From "Perceptron" to first "Macintosh"

@InoSMI.ru
Show original
The technician holds a hand of the android of Rob's Open Source Android © AP Photo, Alastair Grant

Five seasons Artificial intelligence. It totally beats the person in go, assumes control of its car and replaces it at work, and at the same time can increase efficiency of medicine. Its long history originates in 1958 from the huge car which could distinguish the right and left.


1:0. Then 2:0. And 3:0. In March, 2016 in the Seoul hotel "Four Seasons" the final meeting after which did not remain also to a doubt shadow took place: the Korean champion of game conceded in guo Li Xiedol with the score 4:1 to the computer under control of the Alfago program developed by the affiliated Google Inc. "Dipmaynd". For the first time in the history the <16> machine trainings mechanism and "Neural networks" completely surpassed a human brain in this game which as it is considered, is more difficult to feign, than chess. Many experts emphasize that expected such result, only in some years.


For wide audience it became the proof of power of new technologies "deep training" which now is cornerstone of voice assistants, independent cars, Face recognition, machine translation, and also statement of the medical diagnosis is promoted …

to
to
to


by Interest to technologies automatic training, which is shown by the American and Chinese corporations in the field of high technologies (Google Inc., "Amazon", Social network Facebook, "Microsoft", "Canoe", "Tensent"), covers all planet and even more often jumps from scientific headings of newspapers in economic, analytical and social materials. The matter is that Artificial intelligence not only promises big changes in economy, but also causes thoughts of the new destructive weapon, general supervision of citizens, replacement of employees with robots, ethical problems …


But from where technological revolution of Artificial intelligence undertook? In its stories take-off and falling suffice. She relied on achievements of a neurobiology and informatics (as it is possible to guess from the name), and also that is surprising, physics. Its way went through France, United States of America, Japan, Switzerland and the USSR. In this sphere different schools of sciences faced with each other. In one day they won, in another lost. All needed to show patience, persistence and readiness to risk. In total in this history two winters and three spring.

"The car realizing"


everything Began


just fine. "The American army told about an idea of the car which will be able to go, speak, see, write, to breed and realize itself" — wrote American newspaper "The New York Times" on July 8, 1958. This article in one column describes "Perceptron" which was created by the American psychologist Frank Rozenblatt (Frank Rosenblatt) in laboratories of Kornellsky university. This car worth 2 million dollars those times was two-three refrigerators in size and is braided by a set of wires. During demonstration before the American press "Perceptron" defined, whether there is a square drawn on a leaf on the right or at the left. The scientist promised that at an investment of 100 thousand more dollars its car will be able already to read and write in a year. In practice for this purpose more than 30 years …

were required to


Anyway, the inspiration source which remained invariable up to "Alfago" and his "relatives" was the main thing in this project. The psychologist Frank Rozenblatt was engaged more than ten years in concepts of cybernetics and Artificial intelligence. By the way, he developed "Perceptron" by means of works of two other North American psychologists: Warren Mak-Kallok Warren Mak-Kallok - Kalloka (Warren McCulloch) and Khebb Donald (Donald Hebb). The first published in 1943 joint article with Walter Pitts (Walter Pitts) with the offer on creation of "artificial" neurons which have to make a start from natural and possess mathematical properties. The second presented in 1949 of the rule who have to allow artificial neurons to study a trial and error method as it does a brain.


the Bridge between biology and mathematics became a courageous initiative. Calculating unit (neuron) can be active (1) or inactive (0) depending on incentives from other artificial educations with which it is connected, forming a difficult and dynamic network. If is more exact, each neuron receives a certain set of symbols and compares it to a certain threshold. If the threshold is overcome, value is equal 1, otherwise — 0. Authors showed that their related system can perform logical operations of the "and" and "or" type … and thus carry out any calculation. In the theory.


Such innovative approach to calculations caused the first quarrel in our history. Two concepts met in irreconcilable opposition which proceeds and to this day. On the one hand there are supporters of Neural networks, and with another — advocates of "classical" computers. The last rely on three principles: calculations have mainly consecutive character, memory and calculations are provided with accurately certain components, any intermediate value has to be equal 0 or 1. At the first all in a different way: the network provides both memory, and the calculations, the centralized controls is not present, and intermediate values are allowed.


"Perceptron" too possesses ability to be trained to distinguish, for example, drawing or to classify signals. Approximately as the shooter corrects a sight. If the bullet leaves to the right, it displaces a trunk to the left. At the level of artificial neurons it means weakening of who pull to the right, to please what pull on the left, and allows to hit the mark. It is necessary to create only this ball of neurons and to find a way to connect them.


