User:Vidstige~enwiki/History of AI

The history of artificial intelligence (AI) can be said to have started in 1943 when Warren McCulloch and Walter Pitts presents their work on artificial neurons, using knowledge from physiology, logic and maths. It is not until 1956 though, on a 2 month Dartmouth workshop, the field is officially given its name: artificial intelligence. On this workshop, the Logic Theorist, a reasoning program is presented by Allen Newell and Herbert Simon.

The general distrust of computers ability to do anything clever, provided the early researchers with lots of relatively easy problems to impress the public with. More noteworthy successes includes the General Problem Solver (Allen Newell and Herbert Simon) and the Geometry Theorem Solver (Herbert Gelernter). LISP was created 1958 by John McCarthy, the same year as he describes the Advice Taker (Programs with Common Sense), considered as the first complete AI system, although hypotethical.

The concept of microworlds where founded, a domain of limited problems which appear to require intelligence to solve, which of the blocks world is the most famous. Works on neural networks, based on McCulloch and Pitts, was done and Frank Rosenblatt proved the famous perceptron convergence theorem.

After a row of successes, researchers began to try their luck on harder problems, which require more domain knowledge. They soon relized that there previous methods where not enough and that even if the problem was solvable in theory, it might take forever to solve it practically.

Knowledge-based system, or expert systems, came as an answer to the need for more domain knowledge to solve advanced problems. By coding a lot of domain knowledge into the program, the programs got a better idea of which steps that where useful to try when solving the problem, making them spending less them on impossible solutions.

Expert systems are taking into commercial service in the 1980s. At this time a Japanese project to build computers that run prolog programs like machine code was started, and the fear in the U.S. for getting behind Japan made researchers having no trouble finding founding. The 1980 also saw a rise in the interest for neural networks.