For many millennia, a person has tried to determine how he thinks, what processes are going on in his head. So in the field of artificial intelligence (AI), scientists have to solve an even more difficult task. Indeed, in this area, specialists have to not only understand the essence of intelligence, but also create intellectual entities.
First of all, it should be noted that artificial intelligence is a fairly young science. The first experiments in this area appeared shortly after the end of the Second World War, and the term "artificial intelligence" appeared a little later - in 1956. At the same time, if it is quite difficult to make a great discovery in other fields of science, then this field of science opens up great prospects for the manifestation of talent.
At the present time, the problem of artificial intelligence includes a large list of various scientific areas, including such general concepts as perception and learning, and special tasks, in particular, theorem proving, playing chess, and diagnosing diseases.
In this area, the analysis and systematization of intellectual tasks is carried out, thus, artificial intelligence concerns all spheres of human intellectual activity, and therefore it can be considered a universal field of science.
From all of the above, we can conclude that the field of scientific intelligence is a very interesting area of science. Interestingly, there is no single definition of AI. In various scientific works devoted to him, there are different interpretations of this phenomenon. They can cover not only thought processes, but also formulations about the behavior of an individual.
If you carefully study the history of the development of artificial intelligence, you can see that research was carried out in several directions. And this suggests the conclusion that there were certain controversial situations between those scientists who were engaged in research of human abilities, and those who were engaged in the problems of rationality.
A scientific approach that focuses on the study of a person should be based on the advancement of a large number of hypotheses, as well as experimental proof of them. At the same time, the approach focused on the study of the concept of rationality is a kind of combination of technology and mathematics.
In order to test whether a computer is capable of performing actions like a human, an approach was developed that relied heavily on the Turing test. It got its name from its creator, Alan Turing. The test is used as a satisfactory functional definition of intelligence. The English mathematician who laid the foundations of computer technology, in 1950, published a scientific article called "Computing Machines and the Mind", which proposed a test that could determine the intellectual level and nature of the intelligence of a computer.
The author of the test came to the conclusion that there is no point in developing a large list of requirements in order to create artificial intelligence, which, among other things, can turn out to be very contradictory, therefore, he proposed a test that was based on the fact that in the end it would be impossible to distinguish the behavior of an object endowed with artificial intelligence from the behavior of human beings. Thus, a computer can successfully pass the test if the human experimenter, who asked him questions in writing, cannot determine from whom the answers were actually received - from a person or from a certain device.
At the same time, the author derived a formula that determined the boundary when artificial intelligence could reach the natural level. According to Turing's findings, if a computer can trick a person into answering 30 percent of the questions, then it can be assumed that he has artificial intelligence.
At the same time, in order for the computer to be able to answer the questions posed, it must perform a large amount of actions. So, in particular, it must have such capabilities as means of information processing in a natural language, which would allow quite successfully communicate with the device in one of the languages existing in the world. In addition, it should be equipped with means of knowledge representation, with the help of which the device will be able to write new information into memory. There should also be a means of automatically generating conclusions, which would provide an opportunity to use the available information to search for answers to the questions posed and formulate new conclusions. Machine learning tools are designed to provide the computer with the ability to adapt to new circumstances, and in addition, detect the signs of a standard situation.
The Turing test deliberately excludes the possibility of direct physical interaction between the person conducting the experiment and the computer, because the process of creating artificial intelligence does not require physical imitation of a person. In this case, in the case of using the full version of the test, the experimenter can use the video signal in order to test the computer's ability to perceive.
Therefore, when passing the full Turing test for the above tools, it is necessary to have machine vision for the perception of an object, as well as robotics tools to be able to manipulate objects and move them.
All of this ultimately forms the basis of artificial intelligence, and the Turing test has not lost its importance even after half a century. At the same time, it should be noted that scientists who study and create artificial intelligence almost never solve problems aimed at passing this test, believing that it is much more important to study in detail the principles that underlie intelligence than to create a copy of one from carriers of natural intelligence.
At the same time, the Turing test was recognized as the standard, but until recently, scientists have not been able to create a program that would successfully overcome the test. Thus, scientists could easily determine whether they were talking to a computer or a person.
However, a few months ago, information appeared in the media that scientists, for the first time in fifty years, managed to come close to creating an artificial intelligence that was able to think like a person. As it turned out, the authors of the program were a Russian group of scientists.
At the end of June, the UK hosted a worldwide cybernetic intelligence science competition sponsored by the University of Reading. The competition was held at the main encryption center in Blatchley Park. Russian scientists presented a program called "Eugene". Besides her, 4 more programs took part in testing. The Russian development was recognized as the winner, answering 29.2 percent of the questions asked in the same way as a person. Thus, the program lacked only 0.8 percent in order for the long-awaited event to come true - the emergence of artificial intelligence.
American scientists also keep up with the Russians. So, they managed to create software bots that were developed specifically for a computer game. They passed the modified Turing test without any problems and quite confidently. It should be noted that this was done with much more success than the people who tested it with bots. And from this, we can draw certain conclusions that artificial intelligence has managed to reach the level when the automatic system is no longer able to determine where the person is responding and where is the computer.
Of course, it is too early to argue that overcoming such a specific version of the Turing test, which is a game shooter, is an indicator of the creation of artificial intelligence by a person. At the same time, this gives every right to say that artificial intelligence is gradually approaching human, as well as the fact that game bots have already reached the level of development at which they can quite successfully deceive automatic systems designed to determine human behavior.
Scientists from the University of Texas Jacob Schrum, Risto Miikkulainen and Igor Karpov became the creators of game bots. They managed to create artificial intelligence that can play the game on a human level. A huge virtual platform was created, where many bots and real people fought. Most played anonymously. More than half of the game bots were identified by the judges as humans. At the same time, they considered some people to be bots. Thus, the conclusion suggests itself that computer characters already in games behave like people.
The experiment was carried out as part of a competition called BotPrize, which started in America back in 2008. Scientists and developers, whose computer programs will be able to deceive people, can become its participants. Posing as quite real players. But the first successes in this area were achieved only in 2010.
The winners will receive an award of £ 4,500 and will continue to work on their programs. And there is still something to strive for, because in order to recognize the creation of artificial intelligence, the program must convince everyone that it is a person, during the conversation. And this requires deep knowledge of the work of the human brain and the principles of speech formation. Currently, no one has succeeded in passing the Turing test in its original version. But it is quite possible to assume that this may happen in the near future …