Anyway, the enthusiasm significantly died away in 1968 with a publication of the book "Perceptrons" Papert Seymur (Seymour Papert) and Marvin Li Minski (Marvin Minsky). In it they showed that the structure of perceptrons allows to solve only the simplest problems. It was the first winter Artificial intelligence, whose first spring, it is worth recognizing, did not bear special fruits. And the wind blew not ABBA from where: Marvin Li Minski stood at the origins of emergence of the concept "Artificial intelligence" in 1955.


Collision of Artificial intelligence and UI


on August 31 that year they with the colleague John Makkarti (John McCarthy) sent to ten experts the invitation to participate next summer in two-month work on the concept for the first time used at that moment Artificial intelligence in Dartmutsky college. There were present including Warren Mak-Kallok and Shennon Klod (Claude Shannon), the father of the theory of informatics and theoretical base of telecommunications. He brought the Minsk and John Makkarti into laboratories of companies "Bell", from where subsequently there were transistors and lasers. Besides, they became one of the centers of revival of Neural networks in the 1980th years.


In parallel with it created two new currents which battlefield there was Stanford university. On the one hand the abbreviation of Artificial intelligence, "Artificial intelligence" flaunted, in other than Neural networks understanding which was defended by John Makkarti (it left Massachusetts Institute of Technology and created the laboratory in Stanford). On the other hand — UI, "the strengthened intelligence", Douglas Engelbart's (Douglas Engelbart) reflecting new approach. That was employed in 1957 by Stenfordsky research institute (the independent establishment created in 1946 of higher education institution which interacted with the private sector).


Behind Douglas Engelbart's shoulders was a difficult way. He was a technician and was engaged in a radar in the years of World War II, but then resumed training and defended the dissertation. Before arrival to Stanford it even created the enterprise, however it existed only two years. On a new place it began to realize the vision of strengthening of human abilities. He said that accurately imagines as "colleagues sit in different rooms at similar workplaces which are connected with one information system, and can closely interact and exchange data" — the sociologist Bardini Thierry tells (Thierry Bardini).


It is vision received practical reflection in December, 1968, in ten years after emergence of "Perceptron", during oNLine System system demonstration with a text editor on the screen, hyperlinks on documents, schedules and a mouse. Douglas Engelbart was a seer, but, probably, looked forward too far really to declare itself.

the January, 1984, first "Macintosh"


John Makkarti in turn called


this system excessively "dictatorial" as that imposed special approach to text structurization. This courageous scientist whom, as well as Engelbarta, the American army financed, submitted own, symbolical concept Artificial intelligence. In it it relied on LISP developed by it one of the first programming languages. The idea was in imitating thought process by a logical chain of rules and symbols and to create thereby thought or at least cognitive function. There is nothing the general with networks of independent neurons which can be trained, but is not able to explain the choice. Except for spilling punch a robo-hand which made laugh all, having overturned the glasses, new approach was quite successful in respect of that called long time "expert systems". Chains of rules allowed cars to analyze data in the most different spheres, whether it be finance, medicine, industrial production, the translation.


B 1970 one colleague Minsk made such statement to the Layf magazine: "In eight years at us the car with intelligence will be of the average person at the level. That is the car which will be able to read Shakespeare to change oil in the car, to joke, fight".


the Victory of symbolical approach


Apparently, Artificial intelligence does not love prophecies. In 1973 to England left the report which cooled hotheads: "The majority of research associates who work over Artificial intelligence and in adjacent areas, recognize that are disappointed reached for the last 25 years. (…) In one camp the discoveries made to the present moment did not yield the promised results".

the Followed years confirmed with


this diagnosis. In the 1980th years being engaged Artificial intelligence the enterprises went bankrupt or replaced field of activity. John Makkarti demolished the building of laboratory in 1986.

Douglas Engelbart won against


. In January, 1984 Apple Inc. released first "Macintosh", having realized in practice the most part of ideas of the engineer.


Thus, got a victory not to Artificial intelligence of which the Minsk and John Makkarti, and to Engelbart's strengthened intelligence dreamed. All this caused development effective personal computers. And Artificial intelligence reached a deadlock. Symbolism was stronger than Neural networks. Nevertheless on it our history does not come to an end, and they still will declare themselves.

Full article is available at its source

Go to full article