Some people say that 2016 is the first year of artificial intelligence, and artificial intelligence technology has mushroomed in all walks of life. At the beginning of 2017, I began to translate Artificial Intelligence (2nd Edition). When I was about to finish the first translation of the book, news came from the scientific and technological circles that Alphago had defeated The Go player Ke Jie. Thus, the book became a veritable and unpublished “ancient book”. The book can be called a classic textbook, full of contents, clear logic, allusion to the classics, is a rare textbook on artificial intelligence, it is also known as the “encyclopedia of artificial intelligence field”.
Artificial Intelligence (2nd Edition)
By Stephen Lucci
Back in 1997, when Deep Blue beat Kasparov, there were optimistic predictions that computers would not be able to beat humans so easily in the ancient Oriental wisdom of Go, which symbolized the last high ground of human intelligence. Only 20 years later, however, the predictions were shattered.
From Alan? From Turing’s crack of the Enigma cipher machine to the victory of the Second World War, to the Dartmouth symposium where the term ‘artificial intelligence’ was coined, to the present day, artificial intelligence has undergone 60 years of development. During this period, “mountains and rivers doubt no way, bright flowers and another village”, artificial intelligence has experienced three waves, two cold winter baptism. At present, promoted by deep learning algorithms, ARTIFICIAL intelligence, with cloud computing, big data and convolutional neural network, has broken through the bottleneck of natural language speech processing and image recognition, bringing earth-shaking changes to human beings.
It is not an exaggeration to describe the development of artificial intelligence with the poem “Suddenly like a night of spring breeze, thousands of trees pear blossom”. “If you’re lucky, machines can treat you like a pet,” the industry joke goes.
A joking words but said how many people sad, artificial intelligence has begun to replace human in all aspects of the possibility: in the future in the production workshop, we can no longer see human workers busy figure; In supermarkets, we don’t see cashiers working. In restaurants, we don’t see chefs or waiters. Artificial intelligence can help us complete many tasks and assist us in making decisions.
The debate about whether artificial intelligence is a gift to mankind or a disaster to mankind has been raging and divided. Artificial intelligence may be Pandora’s box, but there is hope that has carried it through a rocky 60 years.
On the optimistic side, in the future, science fiction stories may appear in daily life, and labor may become a need to keep healthy. But it is still too early to be a done deal. Referred to as the book in the robot parts, and artificial intelligence in the midst of a toddler “infancy”, when I wrote this article translator sequence, humanoid robots or very early in the sports ability, therefore, someone said a sentence joke: “if you want to block the terminator, shut the door just more difficult to grasp the door technology (robot).”
The book can be regarded as a classic textbook, full of contents, clear logic, allusion to the classics, is a rare textbook on artificial intelligence. Artificial intelligence covers everything, including natural language processing, knowledge representation, intelligent search, planning, machine learning, artificial neural network complex systems, data mining, genetic algorithms, fuzzy control, etc. Facing the rapid development of artificial intelligence and massive knowledge, computer science and engineering related professional readers, and the circle, rather than retreat webs, solid foundation. “Paper come zhongjue shallow, and must know this to practice.”
To learn artificial intelligence, readers need to guard against arrogance, carefully understand the algorithm, and translate the algorithm into a computer program. Therefore, I recommend that after reading a chapter, readers write the code themselves, and actually run the program on a machine. As a Chinese saying goes, Rome wasn’t built in a day. To become a leader in artificial intelligence, readers need to be prepared to fight a long and tough battle, and keep learning new technologies and bringing forth new ones. Only in this way, can become the mainstay of the society, leading the trend of The Times.
Why did I write Artificial Intelligence (2nd Edition)
In 2006, James Moor, a professor of philosophy at Dartmouth College, asked me to organize an exhibition game of computer games at AI@50, held to celebrate the 50th anniversary of the Dartmouth Summer Conference. At the Dartmouth Summer Seminar, John McCarthy coined the term “artificial intelligence.”
Some of the original Dartmouth attendees attended AI @ 50, Among them were John McCarthy, Marvin Minsky, Oliver Selfridge and Ray Solomonoff. Professor Lucci also attended AI @ 50, and soon after, he agreed to collaborate with me on an AI textbook.
