History of Artificial Intelligence, part 1. Painting without an artist.
I became interested in this area spontaneously. After that, more and more ideas on this topic began to appear, so I had to streamline it all. I decided to start with the history of the development of artificial intelligence. The article turned out pretty big, so I divided it into two parts. Here is the first part of it, dedicated to cybernetics.
Artificial Intelligence is one of the most interesting ideas of man. Many dissertations and even more works in the science fiction genre have been written about him. You can endlessly argue about him and so do not come to a common opinion. He is everywhere and nowhere. And all because there is no consensus about what is or should be Artificial Intelligence.
At the mention of these two words, someone will remember the same film by Steven Spielberg, someone of Azimov’s book about positron robots. Others will say: "Drop to the ground and open your eyes, because Artificial Intelligence is already among us." However, some point to computer games, in which virtual opponents try to imitate their actions with real ones, they are even called also AI. Perhaps the older generation will recall KAISSA and similar programs that are already able to beat even the world chess champion. For someone all sorts of software wizards and search engines are AI.
')
This question is one of those for which it is difficult to give a definite answer. Moreover, the more you begin to dig into it, the more you meet discrepancies, contradictions and even more questions. The definition seems to be as it should be, but it is only hidden under a mountain of different sciences, from above there is a decent layer of fiction, followed by a small layer of philosophy, delusions and subjective romanticism, well, and at the very top, pseudo science and screaming advertising.
As you can see, the truth is not so easy to reach. Different people have different opinions on this matter, and therefore it becomes difficult to formulate something in common. To understand this question, let's see what history will say about this. It turns out that the very name Artificial Intelligence was invented not by science fiction writers, but by scientists. But first things first.
It was back in 1956 (by this time there was already a series of books by Azimov about robots). In the USA, a seminar was held at Stanford University, at which they proposed the term Artificial Intelligence. The participants in that seminar vaguely imagined what it was and why it was needed. The topic of the seminar was logical tasks and ways to automate their solutions. A start was made.
After 25 years, Barr and Feigenbaum proposed the following definition, which no one has challenged for almost two decades.
Artificial Intelligence (AI) is an area of ​​computer science that develops intelligent computer systems, i.e. systems that have the capabilities that we traditionally associate with the human mind - understanding the language, learning, the ability to reason, solve problems, etc.
Here they are the basic properties of AI - language understanding, learning and the ability to think. This is the goal to which its future creators should strive. The definition is very laconic, just in the style of mathematicians, but for now let's dwell on it.
As it usually happens, the “future creators” were divided into two camps from the very beginning: cybernetics and neurocybernetics.
Cybernetics immediately abandoned the idea of ​​a complete copy of human thought processes. It is enough for them that the device and the person give the same answers (output data) to the same questions (input data). And how it happens, what mechanisms and algorithms are inside the car, it does not matter. Therefore, it is called cybernetics "black box".
As soon as computers became widespread, cybernetics mastered them and immediately set to work. They started with small games. The toys were like tic-tac-toe, Scrabble (Scrabble), where the computer's run was calculated by a brute force method. When finding the answer, the search was stopped, and if it was not expected ... It turned out that a combinatorial explosion or speaking Russian, the CPU is 100%.
Despite the fact that some tasks were solved by this method, I still will not turn the language to call it Artificial Intelligence, because a person doesn’t go through all sorts of possible actions in his head before doing something.
The developers paid attention to this and introduced a special technology into the search algorithm - heuristics. Its essence is to limit the search space to special conditions. For each situation they are different.
For example, a computer is trying to find a word that must contain the letter “b” (Scrabble). If he tries to go through absolutely all the options from the available letters, it will take a lot of time and will be ineffective. After all, having met a combination of the letters "brp", it makes no sense to select the next fourth letter, since there are no words that would begin with "brp". Therefore, for such a game, the task of heuristics is to find impossible combinations of letters and stop searching when it encounters such a combination.
Heuristic search is used now, in particular, in some anti-virus scanners.
Other developments that cybernetics did at about the same time are proof of theorems. But, if the computer coped with mathematical tasks without any problems, they all easily yielded to software formalization, then more everyday problems have already arisen. We note that such systems have nothing more to do with an ordinary person than the previous ones, since most of us do not have outstanding abilities in logical reasoning.
Time passed, technology evolved. And so, in the first half of the 70s, more interesting systems began to appear, capable of recognizing and communicating on natural language. One of the first in this area is SHRLDU, created by Winograd in 1972.
