is complex and the number of characters and activity space increases, the logic becomes complex (exponentially), and human programming becomes cumbersome and error-prone. Once an error occurs, the original program must be modified, recompiled, debugged, and finally provided with a new version or a new patch for the user, which is very troublesome. In the latter approach, the programmer designs an intelligent system (a module) for each character to control, this intelligent system (module) does not know anything at first, like a newborn baby, but it can learn, can gradually adapt to the environment, cope with various complications. Such a system often makes mistakes at first, but it learns from them and is likely to correct them the next time it runs, or at least not permanently, without releasing a new version or patch. Using this method to achieve artificial intelligence requires programmers to have biological thinking methods, and the entry is more difficult. But once in the door, it can be widely used. Because this method does not need to make detailed provisions on the activity law of the role when programming, it is usually less labor-intensive than the previous method when applied to complex problems.

Gap with human

In 2023, Institute of Automation, Chinese Academy of Sciences A newly completed study by a team from the Institute of Automation of the Chinese Academy of Sciences (CAS) found that artificial intelligence-based neural networks and deep learning models are "blind" to hallucinatory contours, and the "competition" between humans and artificial intelligence is "winning a game" in hallucinatory cognition. ^[13]

Professional institution

broadcast

EDITOR

America

⒈ MASSACHUSETTS INSTITUTE OF TECHNOLOGY Massachusetts Institute of Technology

⒉ STANFORD UNIVERSITY Stanford University (CA)

⒊ CARNEGIE MELLON UNIVERSITY Carnegie Mellon University (PA)

⒋ UNIVERSITY OF CALIFORNIA-BERKELEY University of California, Berkeley

⒌ UNIVERSITY OF WASHINGTON University of Washington

⒍ UNIVERSITY OF TEXAS-AUSTIN University of Texas at Austin

⒎ UNIVERSITY OF PENNSYLVANIA University of Pennsylvania

UNIVERSITY OF ILLINOIS-URBANA-CHAMPAIGN. University of Illinois - Urbana - Champaign

⒐ UNIVERSITY OF MARYLAND-COLLEGE PARK University of Maryland, College Park

⒑ CORNELL UNIVERSITY (NY) ⒑ Cornell University

⒒ UNIVERSITY OF MASSACHUSETTS-AMHERST University of Massachusetts The AMHERST Campus

⒓ GEORGIA INSTITUTE OF TECHNOLOGY Georgia Institute of Technology

UNIVERSITY OF MICHIGAN-ANN ARBOR. University of Michigan - Ann Arbor

⒕ UNIVERSITY OF SOUTHERN CALIFORNIA University of Southern California

⒖ COLUMBIA UNIVERSITY Columbia University (NY)

UNIVERSITY OF CALIFORNIA-LOS ANGELES University of California, Los Angeles

⒘ BROWN UNIVERSITY Brown University (RI)

⒙ YALE UNIVERSITY Yale University (CT)

⒚ UNIVERSITY OF CALIFORNIA-SAN DIEGO University of California, San Diego

⒛ UNIVERSITY OF WISCONSIN-MADISON University of Wisconsin-Madison

China

1. Institute of Automation, Chinese Academy of Sciences

2. Tsinghua University

3. Peking University

4. Nanjing University of Science and Technology

5. University of Science and Technology Beijing

6. University of Science and Technology of China

7. Jilin University

8. Harbin Institute of Technology

9. Beijing University of Posts and Telecommunications

10. Beijing Institute of Technology

11. Xiamen University Artificial intelligence research institute

12. Xi 'an Jiaotong University Intelligent vehicle research institute

13. Central South University Institute for Intelligent Systems and Intelligent Software

14. Xidian University Intelligent institute

15. Institute of Image and Artificial Intelligence, Huazhong University of Science and Technology

16. Chongqing University of Posts and Telecommunications

17. Wuhan Institute of Technology

Main achievement

broadcast

EDITOR

Man-machine chess

From February 10 to 17, 1996, GARRY KASPAROV scored a 4-2 victory against DEEP BLUE.

From May 3-11, 1997, GARRY KASPAROV lost to the improved Deep Blue with a 2.5:3.5 score.

In February 2003 GARRY KASPAROV drew 3-3 with DEEP JUNIOR.

In November 2003 GARRY KASPAROV drew 2-2 with "X3D Germans" (X3D-Fritz).

Pattern recognition

Using the $pattern recognition engine, the branch has 2D recognition engine, 3D recognition engine, standing wave recognition engine and multidimensional recognition engine

2D recognition engine has been introduced fingerprint recognition, portrait recognition, text recognition, image recognition, license plate recognition; The standing wave recognition engine has been launched Speech recognition

Automatic engineering

Automatic driving (OSO system)

Money printing factory

Falcon System (YOD Drawing)

Knowledge engineering

How to use artificial intelligence and software technology to design, construct and maintain knowledge system

Expert system

Intelligent search engine

Computer vision and image processing

Machine translation And natural language understanding

Data mining And knowledge discovery

Brief history of development

broadcast

EDITOR

The legend of artificial intelligence goes back to ancient times Egypt However, with the development of electronic computers since 1941, technology has finally been able to create machine INTELLIGENCE. The term "ARTIFICIAL INTELLIGENCE" was first proposed at the DARTMOUTH Society in 1956. Since then, researchers have developed numerous theories and principles. In its short history, the development of artificial intelligence has been slower than expected, but it has been moving forward, since its emergence 40 years ago, there have been many AI programs, and they have also influenced the development of other technologies.

