Istorijski razvoj veštačke inteligencije
Veštačka inteligencija je nauka o stvaranju inteligentnih računarskih programa i mašina u pokušaju da imitira ljudske nivoe inteligencije.
Artificial intelligence is the field of science about creating intelligent computer programs and machines in an attempt to mimic human levels of intelligence. From ancient Greek syllogisms, through all mathematical theories of logic to today's models of artificial intelligence, there has been a desire among people to understand thinking and design independent intelligent agents who are able to understand the world around them and make independent decisions. u skladu sa situacijom.
Thus, in the middle of the last century, top scientists began to work more intensively on that idea.
The first among them was the English mathematician Alan Turing, who in 1950 published an article in the magazine Mind, entitled "Computers and Intelligence", and thus laid the foundation stone in the philosophy of artificial intelligence by proposing a test with which we can determine whether a machine is intelligent - if in conversation with it we cannot distinguish it from man, then it is an intelligent machine.
Then, in the mid-fifties of the XX century, several scientists began to explore the possibility of creating an "artificial brain". Research has shown that the brain is a network of interconnected cells, neurons. Neurons transmit information to each other using electrical impulses.
It is believed that the concept of artificial intelligence was set up by McCullough and Pittsin 1943, in a paper that presents a model of artificial neurons based on three sources. A few years later, the first neural computer was created named SNARC.
The formal beginning of artificial intelligence is considered to be the year 1956, when a conference was held in Dartmouth. It was attended by experienced researchers in this field who discussed past achievements and future directions of development. At the suggestion of the American scientist John McCarthy, the creator of the LISPprogramming language, the field was named artificial intelligence.
Intelligent behavioe has begun to be seen as finding the most effective set of actions - steps that lead to problem solving. The solution is reached gradually, step by step, memorizing the previous steps that are used for corrections if the path taken turns out to be a dead end. The step that reduces the difference to the desired goal is chosen the most. The approach has proven to be effective for problems such as finding the shortest path from one point to another or a path through a maze, but that was not enough to solve real-life problems.
John McCarthy then proposed a new way of creating intelligent systems based on logic. McCarthy proposed a system in which heuristics (the science of methods and principles of finding new ones) is an integral part of the system, and programming is reduced to communicating facts-premises to a computer using a suitable formal language. This system was named logical programming, and in the following decades it was developed through the programming language Prolog.
Machine learning was created by Ray Solomonov, an American scientist of Russian origin, presenting a proposal for a machine that draws conclusions based on previous examples with which it was trained. This field aims to develop algorithms that allow computers to improve their own behavior through learning based on empirical data from databases or sensors.
The first algorithm called "Perceptron" was introduced by the American scientist Frank Rosenblatt in 1957. The algorithm enabled the application of neural networks to the classification problem. Based on a set of input signals from the outside world (image, sound, numbers…), the neural network has generated at its output the class to which the inputs belong. During the 1990s, the first efficient neural network-based text and speech recognition systems were developed.
Over time, the difference between top chess players and computers became less and less noticeable, so in May 1997, Deep Blue, a chess computer developed by IBM, defeated the then current world chess champion Gary Kasparov.
DARPA Grand Challenge is a competition organized by the US government, where competitors were offered a cash prize for constructing a usable autonomous vehicle. At the first competition, held in 2004, none of the vehicles managed to cross the planned 240 km long route. However, as early as the following year, as many as five vehicles reached the finish line. The routes were first formed through desert and mountainous areas, while in the following years the competition was successfully held in urban areas as well.
Answering the questions in natural language has become the next big challenge for all fans of artificial intelligence. In February 2011, IBM's Watson system convincingly won one well-known American quiz of the best live competitors. During the competition, Watson had to follow the rules that applied to other participants: he was asked questions in natural language, he was not allowed to use external resources, but he was a local copy of Wikipedia, and he had to build tactics that involved assessment of certainty in the correctness of one's own answer. IBM Watson uses natural language document analysis and statistics to come up with an accurate answer to a question. Today, this feature is widespread, even in our mobile phones.