Our view is that ARTIFICIAL intelligence is made up of People, ideas, methods, machines, and outcomes.
First, artificial intelligence is made up of humans. Humans have ideas and turn those ideas into methods. These ideas can be represented by algorithms, heuristics, programs, or systems that serve as computational backbones. Finally, we get the products of these machines (programs), which we call the “results.” Each result can be measured in terms of value, effectiveness, efficiency, etc.
We found that existing books on ARTIFICIAL intelligence often failed to mention some of these areas. Without humans, there would be no artificial intelligence. So we decided to introduce the people who contributed to the success of AI by adding a “People anecdote” column to the book. The people we introduce throughout the book’s 17 chapters include the people who came up with the ideas and implemented the development methods.
Ai and computer science are relatively young compared to other sciences such as mathematics, physics, chemistry and biology. But AI is a truly interdisciplinary discipline, combining many elements from other fields.
Machines/computers are the tools of artificial intelligence researchers. Machines/computers allow researchers to experiment, learn and improve problem-solving methods that can be applied to many interesting fields that may be beneficial to humans. Finally, by applying AI to a wide variety of problems and disciplines, we get measurable results that remind us that AI must also be interpretable. In many places in this book you will find a discussion of the difference between ‘performance’ and ‘competence’. Both will be needed as AI matures and advances.
So far, by personally teaching ai courses and reading AI textbooks, we have found that most of the available textbooks lack one or more of these areas. The names and great contributions of Turing, McCarthy, Minsky, Michie, McLelland, Feigenbaum, Shortliffe, Lenat, Newell and Simon, Brooks and many others should be familiar to students.
However, this is not a history book! We felt that the subject was so interesting, so broad, and so full of potential that the book should be enriched by the fascinating ideas and excellent work of the people working in the field.
What is the difference between version 2 and Version 1?
It has been a long time since the first edition of this book was published. Artificial intelligence concepts, methods and systems are increasingly being integrated into People’s Daily activities. For example, at the time version 1 was written, many cars were built to be capable of parallel parking; It has become commonplace to equip cars with crash-avoidance systems. Technology that science fiction enthusiasts fantasize about (such as drones and robots) is now a reality, and more and more drones and robots are becoming a part of everyday life.
GPS systems, mobile apps and social networks that surfaced in the 2000s or earlier are now ubiquitous. These technologies, including the best transportation routes, health advice and personal attendants, are already used in every aspect of people’s lives, and each often uses some form of ARTIFICIAL intelligence.
Advances in natural language and speech processing have dramatically changed the way humans interact with machines. The second edition adds a chapter 10, which introduces and discusses decision trees for machine learning. Thus, together, chapters 10, 11 (Machine learning part 2: Neural Networks) and 12 (Nature-inspired Search) provide the basis for further research. A new section (Section 13.10) of Chapter 13 (Natural Language Processing) introduces the theory, methods, and applications of speech comprehension. Chapter 13 also adds a section on metaphors in natural language processing. Chapter 15 provides an overview of the field of robotics, including recent applications, and concludes with a look into the future along with Chapter 17 (Memorabilia). New exercises have been added to many chapters.
Alan Turing
Overview of Artificial Intelligence
In early times, man had to fight nature with tools and weapons like the wheel and fire. Gutenberg’s invention of the printing press in the 15th century brought about widespread changes in people’s lives. In the 19th century, the industrial Revolution used natural resources to develop electricity, which led to the development of manufacturing, transportation and communication. In the 20th century, mankind continued to advance through the exploration of the sky and space, through the invention and miniaturization of computers, which became personal computers, the Internet, the World Wide Web and smartphones. The past 60 years have seen the birth of a world in which vast amounts of data, facts and information have to be converted into knowledge (one example is data contained in the human genetic code, as shown in Figure 1.0). This paper introduces the conceptual framework of artificial intelligence, and expounds the fields and methods of its successful application, recent history and future prospects.