Computer age

An invention in 1941 revolutionized every aspect of information storage and processing. This invention, both in the United States and in Germany, was the electronic computer. The first computer took up several rooms air conditioner A large room, yes Programmer In 1949, improved computers that could store programs made it easier to input them, and the development of computer theory led to computer science, and eventually to artificial intelligence. The invention of the computer, which processes data electronically, provides a medium for the possible realization of artificial intelligence.

Although computers provided the necessary technical foundation for AI, it was not until the early 1950s that anyone noticed the connection between human intelligence and machines. NORBERT WIENER was one of the first Americans to study feedback theory. The most familiar example of feedback control is the thermostat. It compares the collected room temperature to the desired temperature and responds by turning the heater up or down, thereby controlling the ambient temperature. The importance of this study of feedback loops lies in the fact that WIENER theorizes that all intelligent activities are the result of feedback mechanisms. And feedback mechanisms are possible machine This discovery had a major impact on the early development of AI.

In late 1955, NEWELL and SIMON created a program called the LOGIC THEORIST. This program is considered by many to be the first AI program. It represents each problem as a tree model, and then selects the one most likely to reach the correct conclusion to solve the problem. The impact of the Logic Expert on the public and the field of AI research made it an important milestone in the development of AI. In 1956, JOHN MCCARTHY, considered the father of artificial intelligence, organized a society. It brought together many experts and scholars interested in machine intelligence for a month of discussion. He invited them to VERMONT for the DARTMOUTH Summer Institute on Artificial Intelligence. From then on, the field was named "Artificial intelligence." Although the DARTMOUTH Society was not very successful, it did bring together the founders of AI and lay the foundation for future AI research.

In the seven years since the DARTMOUTH meeting, AI research has advanced rapidly. Although the field is not clearly defined, some ideas from the conference have been reconsidered and used. CARNEGIE MELLON University and MIT are starting to form AI research centers. Research faces new challenges: The next step is to build systems that can solve problems more efficiently, such as reducing searches in the "logic expert"; And then there's building systems that can learn by themselves.

In 1957 a new program, the first version of the "Universal Problem Solver" (GPS), was tested. This program was developed by the same group that made "Logic Experts." GPS extends WIENER's feedback principle and can solve many common sense problems. Two years later, IBM created an AI research group. HERBERT GELERNETER spent three years making a program to solve geometric theorems.

While more and more programs were emerging, MCCARTHY was working on a breakthrough in the history of AI. In 1958, MCCARTHY announced his new achievement: LISP. LISP is still in use today."LISP" means "LIST PROCESSING" and it was quickly adopted by most AI developers.

In 1963, MIT received a $2.2 million grant from the United States government to study machine-assisted recognition. The grant comes from the Defense Advanced Research Projects Agency (DARPA). ARPA ), has guaranteed that the United States is ahead of the Soviet Union in technological progress. The program attracted computer scientists from around the world and accelerated the pace of AI research.

competition

LOEBNER (Artificial Intelligence)

Using human intelligence to create a machine brain that is comparable to the human brain (artificial intelligence) is a very tempting field for humans, and humans have struggled to realize this dream for many years. From the perspective of a language researcher, it is very difficult to make machines and humans freely communicate, and it may even be a never-answered question. Human language, human intelligence, is so complex that we have not even scratched the edge of the extension of its guiding nature.

Bulk program

In the following years, a large number of programs appeared. One of them is called "SHRDLU". SHRDLU is part of the "Miniature Worlds" project, which involves research and programming in miniature worlds (e.g. with only a limited number of geometric forms). Researchers at MIT, led by MARVIN MINSKY, have found that computer programs can solve spatial and logical problems in the face of small-scale objects. Others, such as "STUDENT", which appeared in the late 1960s, could solve algebra problems, and "SIR" could understand simple English sentences. The results of these programs are helpful in dealing with language understanding and logic.

Another development in the 1970s was expert systems. An expert system can predict the probability of a solution under certain conditions. Because of the enormous capacity of computers at that time, it was possible for expert systems to draw patterns from the data. Expert systems are widely used in the market. For a decade, expert systems have been used to predict the stock market, help doctors diagnose diseases, and instruct miners to locate mineral deposits. All of this is made possible by the ability of expert systems to store patterns and information.

In the 1970s, many new methods were used in AI development, such as MINSKY's theory of construction. In addition, DAVID MARR proposed new theories in machine vision, such as how to distinguish an image from basic information such as shadows, shapes, colors, borders and textures. By analyzing these messages, it is possible to infer what the image might be. Another result of the same period was the PROLOGE language, introduced in 1972. During the 1980s, AI advanced more rapidly and became more commercially available. In 1986, AI-related hardware and software sales in the United States reached $425 million. Expert systems are particularly in demand because of their utility. Companies like Digital Electric used XCON expert systems to program VAX mainframes. Dupont, General Motors and Boeing also rely heavily on expert systems. To meet the needs of computer experts, several companies that produce expert system-assisted crafting software, such as TEKNOWLEDGE and INTELLICORP, were founded. In order to find and correct errors in existing expert systems, other expert systems have been designed.

