The year is 1997. The dim colors of biggies show two men in suits playing chess. One of these men was the biggest name in chess at the time. Kasparov. The other person does From IBM’s artificial intelligence computer Deep Blue An agent that plays incoming commands. This match is making news around the world because it is the first time that an artificial intelligence has challenged a chess master.

A few years have passed and the Internet has just begun to become widespread and invade homes. Everyone, whether they play chess or not, plays chess on the Internet. on Chessmaster He starts the game on the hardest difficulty setting and makes his opponent’s moves in Chessmaster. They also use Chessmaster’s reactions against their human opponents. A maniac named Botman exaggerates this artificial intelligence thing and adds bots to the Counter-Strike game and makes the game playable at home without internet.

The year is 2024. An academic teaching at a university, students to make sure the AI ​​doesn’t do the homework. Trying to find new methods. What happened in the meantime that artificial intelligence, which only makes movements in games or makes simple predictions, began to do homework and write programs on its own?

In fact, artificial intelligence has been on the agenda for a long time.

in 1956 John McCarthyHe came up with a remarkable name for the event where he brought together the most important scientists of his time: Artificial Intelligence. Before the name was invented, computer scientist Arthur Samuel developed a program because he couldn’t find anyone to play checkers with in his spare time from work and invented the first artificial intelligence – albeit without knowing its name.

Although the computers of this period were enormous, they were far behind today’s standards in processing capacity. This added an interesting feature: artificial intelligence had to make predictions like you and me. Can be played simultaneously Because he cannot calculate countless movesmade the move that seemed most ideal at the time.

Over time, artificial intelligence began to improve itself. When their moves caused them to lose a game, they started turning to the next best alternative. This resulted in the method we call reinforced learning. Only with this method artificial intelligence It didn’t act more or better, it just acted faster. If you continuously chop tomatoes, you will learn how to cut tomatoes without crushing them and complete the process without cutting yourself, but this exercise will not be of much use if you are told: “Here, chop the pineapple.” Old artificial intelligence also had such a problem. This artificial intelligence was called symbolic artificial intelligence.

Artificial intelligence became very popular, but could not continue: the “AI winter” has arrived

In the old version of artificial intelligence small problem There were: inventors. While old models were designed, humans were believed to solve every problem with logic, while artificial intelligence was designed at the level of Aristotelian logic.Elephants fly, Mehmet is an elephant, so Mehmet flies too.” It was considered sufficient to think at this level. People Because they are not rational beings at all. artificial intelligence could not meet the needs. Companies promised systems similar to robots in science fiction films, but they were not successful.

As a result of these incompatibility and understanding problems, artificial intelligence research boomed in the 1970s and 1980s. 1987-1993 It almost stopped between now and this period was called “artificial intelligence winter”. At a time when the country, which claims to be the cradle of democracy, appointed as head of state an actor it saw as a “cowboy”, it was natural for people to resent a machine that made decisions based entirely on on logic and logic. ran the codes accordingly.

Then came neural networks

Neural networks have millions of codes heavy and one-time work It came to replace the artificial intelligences that do this. These artificial intelligence models were considered the first artificial intelligence that could ‘really think’. In fact, this system, which has existed throughout the history of artificial intelligence research, first came to prominence in 1986 with the famous paper called backpropagation. This article lays the foundation for why we should click pictures of cats and dogs to convince sites that we are not robots. In its simplest form, artificial intelligence divides images into layers and classifies images or data with layer values. It also allows for more optimized work and training.

After this breakthrough in the neural network, in 2012 AlexNet came. In this groundbreaking paper on deep learning and neural networks, although the concept of AlexNet was put forward by Alex Krizhevsky, the lead author of the study, one of the names who signed the study was also Ilya Sutskever, the estranged founder and former manager of OpenAI. This article laid the foundation for the new era by showing how deep learning can be used with a lot of data and a lot of computing power.

During this period, as artificial intelligence tools that perform more complex calculations, are smarter, and can be used for different purposes emerged, concepts such as machine learning and large language models also began to become popular. Here we also saw primitive examples of artificial intelligence that is more responsive to humans in general and closer to truly understanding people.

Artificial intelligence has truly reached a new level

In the old approach to artificial intelligence, connections are made according to codes on a two-dimensional plane and in a very limited area. without transferring information to different areas transaction was in progress. In the new approach, a lot of information is taken into account at the same time and the skills acquired in different areas are transferred to other media. In other words, artificial intelligence, which draws arrows on paper, now works with a three-dimensional structure and each part is activated like a brain.