Figure 1.0 contains data in the human genetic code
Definition of artificial intelligence
In everyday speech, the word “artificial” means synthetic (i.e., man-made), which usually has a negative connotation, i.e., “man-made objects are inferior in quality to natural objects. However, man-made objects are usually superior to real or natural objects. Artificial flowers, for example, are buds or flower-like objects made of silk and thread. They do not need sunlight or water as nourishment, but can provide a practical decorative function for a home or business.
Artificial flowers may not have the same feel or smell as natural flowers, but they look exactly like real ones.
Another example is artificial light produced by candles, kerosene lamps, or electric bulbs. Obviously, we can only get sunlight when the sun is in the sky, but artificial light is superior to natural light in that we can get it all the time.
Finally, consider that artificial transportation devices (such as cars, trains, planes, and bicycles) offer many advantages in speed and durability over running, walking, and other natural forms of transportation (such as horseback riding). But artificial forms of transportation also have some significant downsides — the earth’s ubiquitous highways, an atmosphere filled with car exhaust, and peace of mind (and sleep) often interrupted by the noise of airplanes.
Like artificial light, artificial flowers and transportation, ARTIFICIAL intelligence is not natural, but artificial. To determine the advantages and disadvantages of AI, you must first understand and define intelligence.
What is thinking? What is intelligence?
The definition of intelligence can be more elusive than that of man. R. Sternberg offers the following useful definitions on the subject of human consciousness: Intelligence is the individual’s cognitive ability to learn from experience, think rationally, remember important information, and cope with the demands of daily life.
We are all familiar with the problem of standardized testing, for example, given a sequence of numbers: 1,3,6,10,15,21. Ask for the next number.
You might notice that the difference between consecutive numbers is spaced by 1. For example, from 1 to 3 the difference is 2, from 3 to 6 the difference is 3, and so on. So the correct answer to the question is 28. The question is designed to measure our proficiency in identifying salient features in patterns. We find patterns through experience.
Try your luck with the following sequence:
A. 1, 2, 2, 3, 3, 3, 4, 4, 4, 4,?
B. 2, 3, 3, 5, 5, 5, 7, 7, 7, 7, 7,?
Now that you have settled on the definition of intelligence, you may have the following questions.
(1) How to judge whether some people (or things) have intelligence?
(2) Do animals have intelligence?
(3) If animals have intelligence, how can their intelligence be assessed?
Most people can answer the first question easily. We assess their intelligence by interacting with other people (making comments or asking questions, for example) to see their reactions, repeating this process several times a day. Although we don’t have direct access to their thoughts, we believe that the indirect approach of q&A can provide us with an accurate assessment of internal brain activity.
If you insist on using question-and-answer methods to assess intelligence, how do you assess animal intelligence? If you’ve ever owned a pet, you probably already have your answer. Puppies seem to remember people they haven’t seen in a month or two and can find their way home when they get lost.
Kittens often act excited when they hear the sound of cans opening at dinnertime. Was it simply a matter of Pavlovian reflexes, or did the cat consciously associate the sound of the can with the pleasure of dinner?
An interesting anecdote about animal intelligence is that around 1900 there was a horse in Berlin, Germany, called Clever Hans, who was said to be proficient in mathematics (see Figure 1.1).
Figure 1.1 “Clever Hans” — a horse doing calculations?
The audience was stunned when Hans added or calculated square roots. Thereafter, it was observed that Hans could not perform well without an audience. In fact, Hans’ genius lies in his ability to recognize human emotions, not his mastery of mathematics.
Horses generally have a keen sense of hearing, and as Hans approached the correct answer, the audience became relatively excited and their hearts beat faster. Perhaps Hans had an uncanny ability to detect these changes and get the right answer. While you might be reluctant to attribute Hans’s behavior to intelligence, you should refer to Sternberg’s earlier definition of intelligence before jumping to conclusions.
Some creatures only exhibit swarm intelligence. For example, ants are simple insects, and the behavior of individual ants can hardly be classified under the subject of ARTIFICIAL intelligence. But ant colonies, on the other hand, have shown remarkable ability to solve complex problems, such as finding the best route from nest to food source, carrying heavy loads and forming Bridges. Collective intelligence results from effective communication between individual insects. Chapter 12 discusses emergent intelligence and cluster intelligence relatively much in its discussion of advanced search methods. Brain mass and the ratio of brain mass to body mass are generally regarded as indicators of animal intelligence. Dolphins are on par with humans on both measures. Dolphin breathing is autonomously controlled, which can be explained by the excessive brain mass and the interesting fact that the two halves of the dolphin brain alternate in hibernation.