Daily life

People are beginning to feel the impact of computer and artificial intelligence technology. Computer technology no longer belongs to a small group of researchers in a lab. Personal computers and numerous technical magazines have brought computer technology to the fore. With foundations like the American Association for Artificial Intelligence. Because of the need for AI development, there has also been a flurry of researchers entering Private company The upsurge. More than 150 companies like DEC (which employs more than 700 people working on AI research) have spent a total of $1 billion on internal AI development groups.

other AI The field also entered the market in the 1980s. One of them is machine vision. MINSKY and MARR's work is now used in production line cameras and computers for quality control. Although still rudimentary, these systems can already tell the difference in the shape of objects from black to white. By 1985, there were more than 100 companies in the United States Machine vision system Sales totaled $80 million.

But the 1980s were not all good years for the AI industry. In 1986-87, demand for AI systems dropped and the industry lost nearly half a billion dollars. Two companies, TEKNOWLEDGE and INTELLICORP, together lost more than $6 million, about a third of their profits. The huge losses forced many research leaders to cut funding. Another disappointment has been the so-called "smart trucks" supported by the Defense Advanced Research Projects Agency. The goal of the project is to develop a device that can perform many field missions Robot . The PENTAGON halted funding for the project because of its shortcomings and lack of success.

Artificial intelligence robot (2 photos)

Despite these setbacks, AI is slowly coming back. New technologies were developed in Japan, just as they were pioneered in the United States Fuzzy logic It can make decisions under uncertain conditions. There are also neural networks, which are seen as a possible route to artificial intelligence. In short, AI was introduced to the market in the 1980s and showed practical value. It is certain to be the key to the 21st century. Ai technology put to the test During Operation Desert Storm, the military's smart devices were put to the test of war. Artificial intelligence technology is used in missile systems and early warning displays and other advanced weapons. AI technology is also finding its way into the home. The increase in intelligent computers has attracted public interest; Some applications such as speech and text recognition for Macs and IBM compatibles are already available; Using fuzzy logic, AI technology simplifies camera equipment. Greater demand for AI-related technologies is driving new advances. Artificial intelligence has and will continue to inevitably change our lives.

Contrast of strength and weakness

One of the more popular definitions of artificial intelligence, and one of the earlier definitions in the field, is defined by John McCarthy As JOHN MCCARTHY put it at the DARTMOUTH CONFERENCE in 1956, artificial intelligence is about making machines behave as if they were intelligent as humans. But this definition seems to ignore the possibility of strong AI (see below). Another definition refers to artificial intelligence as the intelligence shown by artificial machines. In general, most definitions of artificial intelligence can be divided into four categories, namely, machines "think like a human", "act like a human", "think rationally" and "act rationally". Here "action" should be understood broadly as taking action, or making a decision to take action, rather than a physical action.

BOTTOM-UP AI

The strong AI view holds that it is possible to build intelligent machines that can really reason and problem-solve, and that such machines can be considered sentient self-awareness Yes. There are two types of strong AI:

Human-like artificial intelligence, that is, machine ponder and reasoning Just like the human mind.

Non-humanoid artificial intelligence, in which machines generate perceptions and consciousness that are completely different from those of humans, and use reasoning that is completely different from those of humans.

Weak AI (TOP-DOWN AI)

The weak AI view holds that it is impossible to build intelligent machines that can actually reason and solve problems; these machines only appear to be intelligent, but do not really have intelligence and do not have autonomous consciousness.

Mainstream research is focused on weak artificial intelligence, and it is generally believed that considerable achievements have been made in this field of research. The research of strong artificial intelligence is at a standstill.

The philosophical debate over strong artificial intelligence

" Strong artificial intelligence The term was originally John Rogers Hiller Created for computers and other information processing machines, defined as:

"The strong AI view is that computers are not just a tool to study the human mind; Instead, with the proper program, computer It has a mind." (J SEARLE IN MINDS BRAINS AND PROGRAMS. THE BEHAVIORAL AND BRAIN SCIENCES,VOL. 3,1980) This is the act of instructing a computer to perform intelligent activities. Here the meaning of intelligence is ambiguous and uncertain, such as those mentioned below are examples. When using a computer to solve a problem, one must know clear procedures. However, there are many cases in which people manage to solve problems skillfully according to the HEU-RISTIC method even when they are not clear about the procedure. Such as the recognition of written words, graphics, sounds, etc., the so-called cognitive model is an example. In addition, the ability to improve due to learning and inductive reasoning, reasoning based on analogy, etc., are also examples. In addition, although the solution procedure is clear, but it takes a long time to implement, for such problems, people can find quite a good solution in a very short time, such as competitive games and so on. Moreover, a computer cannot understand its meaning without being given sufficient logically correct information, while a human being can grasp its meaning even if he is given only insufficient and incorrect information with appropriate supplementary information. Natural language is an example. Processing natural language with computers is called natural language processing.

