History of AI

AI Robot

We have described the outline and definition of Artificial Intelligence (AI) in the following knowledge.

AI is being researched and progressed every day with the aim of “It learn, guess and judge, just like human beings” .
This knowledge describes the history of AI.


Sponsored Links

History of AI

The beginning of AI

As the beginning of AI, “Logic Theorist” is called the world’s first AI program. In addition, the term Artificial Intelligence was used extensively for the first time at the “Dartmouth Conference,” where “Logic Theorist” was also announced.

  • ”Logic Theorist”
    • It was developed by “Allen Newell”, “Herbert Alexander Simon” and “John Clifford Shaw” from 1955 to 1956.
    • It is the world’s first program that simulates intentionally designed and developed human logical reasoning.
    • It proved the theorem of “Principia Mathematica”, a book on the basics of mathematics, by combining various axioms.
  • “Dartmouth Conference”
    • It was held in 1956.
    • The term “Artificial Intelligence” was used by “John McCarthy”, the proposer of the conference.
    • A demonstration of “Logic Theorist” was also held during this conference.


1st AI boom

The 1st AI boom began in the late 1950s with the “Dartmouth Conference” as the starting point.
Active investment and research was done in the research of “search” and “inference”. This searches for a solution by brute force while considering conditional branching of the problem. And it infers how to get to the solution quickly.
As AI algorithms, “search tree”, “planning”, “neural network”, etc. were created.

However, “search” and “inference” could solve toy problems such as mazes and chess, but could not solve complicated real-life problems. Many things happen in a real problem. Since all the factors that occur are part of the problem, the combination of the factors increases explosively (combination explosion). And it was not possible to search for all the factors with a computer that has only finite information processing capability (frame problem).
Therefore, in the late 1960s and early 1970s, the “ALPAC Report” was issued in the United States and the “Lighthill Report” was issued in the United Kingdom, and investment was discontinued. And the 1st AI boom is over.


2nd AI boom

The 2nd AI boom began in the 1980s.
At this time, research was actively conducted to put “knowledge” into computers. This is an approach that attempts to solve a complex real problem by directly inputting a large amount of real-world knowledge into a computer because “search” could not solve a complicated real problem.
As an AI algorithm, an expert system was created. It first inputs a great deal of specialized information into the computer. By doing so, it is possible to return the input information as an answer to the question asked. This movement causes the computer to behave like an expert.

However, at that time, computers could not collect and accumulate the necessary knowledge by themselves. Therefore, it was necessary to describe and input knowledge in a state that humans can understand on a computer.This work performed by humans was a huge amount of work and was not easy. Therefore, this boom also ended in the early 1990s.
An example of this work is introduced. There is a Cyc project that tries to build a huge database that collects all the knowledge that human beings have. It started in the United States in 1984. However, it has not been completed even after more than 30 years. From this example, it can be seen that this work is enormous.


3rd AI boom

The third AI boom began in the latter half of the 2000s.
This is triggered by the activation of research on machine learning and deep learning.
Deep learning was announced in 2006. Then, in 2012, at the competition ILSVRC, which competes for the recognition rate of objects, the team of “University of Toronto” won with an overwhelmingly good recognition rate by using deep learning rather than the conventional method. Deep learning received a lot of attention from this event.

Furthermore, in 2012, Google announced that “AI succeeded in spontaneously recognizing cats without human teaching.”
First, they randomly picked 10 million 200×200 pixel-sized images from videos on YouTube. They then used the image without labels and teacher data, trained it for a week by deep learning, and developed an AI model. This developed AI model was able to recognize images of cats.
In other words, it became possible to recognize a “cat” only by a machine without any human teaching. To be precise, because it is unsupervised learning, it does not classify what was taught as “cats”, but rather groups “groups of creatures with the concept of cats”.

Originally, machine learning and neural networks were proposed before this period, but they have not been used much until now. Behind this boom was the spread of the Internet, the birth of search engines, and the birth of cloud services. In other words, these events made it possible to accumulate a large amount of data and use a huge amount of computer resources, and to process data by combining them. And this progress has contributed to the improvement of machine learning and deep learning.


Present and Future

The 3rd AI boom is not over yet. This is because, unlike the 1st boom and 2nd boom, many of the AI currently in use are at a practical level and have come to be used in various real-life products.

With the development of AI, AI can replace the work that humans are doing now.Therefore, society is expected to become more convenient and change significantly.
Furthermore, according to one theory, it is said that “Singularity” that artificial intelligence exceeds the capabilities of the human brain will be in 2045.


The roots of AI currently in use

Currently, various artificial intelligences are used on a daily basis. Many of them have evolved day by day in the history described above and are in their present form. We give an example of its roots.

  • Siri from Apple Inc.
    • ELIZA (Natural language processing program) -> Siri
  • Autonomous driving / Various image recognition
    • Machine learning / Neural network -> Deep learning -> Autonomous driving / Various image recognition
  • AI program for chess and shogi
    • Search Tree / Tower of Hanoi -> Planning -> AI program for chess and shogi
  • IBM Watson
    • Dendral / Mycin -> Expert system -> IBM Watson

Sponsored Links