On animal self-awareness tests, such as the mirror test, the dolphins scored well, recognizing that the image in the mirror was actually their own image. Visitors to parks like SeaWorld can see dolphins perform complex tricks. This suggests that dolphins have the ability to remember sequences and perform complex body movements.
Tool use is another “touchstone” for intelligence, and it is often used to distinguish Homo erectus from previous human ancestors. Dolphins and humans share this trait. For example, dolphins use deep-sea sponges (multicellular animals) to protect their mouths when foraging. Clearly, intelligence is not a uniquely human trait. Many life forms are intelligent to some extent.
You should ask yourself the following questions: “Do you think being alive is a necessary prerequisite for having intelligence?” “Or” Can inanimate objects, such as computers, possess intelligence?” The stated goal of AI is to create computer software and/or hardware systems that can match the human mind — in other words, exhibit characteristics associated with human intelligence. A key question is “Can machines think?” More generally, you might ask, “Do humans, animals, or machines possess intelligence?”
At this point, it is wise to emphasize the difference between thinking and intelligence. Thinking is a tool for reasoning, analyzing, evaluating and forming ideas and concepts. Not all thinking objects are intelligent. Intelligence may be efficient and effective thinking. Many people view this question with prejudice, saying, “A computer is made of silicon and electricity, and therefore cannot think.” Or go to the other extreme: “Computers perform faster than humans, and therefore have higher IQs.” The truth probably lies somewhere between these two extremes.
As we have discussed, different animal species have different degrees of intelligence. We will describe software and hardware systems developed in the field of artificial intelligence, which also have varying degrees of intelligence. We don’t care much about assessing animal intelligence and haven’t developed standardized animal INTELLIGENCE tests, but are interested in tests to determine whether machine intelligence exists.
Perhaps Raphael put it best: “ARTIFICIAL intelligence is a science that allows machines to do what humans need intelligence to do.”
The Turing test
The previous section asked “How do you determine intelligence” and “Do animals have intelligence?” These two problems have been solved. The answer to the second question is not necessarily a simple “yes” or “no” – some people are smarter than others, some animals are smarter than others. Machine intelligence runs into the same problem.
Alan Turing sought an operable answer to the question of intelligence, separating function (what intelligence can do) from implementation (how intelligence can be achieved).
Additional information
Abstraction is a strategy that ignores the implementation of objects or concepts (such as internal work) so that you can get a clearer picture of artifacts and their relationship to the outside world. In other words, you can treat the object as a black box and focus only on the input and output of the object (see Figure 1.2).
Figure 1.2 Input and output of black box
In general, abstraction is a useful and necessary tool. For example, if you want to learn how to drive, it might be a good idea to treat the car like a black box. Instead of learning the automatic transmission and powertrain hard at first, you can focus on system inputs such as gas pedals, brakes, turn signals, and outputs such as forward, park, left, and right. Also use abstract data structure course, so if you want to understand the behavior of the stack, you can focus on basic stack operations, such as pop (pop up one) and push (insert one), rather than into how to construct a list of details (for example, using the linear list or circular linked list, or use the links list or a continuous distribution of space).
Definition of the Turing test
Alan Turing came up with two simulation games. In simulations, a person or entity acts as if it were another person. In the first simulation, a person was in a room with a curtain in the center, and there were two people on either side of the curtain. The person on one side, called the questioner, had to determine whether the person on the other side was a man or a woman. The questioner (whose gender doesn’t matter) accomplises this task by asking a series of questions. The game assumes that a man is likely to lie in his answers, while a woman is always honest. To make it impossible for the questioner to determine gender from speech, communication is conducted by computer instead of speech, as shown in Figure 1.3. If there is a man on the other side of the curtain and he succeeds in deceiving the questioner, then he wins.