The debate about strong AI is different from the broader one monism and dualism (DUALISM) debate. The argument goes: If the only way a machine works is to be right encoding The data is transformed, so does this machine have a mind? Hiller doesn't think that's possible. He gave the example of a Chinese room to show that if the machine is merely transforming data, and the data itself is a coded representation of something, then the machine cannot have any understanding of the data it is processing without understanding the correspondence between the code and the actual thing. Based on this argument, Hiller argued that even if a machine passes the Turing Test, it does not necessarily mean that the machine really has a mind and consciousness like a human.

As well as Philosopher Hold a different view. DANIEL C. DENNETT, in his book CONSCIOUSNESS EXPLAINED, argues that a human being is nothing more than a machine with a soul, so why do we think that humans can have intelligence and ordinary machines can't? He believes that it is possible for a data conversion machine like the one described above to have a mind and consciousness.

Some philosophers believe that if weak AI is achievable, then strong AI is also achievable. For example, SIMON BLACKBURN, in his introductory philosophy textbook THINK, argues that a person's actions that appear to be "intelligent" do not really mean that the person is intelligent. I can never know if another person is really intelligent like me, or if she/he just seems intelligent. Based on this argument, since weak AI thinks it can make a machine appear intelligent, it cannot be completely denied that the machine is really intelligent. BLACKBURN thinks this is a subjective problem.

It should be pointed out that weak AI is not completely opposed to strong AI, that is, even if strong AI is possible, weak AI still makes sense. At the very least, today's computers can do things, like arithmetic, that were thought to require intelligence more than a century ago.

Policy measure

On June 17, 2019, the National New Generation Artificial Intelligence Governance Professional Committee issued the" A new generation of AI governance principles - Developing responsible AI The framework and action guidelines for AI governance are proposed. This is an important achievement of China's efforts to promote the healthy development of a new generation of artificial intelligence, strengthen research on artificial intelligence laws, ethics and social issues, and actively promote global governance of artificial intelligence. ^[3]

Research topic

broadcast

EDITOR

The research direction of artificial intelligence has been divided into several sub-fields, and researchers hope that an AI system should have certain specific capabilities, which are listed and explained below.

Solve the problem

Early AI researchers directly mimicked humans in step-by-step reasoning, much like the way humans think when playing a board game or performing logical reasoning. In the 1980s and 1990s, use Chance and economics On the concept, AI research has also developed highly successful methods for dealing with uncertain or incomplete information.

For difficult problems, which may require a lot of computing resources, there is a "possible combination explosion" : when the problem exceeds a certain size, the computer will need an astronomical order of memory or computing time. Finding more efficient algorithms is a priority AI research project.

The human model of problem solving is usually to use the fastest, intuitive judgment, rather than the conscious, step-by-step derivation that early AI research often uses. Ai research has made progress on this "subrepresentational" approach to problem solving: Materialized AGENT research emphasizes the importance of perceptual motion. Neural network research attempts to reproduce this skill by simulating human and animal brain structures.

Knowledge representation

AN ONTOLOGY REPRESENTS KNOWLEDGE AS A SET OF CONCEPTS WITHIN A DOMAIN AND THE RELATIONSHIPS BETWEEN THOSE CONCEPTS.

Main article: Knowledge representation and General knowledge base

project

An intelligent AGENT must be able to set goals and achieve those goals. They needed a way to build a predictable model of the world (a mathematical representation of the state of the world and how their actions would change it) so that they could choose the actions that would have the most effect. In traditional programming problems, the intelligent AGENT is assumed to be the only one in the world that has influence, so what it does is already determined. However, if this is not the case, it must periodically check whether the state of the world model matches its own predictions. If not, it must change its plans. therefore Intelligent agent Must have the ability to reason in a state of uncertainty. In multi-agent, multiple agents plan to achieve certain goals through cooperation and competition, and an overall emergent behavior goal can be achieved by using evolutionary algorithms and swarm intelligence.

study

Main article: Machine learning

The main purpose of machine learning is to gain knowledge from users and input data, which can help solve more problems, reduce errors, and improve the efficiency of problem solving. For AI, machine learning has been important from the beginning. In 1956, at the original Dartmouth Summer Conference, Raymond Solomonov wrote a paper on the probability of not spying Rote learning : A machine of inductive reasoning.

Natural language processing

Main article: Natural language processing

Motion and control

Main article: Robotics

perception

Main article: Machine sensing , Computer vision and Speech recognition

Machine perception is the ability to use data input by sensors (such as cameras, microphones, sonar, and other special sensors) and then infer the state of the world. Computer vision can analyze image input. There are also speech recognition, face recognition and object recognition.

Social contact

Main article: Affective computing

KISMET, a robot with social abilities such as expressions

Emotional and social skills are important for an intelligent AGENT. First, by understanding their motivations and emotional states, agents are able to predict the actions of others (this involves factor game theory, decision theory, and detection of emotions and emotional perception that can shape people). Also, for good Man-machine interaction Intelligent agents also need to show emotion. At the very least it must appear to deal politely with humans. At the very least, it should have a normal mood in itself.

creativity

Main article: Computer creativity

A subfield of artificial intelligence that represents creativity as defined both theoretically (from a philosophical and psychological point of view) and practically (creativity that can be considered as the output of a system produced by a specific implementation, or a system that identifies and evaluates creativity). Related fields of research include artificial intuition and artificial imagination.