She understood a decent amount of English words, limited by the scope - the world, created from the details of the children's designer. All vocabulary, all questions that could get a normal answer, were associated only with this world. Other systems similar to it were also limited to some field of application. About self-study out of the question.
Later, more advanced systems appeared that could support almost all topics. But this can only be argued with great exaggeration, since the ownership of topics was rather superficial. If you ask a question that requires more in-depth knowledge or a difficultly worded question, the program refused to digest it and began to repeat after you, constantly asking again. Its main drawback is the lack of full-fledged self-study. So systems of this type still leave much to be desired.
Another of the more recent developments is knowledge representation systems. They were part of expert systems. These systems are created in order to act as an adviser. They can analyze the situation, find a solution to the problem, carry out diagnostics and everything else in the same vein.
But here, not everything is so smooth. For each such goal you need to create your own expert system. In addition, knowledge of such a system always covers only one narrow field of application. And finally, the organization of knowledge in each system is different, since it directly depends on the goals and content. So, creating a system that could be an expert in various fields is out of the question.
Well, these were just some of the developments in cybernetic. I deliberately did not write here about the development of robots, since the achievement in this area is already widely known. Each of these developments has achieved great success. If it is speech recognition, then it is already at a high level, the thinking of expert systems is also not lagging behind, for good reason so much money is invested in it. Training is now also implemented, although not so well - again, it is strongly limited to the scope. I will return to training when we talk about neural networks.
As you can see, we did not manage to find an AI candidate. Now we are talking more about “some characteristics of Artificial Intelligence” than about the AI ​​itself. All its components have already been developed. It would seem that prevents us from taking and combining them into one interconnected system? For example, take a neurocomputer, lay in it a simplified model of our world (an expert system) and empower with the ability to understand natural speech?
We can do that today. But for some reason it seems to me that we will not get exactly what we would like. Now ask yourself this question - what do you expect from AI? Do your expectations coincide with the definition of Barr and Feigenbaum, who is already 27 years old? In the meantime, you are thinking about it, I will tell you a little about those very neural networks ...
Most of the information on developments is taken from the book of Peter Jackson - Introduction to expert systems.
Table of contents:
History of Artificial Intelligence, part 1. Painting without an artist.
The history of Artificial Intelligence, part 2. Neural network AI - inevitable or impossible?
Making Artificial Intelligence
Artificial Intelligence is one of the most interesting ideas of man. Many dissertations and even more works in the science fiction genre have been written about him. You can endlessly argue about him and so do not come to a common opinion. He is everywhere and nowhere. And all because there is no consensus about what is or should be Artificial Intelligence.
At the mention of these two words, someone will remember the same film by Steven Spielberg, someone of Azimov’s book about positron robots. Others will say: "Drop to the ground and open your eyes, because Artificial Intelligence is already among us." However, some point to computer games, in which virtual opponents try to imitate their actions with real ones, they are even called also AI. Perhaps the older generation will recall KAISSA and similar programs that are already able to beat even the world chess champion. For someone all sorts of software wizards and search engines are AI.
')
This question is one of those for which it is difficult to give a definite answer. Moreover, the more you begin to dig into it, the more you meet discrepancies, contradictions and even more questions. The definition seems to be as it should be, but it is only hidden under a mountain of different sciences, from above there is a decent layer of fiction, followed by a small layer of philosophy, delusions and subjective romanticism, well, and at the very top, pseudo science and screaming advertising.
As you can see, the truth is not so easy to reach. Different people have different opinions on this matter, and therefore it becomes difficult to formulate something in common. To understand this question, let's see what history will say about this. It turns out that the very name Artificial Intelligence was invented not by science fiction writers, but by scientists. But first things first.
It was back in 1956 (by this time there was already a series of books by Azimov about robots). In the USA, a seminar was held at Stanford University, at which they proposed the term Artificial Intelligence. The participants in that seminar vaguely imagined what it was and why it was needed. The topic of the seminar was logical tasks and ways to automate their solutions. A start was made.
After 25 years, Barr and Feigenbaum proposed the following definition, which no one has challenged for almost two decades.
Artificial Intelligence (AI) is an area of ​​computer science that develops intelligent computer systems, i.e. systems that have the capabilities that we traditionally associate with the human mind - understanding the language, learning, the ability to reason, solve problems, etc.
Here they are the basic properties of AI - language understanding, learning and the ability to think. This is the goal to which its future creators should strive. The definition is very laconic, just in the style of mathematicians, but for now let's dwell on it.
