Minsky was born in New York City on August 9, 1927. His father was an ophthalmologist and an artist at the same time. His mother is an active Zionist (what a funny coincidence: McCarthy's mother is a Jew from Lithuania, an active feminist). Elementary and middle school, Minsky went to a private school and showed interest in electronics and chemistry. The scope of his activities is basically the intellectual class. After graduating from high school in 1945, Minsky enlisted in the army and received training with about 120 other recruits at the Great Lakes Naval Training Center (Great Lakes Navaltraining Center) north of Chicago. According to Minsky himself later, this was the first and last time he had mixed up with non-academic personnel. After retiring from the army, he entered Harvard University in 1946 to major in physics, but his elective courses are quite extensive, from electrical engineering, mathematics, to genetics and other subjects involving multiple disciplines. For a while, he also participated in research projects in the Department of Psychology. . Some theories and theories about the origin of the mind that were popular at that time made it difficult for him to accept. For example, the neo-behavioral psychologist Skinner (Burrhus Frederic Skinner, 1904-1990) put forward theories based on the facts of animal behaviors, combining human learning with animal learning. Equivalently, Minsky disagreed, and inspired his determination to clarify this difficult problem. Later, he gave up physics to reform mathematics, and graduated in 1950, and then entered the graduate school of Princeton University for further studies.
Proposed basic theories about thinking
Before World War II, Turing was here to study whether machines can think about this issue. , Minsky also began to study the same problem here. In 1951, he put forward some basic theories on how thoughts sprouted and formed, and built a learning machine called Snare. Snare is the world's first neural network simulator. Its purpose is to learn how to walk through the maze. Its composition includes 40 "agents" (agent, domestic sources also translate it as "subject" and "agent") And a system that rewards success. Agent-based computing and distributed intelligence are currently a hot spot in artificial intelligence research. Minsky may be one of the earliest scholars who proposed the concept of agent. Although Snare is still relatively rough and not flexible enough, after all, it is one of the earliest attempts in artificial intelligence research. On the basis of Snare, Minsky comprehensively used his multidisciplinary knowledge to solve the problem of enabling the machine to predict the results of its current behavior based on the knowledge of the past behavior, and used the "Neural Nets and the Brain Model Problem" (NeuralNetsandtheBrainModelProblem). ) Completed his doctoral dissertation on the subject and obtained his doctorate in 1954.
Organizing and launching an artificial intelligence conference
Minsky is one of the founders of artificial intelligence and the organizer of the Dartmouth Conference.
After completing his studies, he stayed at school for 3 years, during which he initiated and organized the "Dartmouth Conference" which became the starting point of artificial intelligence together with McCarthy, Shannon and others. At the conference, Minsky's Snare, McCarthy's α-β search method, and Simon and Newell's "logic theorist" (10sicTheorist) were the three highlights of the conference. In 1958, Minsky transferred from Harvard to MIT, and McCarthy also came to MIT from Dartmouth to join him, where they jointly created the world's first artificial intelligence laboratory.
Contributions to artificial intelligence
Minsky’s contributions to artificial intelligence are multifaceted. In 1975 he pioneered frame theory (frametheory). The core of frame theory is to express knowledge in the form of frame. The top level of the framework is fixed, representing fixed concepts, objects, or events. The lower layer is composed of a number of slots, which can be filled with specific values to describe the characteristics of specific things. Each slot can have several facets, and additional descriptions are made on the slot, such as the value range and evaluation method of the slot. In this way, the framework can contain a variety of information, such as information describing things, information about how to use the framework, expectations for what will happen next, what to do if it doesn't happen, and so on. By using multiple related frames to form a frame system, knowledge can be expressed completely and accurately.
Develop the world’s earliest robot
Minsky also combined artificial intelligence technology and robotics technology to develop the world’s first robot RobotC that can simulate human activities, making the robot Technology has leapt to a new level. Another big move by Minsky was to create the famous "Thinking Machines, Inc." (Thinking Machines, Inc.) to develop intelligent computers. In the mid-1980s, the thinking machine company began to launch the famous "ConnectionMachine" (ConnectionMachine) series CM-1, CM-2 and CM-5, connecting a large number of simple storage-processing units into a multi-dimensional structure, forming a macroscopic structure Large-capacity smart storage, and then through conventional computers to perform control, I/O and user interface functions, can be effectively used for smart information processing. CM-1 is composed of 4 quadrants, each quadrant contains up to 16,384 one-bit processors, and all processors are divided into 4096 groups, forming a 12-dimensional hypercube structure between groups, and its integrated peak speed reaches 60 billion times per second . CM-5 has more nodes and stronger functions.