Multiple intelligence

Most researchers hope that their research will eventually be incorporated into a system with multiple intelligences (called strong AI) that combines all of the above skills and exceeds the capabilities of most humans. Some people think that achieving these goals may require anthropomorphic characteristics, such as Artificial consciousness or Artificial brain . Many of the above problems are considered AI integrity: to solve one of them, you have to solve all of them. Even a simple and specific task such as Machine translation The machine is required to follow the author's arguments (reasoning), to know what is being talked about (knowledge), and to faithfully reproduce the author's intentions (affective computing). Therefore, machine translation is considered to have AI integrity: it may require strong AI, just like humans.

Artificial intelligence impact

(1) Artificial intelligence Natural science The impact of... In disciplines that require the use of mathematical computer tools to solve problems, the help that AI brings is self-evident. More importantly, AI in turn helps humans finally understand the formation of their own intelligence.

(2) The impact of artificial intelligence on the economy. Expert systems penetrate deeper into all walks of life, bringing huge Macro benefit . AI also promotes the development of the computer industry network industry. But at the same time, it also brings the problem of labor employment. Due to the application of AI in science and technology and engineering, it can replace human beings to carry out various technical work and mental labor, which will cause drastic changes in the social structure.

(3) The impact of artificial intelligence on society. With the deep development of artificial intelligence technology and the wide application of robots, people will be liberated from many traditional production activities and have more leisure time. ^[45] Human society ^[46] And thus in artificial intelligence and robotics Assist Down to eternity. ^[46] AI also offers a new model for human cultural life. Existing games will gradually develop into more intelligent interactive cultural entertainment means, and today, the application of artificial intelligence in games has been deeply involved in the development of major game manufacturers.

An ideal AI society is one in which humans and AI get along well. ^[27] With the development of artificial intelligence and intelligent robots, it has to be discussed that artificial intelligence itself is advanced research, and modern scientific research needs to be carried out with a future vision, so it is likely to touch the ethical bottom line. As a sensitive issue that may be involved in scientific research, it is necessary to prevent possible conflicts as early as possible, rather than trying to resolve problems until they become unsolvable. In the development of artificial intelligence, we must first do a good job of risk control, so that the artificial intelligence developed is a blessing for mankind. ^[26]

Application field

Machine translation , Intelligent control , Expert system , robotics , language and image understanding, Genetic programming Robotic factories, automated programming, aerospace applications, massive information processing, storage and management, performing tasks of complexity or scale that cannot be performed by synthetic organisms, and so on.

It is worth mentioning that machine translation is an important branch of artificial intelligence and the first application field. However, from the perspective of the existing achievements, the translation quality of the machine translation system is still far from the ultimate goal. Translation by machine Quality It is the key to the success of the machine translation system. Chinese mathematician, linguist Professor Zhou Haizhong pointed out in his paper "Fifty Years of machine Translation" that in order to improve the quality of machine translation, we must first Work out The problem is the language itself, not the programming; It is certainly impossible to improve the quality of machine translation by relying only on several programs. The other is not yet known in humans brain In the case of fuzzy language recognition and logical judgment, how can machine translation achieve "faithfulness, faithfulness and elegance" degree It's impossible. After smart home, artificial intelligence has become a new outlet for the home appliance industry.

catchword

broadcast

EDITOR

In December 2017, artificial intelligence was selected as one of the "Top ten buzzwords in Chinese media in 2017".

After years of evolution, the development of artificial intelligence has entered a new stage. In order to seize the major strategic opportunities in the development of artificial intelligence, build the first-mover advantage in the development of artificial intelligence in China, and accelerate the construction of an innovative country and a world science and technology power, on July 20, 2017, The State Council issued the" A new generation of artificial intelligence development plan ". The "Plan" puts forward the guiding ideology, strategic goals, key tasks and safeguard measures for the development of China's new generation of artificial intelligence by 2030, laying an important foundation for the further accelerated development of artificial intelligence in China. ^[1]

On January 8, 2024, artificial intelligence was selected 2023 Labor hot words .

Background: In early 2023, chatbot ChatGPT, powered by AI (artificial intelligence) technology, swept the Internet. Subsequently, many domestic and foreign technology companies have released large artificial intelligence models. These large models have a large number of parameters and complex structure of the machine learning model, can handle massive data, complete a variety of complex tasks, such as natural language processing, computer vision, speech recognition and so on.

Watch: Will my job be replaced by AI?

This seems to be the hottest issue for workers in 2023, but it's not a new one. As the manufacturing industry ushered in the intelligent transformation represented by the robot arm, some skilled workers have already faced a career crisis.