Figure 1.3 The first Turing simulation game
In its original form, the Turing test involved a man and a woman sitting behind a curtain, and the questioner had to correctly identify their gender (Alan may have been inspired to invent it by a popular game of the era). The game also inspired him to test machine intelligence). As Erich Fromm wrote [8] : Men and women are equal, but not necessarily the same. For example, different genders have different knowledge about colors and flowers, and spend different amounts of time shopping. What do men and women have to do with intelligence? Alan thought that there might be different types of thinking, and that it was important to understand and tolerate these differences. Figure 1.4 shows the second version of the Turing test.
Figure 1.4 The second Turing simulation game
The second game is more suited to ai research. The questioner was still in the curtained room. This time, it could be a computer or a person behind the curtain. The machines here play the male role and occasionally lie, but the people are always honest. The questioner asks questions and then evaluates the answers to determine whether he is communicating with a human or a machine. If the computer successfully deceives the questioner, it passes the Turing test and is therefore considered intelligent.
It is well known that machines can perform arithmetical calculations many times faster than humans. If the person behind the curtain can get a Taylor series approximation of trigonometric functions in a matter of microseconds, it is easy to tell that the person behind the curtain is a computer and not a person.
Naturally, the chances that a computer can successfully trick the questioner on any Turing test are very small. This test is performed many times in order to get a valid “barometer” of intelligence. Again, in this original Version of Turing’s test, people and computers were behind a curtain and the questioner had to correctly identify them.
Additional information
The Turing test
No computer system passes the Turing test. In 1990, however, philanthropist Hugh Gene Loebner launched a competition to achieve the Turing Test. The first computer to pass the Turing test will be awarded a gold medal and the $100, 000 Loebner prize. Also, the computer that performs best in the competition each year will be awarded a bronze medal and a prize of about $2,000.
What questions would you ask in the Turing Test? Consider the following example:
· (1 000 017)? How much is? A calculation like this might not be a good idea. Remember, the computer is trying to trick the questioner. The computer may not respond in fractions of a second to give the correct answer, it may intentionally take longer, and perhaps make mistakes because it “knows” that humans are unfamiliar with these calculations.
· What is the current weather situation? Suppose the computer probably doesn’t look out the window, so you might try to ask about the weather. But the computer is usually connected to the World Wide Web, so before answering, it also connects to the weather website.
· Are you afraid of death? Because computers can’t fake human emotions, you might ask this question or something like it: “How does black make you feel?” “Or” How does it feel to be in love? But remember, you’re trying to determine intelligence here, and human emotions may not be a valid “barometer” of intelligence.
Alan expected many objections to the idea of ‘machine intelligence’ proposed in his original paper, one of which was the so-called ‘ostrich policy objection’. It is believed that the power of thought transforms man into the soul of all things. Acknowledging that computers can think may challenge this lofty habitat enjoyed only by humans.
Many people believe that it is the human soul that allows people to think, and that if we create machines with this ability, we will usurp the authority of “God.” Alan refuted this by suggesting that people were simply prepared to wait for the spiritually endowed vessel to do the will of ‘God’. Finally, we mention the objection of Lady Lovelace (who is often referred to in literature as the first computer programmer).
Commenting on analytical engines, she said breezily that “this machine alone can’t surprise us”. She reiterated the belief of many that a computer cannot perform any unprogrammed activity. Alan objected, saying that machines never ceased to amaze him. Proponents of this objection, he maintains, accept that human intelligence can instantly infer all the consequences of a given fact or action. Alan’s original paper referred to these readers in collecting these objections and others.
Controversy and criticism of the Turing Test
Ned Block argued that English text was encoded in ASCII, in other words, as a series of zeros and ones within a computer. Thus, a particular Turing test, a series of questions and answers, can be stored as a very large number. Suppose, for example, that there was an upper limit to the length of the Turing test, in which “Are you afraid of dying? Are you afraid of death? The first three characters beginning with “are stored as binary digits, as shown in Figure 1.5.
Figure 1.5 stores the opening character of the Turing test using ASCII code
Assuming that a typical Turing test lasts an hour, during which the tester asks about 50 questions and gets 50 answers, the binary number corresponding to the test should be very long. Now, suppose there was a large database of all the Turing tests, which contained 50 or fewer questions with reasonable answers.