The advocate of “virtual reality”
Minsky is also the advocate of “virtual reality” (virtual reality), although the term and concept of VR only appeared in the 1990s. It became clear. As early as the 1960s, Minsky coined a term called telepresence, which literally means "remote existence" or "remote intervention." Minsky's so-called telepresence is such a device or environment. People are allowed to experience certain events without actually intervening in such events, such as feeling like they are flying an airplane, participating in a battle on the battlefield, swimming underwater, etc. In fact, none of these things happened. Minsky has proposed some schemes to realize telepresence by using devices such as miniature cameras and motion sensors. Isn’t Minsky’s telepresence the same as the virtual reality currently being studied?
As one of the advocates of artificial intelligence, Minsky firmly believes that human thinking process can be simulated by machines, and machines can also be intelligent. One of his popular sayings is: "The brain is nothing more than a machine made of meat" (thebrainhappenstobeameatmachine). As we mentioned earlier, this view of his has been tit-for-tatly criticized by Wilkes, another Turing Award winner, and it needs further scientific research and practice to judge whether it is right or wrong.
For example, there is a car frame as follows:
super-class: transportation< /p>
sub-class: car, van, jeep
Number of wheels:
default: 4 p>
value-class: floating point number
where super-class and sub-class respectively represent the parent class and sub-class of the object, "number of wheels" and "body length" are two slots, reflecting the structural attributes of the car, respectively It is composed of several sides. In the example, there is value, which means the value of the attribute, value-class (or type) means the type of the attribute value, default means the default attribute value, and so on.
Minsky originally proposed the framework as the basis of visual perception, natural language dialogue and other complex behaviors, but once it was proposed, it was because it was both hierarchical and modular. The artificial intelligence community has aroused great repercussions and has become a universal knowledge representation method which is widely accepted and applied. Not only that, some of its basic concepts and structures have also been used by the object-oriented technologies and methods that emerged later. In addition, Minsky’s framework theory has also become the basis of some popular expert system development tools and artificial intelligence languages. For example, the famous KRL (Knowledge Representation Language) was founded in 1979 by D.G. Boborow, who also One of the main developers of Interlisp who won the ACM Software System Award in 1992) is based on the design and implementation of the framework structure. Two other special contributions of frame theory are: 1. It first proposed the concept of "default" and became an important research object for common sense knowledge representation. 2. Develop a "script" representation method from the framework, which can describe events and time sequence, and become one of the foundations of CBR (case-based reasoning).
"Computation: Finite and Infinite Machines, Prentice-Hall, 1967) "Semantic Information Processing" (Semantic Information Processing, MITPr., 1968)< /p>
"Perceptrons, MITPr., 1969; expanded edition, 1988)
"A Framework for Representing Knowledge, McGraw-Hill, 1975)
"The Society of Mind" (The Society of Mind, Simon & Schuster, 1986)
"Robotics" (Robotics, AnchorPr./Doubleday, 1985)
"The Emotion Machine: Commonsense Thinking ,ArtificialIntelligence,andtheFutureoftheHumanMind, 2006)
Minsky is a member of the American Academy of Sciences and the American Academy of Engineering. He served as the third chairman of AAAI (1981-1982). In addition to receiving the Turing Award, he also received the Killian Award from MIT in 1989. In 1990, he won the "Japan Award" established by the Japanese government.
When Minsky received the Turing Award, he gave a speech entitled "The Form and Content of Computer Science", which was published in JournalofACM, April 1970, 197 —215 pages. See also (Turing Award Lectures in the first 20 years) (ACMTuringAwardLecturesTheFirst20Years: 1966-1985, ACMpr.), 219-242 pages. In his speech, Minsky discussed the so-called "form-content confusion" in computing theory, programming language, and education, and expressed his views. Much of the space is criticizing the "New Mathematics" at the time. He advocated that mathematics education for children should not only emphasize the form, but also the content instead of ignoring the content.