In fact, the fundamental purpose of every technological change is to liberate people rather than replace them, and this time is no exception. Whether it is a technical worker who has given up his workstation to a robotic arm, who works with code, or an illustrator who has only recently encountered "AI painting", they have found that AI can help people complete some repetitive and standardized work, but in the face of complex situations or the need for creativity, "old master" is still irreplaceable.

AI will have an impact on certain occupations, but it will also create new jobs. For workers, adapting to new technologies and developing the skills to work with them is the only way to make AI work for us. ^[34]

Development status

broadcast

EDITOR

On July 13, 2021, the Internet Society of China issued the" China Internet Development Report (2021) ". According to the report, in 2020, the scale of the artificial intelligence industry will reach 303.1 billion yuan. ^[4]

In the field of artificial intelligence, the scale of the artificial intelligence industry in 2020 maintained steady growth, reaching 303.1 billion yuan, an increase of 15%, and the growth rate was slightly higher than the global average growth rate. The industry is mainly concentrated in Beijing, Shanghai, Guangdong, Zhejiang and other provinces, and China has made remarkable progress in the field of artificial intelligence chips, deep learning software architecture, and Chinese natural language processing. ^[4]

On June 27, 2022, at the closing ceremony of the 24th Annual Meeting of the China Association for Science and Technology, the China Association for Science and Technology solemnly released 10 cutting-edge scientific issues that have a guiding role in scientific development, including" How to achieve reliable and explainable AI technology routes and solutions ^[8] ".

December 2022 On the 9th, the Supreme People's Court issued the Opinions on Regulating and Strengthening the Judicial Application of Artificial Intelligence. . ^[12]

In March 2023, in order to implement the national Law A new generation of artificial intelligence development plan The Ministry of Science and Technology, together with the Natural Science Foundation, launched the special deployment of "AI for Science", closely combining key issues in basic disciplines such as mathematics, physics, chemistry, astronomy, and focusing on the scientific research needs in key areas such as drug research and development, gene research, biological breeding, and new material research and development. Layout of "artificial intelligence-driven scientific research" frontier science and technology research and development system. ^[15]

According to IIMedia Consulting data, in 2016, the scale of China's artificial intelligence industry has exceeded 10 billion yuan, with a growth rate of 43.3%.

In February 2023, data released by the Ministry of Industry and Information Technology show that the scale of China's AI core industry has reached 500 billion in 2022.

In 2023, the data of the "China Artificial Intelligence Industry Map" released by Qixinbao, a subsidiary of Hehe Information, shows that there were nearly 280,000 AI-related surviving enterprises in 2016 and more than 600,000 in 2022, an increase of more than 114% compared with six years ago.

By the end of 2022, the number of national AI innovation and application pilot zones established by the Ministry of Industry and Information Technology has increased to 11, covering the four strategic regions of the Yangtze River Delta, Beijing-Tianjin-Hebei, Guangdong-Hong Kong-Macao, Chengdu-Chongqing, and the middle reaches of the Yangtze River city cluster. ^[16]

On April 7, 2023, Russian Prime Minister Mikhail Mishustin said in a meeting with members of the State Duma that artificial intelligence currently accounts for about 20% of the Russian economy, and plans to reach at least 50% by 2024. ^[18]

On the evening of December 8, 2023, local time in Brussels, the European Parliament, EU member states and the European Commission reached an agreement on the Artificial Intelligence Act after nearly 40 hours of lengthy negotiations. ^[28]

On January 19, 2024, The State Council Information Office held a press conference, and Tao Qing, spokesman for the Ministry of Industry and Information Technology and director of the Operation Monitoring and Coordination Bureau, said that the number of artificial intelligence enterprises in China exceeded 4,400. ^[38]

On May 22, 2024, the British government announced that it will provide 8.5 million pounds (about $10.81 million) in government research grants to improve society's resilience to the risks posed by the development of new artificial intelligence technologies. ^[49]

Development direction

broadcast

EDITOR

" Cross-cutting frontiers of major fields 2021 (September 13, 2021) China Institute of Science and Education Strategy, Zhejiang University Release) is considered to be current Big data , Deep learning and Computing power Based artificial intelligence in Speech recognition , Face recognition Et al Pattern recognition The application of the technology is relatively mature, but for complex tasks that require expert knowledge, logical reasoning, or domain migration, AI systems are far from capable. ^[6] Deep learning based on statistics focuses on correlation and lacks causal analysis, which makes artificial intelligence systems poor in interpretability, weak in handling dynamics and uncertainty, difficult to interact with humans naturally, and easy to bring security and ethical risks in some sensitive applications. Brain-like intelligence , Cognitive intelligence , Hybrid augmented intelligence Is an important direction of development. ^[5]

2024, by School of Economics and Management, Tsinghua University The Report on Generative Artificial Intelligence Applications in the Financial Industry in 2024, jointly prepared by Du Xiaoman and other institutions, was released. According to the Report, Generative artificial intelligence The application of technology in the financial industry is still in the parallel period of technology exploration and pilot application, and it is expected that the first batch of financial institutions with large model enhancement will enter the mature application period in 1 to 2 years, and it will drive the large-scale application of generative artificial intelligence in the financial industry after 3 years. ^[39]