The computer can then pass the test by looking up tables. Of course, a computer system capable of processing such a large amount of data does not yet exist. But if the computer passes the Turing test, Block asks, “Do you think such a machine is intelligent? Do you feel comfortable?” In other words, Block’s critique was that the Turing test could be passed by mechanical table-lookup rather than intelligence.
John Searle’s criticism of the Turing test was more fundamental. Imagine the questioner asking the question as one would expect — but this time in Chinese. The man in the other room did not speak Chinese, but had a detailed rulebook. Although the Chinese questions are presented in scrawled handwriting, the people in the room refer to the rule book, process the Chinese characters according to the rules, and write the answers in Chinese, as shown in Figure 1.6.
Figure 1.6 Argument in Chinese room
The inquirer gets answers to grammatically correct and semantically sound questions. Does that mean the people in the room are fluent in Chinese? If your answer is “no”, does the combination of man and Chinese rulebook know Chinese? Again, the answer is “no” — the people in the room are not learning or understanding Chinese, but merely processing symbols. Similarly, computers run programs that receive, process, and answer with symbols without having to learn or understand what the symbols themselves mean.
Sell also asks us to imagine a gym where people pass notes to each other, if not a single person holds a rulebook. When a person receives such a note, the rule book determines whether the person should generate an output or simply pass the information to another person in the gym, as shown in Figure 1.7.
Figure 1.7 Variation of argument in Chinese room
Where does the knowledge of Chinese exist now? Does it belong to everyone, or does it belong to the stadium?
Consider one last example. Map the brain of a person who does know Chinese, as shown in Figure 1.8. This person can receive questions in Chinese and accurately explain and answer them in Chinese.
Figure 1.8 The Chinese speaker receives and answers questions in Chinese
Similarly, where does knowledge of Chinese exist? Is it in a single neuron, or is it in a collection of neurons? (It has to exist somewhere!) The key criticism of the Turing test by Block and Searle is that it only looks from the outside and does not provide insight into the internal state of an entity. That is, we should not expect to learn anything new about intelligence by treating intelligent agents (people or machines) as black boxes. However, this is not always true. In the 19th century, the physicist Ernest Rutherford correctly inferred the internal state of matter — which consists mostly of empty space — by bombarding gold foil with alpha particles.
He predicted that the high-energy particles would either pass through the gold foil or be slightly deflected. The results were consistent with his theory of atomic orbitals: atoms consist of a dense core surrounded by orbital electrons. This is our current model of the atom, and many of you who have taken chemistry in high school are very familiar with it. Rutherford succeeded in understanding the internal state of atoms by looking outside.
In short, defining intelligence is hard. It was the difficulty of defining intelligence and determining whether an ‘intelligent agent’ possessed it that led Turing to develop the Turing test. In it, he implicitly argues that any agent that can pass the Turing test must have “brain capacity” to deal with any reasonable intellectual challenge that is comparable to what people generally accept at the human level.
Artificial Intelligence (2nd Edition)
By Stephen Lucci
The Classic American textbook, rated by America Asia as the best textbook since Russell & Norvig’s Artificial Intelligence: A Modern Approach, is more suitable for undergraduates. Known as the encyclopedia of artificial intelligence field.
Based on the theoretical foundation of ARTIFICIAL intelligence, this book presents readers with a comprehensive, novel, colorful and easy-to-understand body of artificial intelligence knowledge. Examples, applications, full-color images, and anecdotes are provided to stimulate reading and learning. Advanced courses in robotics and machine learning, including neural networks, genetic algorithms, natural language processing, planning and complex board games, have also been introduced.
Today’s interactive
What do you think of the future development of artificial intelligence? Why is that? The deadline is 17:00 on August 3rd, leave a message and forward this activity to the circle of friends or book groups of more than 50 people, and 5 readers will be selected by lucky draw as a gift of a paper book (participation in the activity will be direct to the wechat terminal of artificial intelligence, this encyclopedia in the field of artificial intelligence should not be missed).
Click to read the original article to buy the book Artificial Intelligence (2nd Edition)
Read the original