As it usually happens, the “future creators” were divided into two camps from the very beginning: cybernetics and neurocybernetics.
Cybernetics immediately abandoned the idea of ​​a complete copy of human thought processes. It is enough for them that the device and the person give the same answers (output data) to the same questions (input data). And how it happens, what mechanisms and algorithms are inside the car, it does not matter. Therefore, it is called cybernetics "black box".
As soon as computers became widespread, cybernetics mastered them and immediately set to work. They started with small games. The toys were like tic-tac-toe, Scrabble (Scrabble), where the computer's run was calculated by a brute force method. When finding the answer, the search was stopped, and if it was not expected ... It turned out that a combinatorial explosion or speaking Russian, the CPU is 100%.
Despite the fact that some tasks were solved by this method, I still will not turn the language to call it Artificial Intelligence, because a person doesn’t go through all sorts of possible actions in his head before doing something.
The developers paid attention to this and introduced a special technology into the search algorithm - heuristics. Its essence is to limit the search space to special conditions. For each situation they are different.
For example, a computer is trying to find a word that must contain the letter “b” (Scrabble). If he tries to go through absolutely all the options from the available letters, it will take a lot of time and will be ineffective. After all, having met a combination of the letters "brp", it makes no sense to select the next fourth letter, since there are no words that would begin with "brp". Therefore, for such a game, the task of heuristics is to find impossible combinations of letters and stop searching when it encounters such a combination.
Heuristic search is used now, in particular, in some anti-virus scanners.
Other developments that cybernetics did at about the same time are proof of theorems. But, if the computer coped with mathematical tasks without any problems, they all easily yielded to software formalization, then more everyday problems have already arisen. We note that such systems have nothing more to do with an ordinary person than the previous ones, since most of us do not have outstanding abilities in logical reasoning.
Time passed, technology evolved. And so, in the first half of the 70s, more interesting systems began to appear, capable of recognizing and communicating on natural language. One of the first in this area is SHRLDU, created by Winograd in 1972.
She understood a decent amount of English words, limited by the scope - the world, created from the details of the children's designer. All vocabulary, all questions that could get a normal answer, were associated only with this world. Other systems similar to it were also limited to some field of application. About self-study out of the question.
Later, more advanced systems appeared that could support almost all topics. But this can only be argued with great exaggeration, since the ownership of topics was rather superficial. If you ask a question that requires more in-depth knowledge or a difficultly worded question, the program refused to digest it and began to repeat after you, constantly asking again. Its main drawback is the lack of full-fledged self-study. So systems of this type still leave much to be desired.
Another of the more recent developments is knowledge representation systems. They were part of expert systems. These systems are created in order to act as an adviser. They can analyze the situation, find a solution to the problem, carry out diagnostics and everything else in the same vein.
But here, not everything is so smooth. For each such goal you need to create your own expert system. In addition, knowledge of such a system always covers only one narrow field of application. And finally, the organization of knowledge in each system is different, since it directly depends on the goals and content. So, creating a system that could be an expert in various fields is out of the question.
Well, these were just some of the developments in cybernetic. I deliberately did not write here about the development of robots, since the achievement in this area is already widely known. Each of these developments has achieved great success. If it is speech recognition, then it is already at a high level, the thinking of expert systems is also not lagging behind, for good reason so much money is invested in it. Training is now also implemented, although not so well - again, it is strongly limited to the scope. I will return to training when we talk about neural networks.
As you can see, we did not manage to find an AI candidate. Now we are talking more about “some characteristics of Artificial Intelligence” than about the AI ​​itself. All its components have already been developed. It would seem that prevents us from taking and combining them into one interconnected system? For example, take a neurocomputer, lay in it a simplified model of our world (an expert system) and empower with the ability to understand natural speech?
We can do that today. But for some reason it seems to me that we will not get exactly what we would like. Now ask yourself this question - what do you expect from AI? Do your expectations coincide with the definition of Barr and Feigenbaum, who is already 27 years old? In the meantime, you are thinking about it, I will tell you a little about those very neural networks ...
Most of the information on developments is taken from the book of Peter Jackson - Introduction to expert systems.
Table of contents:
History of Artificial Intelligence, part 1. Painting without an artist.
The history of Artificial Intelligence, part 2. Neural network AI - inevitable or impossible?
Making Artificial Intelligence
Source: https://habr.com/ru/post/21863/
All Articles