Cost is one of the factors restricting the development of artificial intelligence at present, and it is also the difficult point to overcome in the future development of artificial intelligence. In November 2023, the "Advanced AI Capacity Report of the Financial Industry" jointly released by a market institution and Jingdong Cloud mentioned that Ai big model The cost of landing is the focus of the industry. Hundreds of billions of level parameters, training cycles calculated in months, corresponding to the storage link means a huge cost, so artificial intelligence if you want to apply in a wider range, there is still a strong Reduce costs and increase efficiency Needs. ^[40]

Code of ethics

broadcast

EDITOR

On September 25, 2021, the 2021 Zhongguancun Forum held a plenary meeting in the Exhibition Center of the Zhongguancun National Independent Innovation Demonstration Zone, at which Xue LAN, director of the National New Generation Artificial Intelligence Governance Professional Committee, issued the" A new generation of AI ethics It aims to integrate ethics into the whole life cycle of artificial intelligence, provide ethical guidance for natural persons, legal persons and other relevant institutions engaged in AI-related activities, and promote the healthy development of artificial intelligence. ^[7]

On March 29, 2023, the UK government published a white paper on AI industry regulation, outlining five principles for AI governance such as ChatGPT. They are: security and robustness, transparency and interpretability, equity , accountability and management, and competitionability. Over the next 12 months, regulators will issue practical guidance to relevant organizations, as well as other tools such as risk assessment templates, setting out some specific rules based on the five principles. Legislation will also be pushed through parliament to enact a specific AI bill. Companies should explain when and how they use AI, and disclose the decision-making process of their systems to "expose" the risks that come with using AI. ^[17]

Application achievement

broadcast

EDITOR

American artificial intelligence company OpenAI The large language model ChatGPT About two months after its launch, it reached 100 million monthly active users in January, making it the fastest growing consumer app in history. Experts predict that ChatGPT Not only is the breakthrough of a new generation of chatbots, but it will also bring great changes to the information industry, but the risks brought by academic fraud, technology abuse, public opinion security and other risks can not be ignored. ^[37]

In June 2022, Michael Chazan Wait to use one Deep learning Artificial intelligence tools that found evidence of humans using fire 1 million years ago are considered one of the most important innovations of all time. ^[9]

South Korea plans to use AI technology for military purposes by 2027, including unmanned operation of self-propelled howitzers drone The use of. ^[11]

In the four-layer structure of artificial intelligence technology "chip-framework-model-application", Baidu is one of the few companies in the world to carry out a full-stack layout in these four layers, from Kunlun Core to fly paddle deep learning framework A word written in the mind From pre-training large models to applications such as Baidu search, there are self-developed technologies at all levels. ^[14]

in e-sports In the field, with the continuous emergence of new technologies such as artificial intelligence, AI can transform into high-end players, practice with e-sports players as "God opponents", and can also be transformed into "God teammates" to assist cooperation, helping e-sports players adjust tactics and improve skills, while improving the self-learning ability of artificial intelligence. ^[19]

Artificial intelligence has been applied to Elderly care service Industry. ninth China International Senior Care Services Expo Recently held in Beijing, intelligent pension products such as artificial intelligence robots have attracted much attention. ^[22]

The large AI model in the financial field shows a scene of "a hundred flowers blooming". Tencent Cloud released a large model solution for the financial industry, Zhaolian Finance, Du Xiaoman, Star Ring Technology, Qifu Technology, etc., have released a large model of finance. AI meets finance, bringing the possibility of innovation and change to the field of fintech. In the traditional financial business, the processing of paper data occupies a lot of manpower, material resources and time. Today, large AI models can easily read the company's annual report in ten seconds, extract important ideas and keywords from it, and generate it Financial analysis Business development forecast and other specialized content. At the same time, AI digital people have become one of the "standard equipment" for many banks to land large model business applications, and digital employees can undertake customer service and other work 24 hours a day. ^[29]

The value potential of artificial intelligence technology in the domestic financial field has attracted much attention, especially in the Mobile banking On the other hand, the user base is large and the scenarios available for application are rich, and these characteristics provide good opportunities for the landing of artificial intelligence technology. In this context, many mobile banks have recently launched iterative upgrades of new versions, focusing on improving the application of artificial intelligence technology. For example, Bank of Communications launched version 8.0 of mobile banking, the new version relies on artificial intelligence big data analysis capabilities, extracts from massive information, and launches fund big data lists to help customers make investment decisions. PSBC mobile banking version 9.0 to create "AI space + digital staff + video customer service" service. Among them, the AI space is entered through the drop-down home page of mobile banking, which intuitively shows customers the information of this month's income and expenditure, common payments, and recent payees, and provides customized services for customers. ^[36]

Financial institutions began to try to apply artificial intelligence technology in the field of risk prevention and control, and use scientific and technological innovation to prevent financial risks. At present, China has actively tried and explored in the field of "artificial intelligence + risk control", which has certain first-mover advantages compared with international financial industry peers. At the World Artificial Intelligence Conference in July 2023, Tencent Released to the public Large model of financial risk control . In November of the same year, Tencent and China Academy of Information and Communications Technology , University of Science and Technology of China , Nanyang Technological University, Singapore Central Plains Consumer Finance, WeBank and other research institutions and financial institutions jointly developed the world's first large-scale model of financial risk control international standards. ^[43]

The new round of job hunting season, "artificial intelligence + interview" (AI interview) is becoming more and more common. As a product of scientific and technological development, AI interview has replaced a lot of repetitive work in the preliminary screening stage in recruitment, which indeed brings convenience to many employers. ^[50]

Related works

broadcast

EDITOR

" Visual reading artificial intelligence Can machines really think? Is the human mind just a complex computer program? This book looks at artificial intelligence, one of the thorniest scientific problems of all time, and focuses on some of the major topics behind it. Artificial intelligence is not just a fictional concept. Half a century of research into the structure of intelligent bodies has shown that machines can beat the greatest human chess players and that humanoid robots can walk and interact with humans. Despite long proclamations that intelligent machines are just around the corner, progress has been slow and difficult. Consciousness and environment are two difficult problems in research. How on earth should we build intelligent machines? It's supposed to work like a brain? Does it need a body? From Turing's seminal research to the leap forward in robotics and new artificial intelligence, this book provides a clear picture of the development of artificial intelligence over the past half century.

" The future of artificial intelligence It explains what intelligence means, explains how the brain works, and tells us how we can build truly intelligent machines - machines that are no longer simply imitations of the human brain, but that are far more intelligent than the human brain in many ways. Hawkins believes that from artificial intelligence to neural networks, previous efforts to replicate human intelligence have not been successful, and the reason is that people do not really understand what intelligence is and the human brain. Intelligence is the ability of the human brain to compare the past and predict the future. The brain is not a computer, and it does not follow the input and produce output in a step-by-step manner. The brain is a vast memory system that stores experiences that, to some extent, reflect the real structure of the world, can remember the sequence of events and their interrelationships, and make predictions based on memory. The brain's memory-prediction system underlies intelligence, sensation, creativity, and perception...

" Philosophy of artificial intelligence The philosophy of artificial intelligence is a branch of philosophy developed along with modern information theory and computer technology. This book is a collection of fifteen representative papers by scholars in the field of artificial intelligence research, which have made groundbreaking contributions to the development of computer science and the establishment of the philosophy of artificial intelligence. These articles summarize the history of the development of artificial intelligence, the trend of the development of the discipline, and the important topics in artificial intelligence. Among these landmark works are: the father of modern computer theory Alan Turing "Computers and Intelligence"; American philosophy Jessel's "Mind, Brain, and Program"; "Distributed Representations" by J. E. Hinton and others, and "Escape from the Chinese House" by M. A. Boden, the British artificial intelligence scholar who edited the book.

" Artificial Intelligence: A Modern approach With detailed and rich information, this book comprehensively expounds the core content of the field of artificial intelligence from the perspective of rational agents, and deeply introduces the main research directions, which is a rare comprehensive teaching material. The book is divided into eight parts: Part I "Artificial Intelligence", Part II "Problem solving", Part III "Knowledge and Reasoning", Part IV "Planning", Part V "Uncertain Knowledge and reasoning", Part VI "Learning", Part VII "Communication, Perception and Action", and Part VIII "Conclusion". The book details a large number of basic concepts, ideas, and algorithms, describes the latest advances in various research fields, and collates detailed historical documents and events. Therefore, this book is suitable for researchers and students of different levels and fields, and can be used as a textbook or teaching bibliography for undergraduates and graduate students in the field of information and related fields, and can also be used as a reference book for scientific research and engineering technicians in related fields.

Related event

broadcast

EDITOR

On December 20, 2023 local time, the British Supreme Court published a judgment showing that an American computer scientist lost his application for a patent for an invention created by his artificial intelligence system. The UK Intellectual Property Office had previously rejected his application for patent registration on the grounds that the inventor must be a human or a company, not a machine, and then Taylor appealed to the UK Supreme Court, which was rejected on the 20th because under UK patent law, "the inventor must be a natural person". ^[30]

According to reports, on November 2, 2023, the Israeli military claimed that it used artificial intelligence (AI) technology to pinpoint and strike targets in the new Israeli-Palestinian conflict. Currently, the Israeli military mainly uses two AI systems for military operations. One is a system for processing large amounts of data and selecting targets for air strikes; The other is an AI model used to calculate ammunition loads and plan raids. Military officials say it is now possible to select targets for air strikes and carry them out in minutes, an unprecedented speed. ^[31]

In December 2023, it was selected as the top ten hot spots in the theoretical vision of 2023. ^[32]

On March 15, 2024, The China Media Group March 15 Gala in 2024 The topic exposed the illegal use of artificial intelligence for fraud activities. ^[44]

On March 21, 2024, the United Nations General Assembly voted to adopt its first draft resolution on artificial intelligence (AI) to ensure that the new technology benefits all countries, respects human rights and is "safe, reliable and trustworthy." ^[47]

Hit the road

Growth task Introduction to editing Editing rule Edit by oneself

I have a question.

Content questioning Online customer service Official post bar feedback

Complaint and suggestion

Report bad information The article was not appealed Complaint infringement information Blocked query and unblocked

Beijing Public network Anbei 11000002000001