Can Artificial Intelligence really be made? Part 1 of 3
[Beginning of Artificial Intelligence]
The term “Artificial Intelligence” is considered to be used by John McCarthy, computer scientist for the first time at Dartmouth conference in 1956. Since then it has been passed 60 years. For ordinary people the definitions and understandings of Artificial Intelligence vary, so it is often a topic of discussions what should we call as the Artificial Intelligence. However, I would like to avoid sticking to its definition but I would like to broadly capture it as a computing processing that can do the things only human can do in the past, or a computing processing that can do better than human.
In the world of movie, one of famous Artificial Intelligence was HAL9000 which appeared in the movie “Space Odyssey” (released in 1968), original book written by Arthur C Clarke. The name HAL is well-known as it is made from the one letter shift from IBM, world famous computer manufacturer then.
This HAL 9000 was equipped in the spacecraft to explore planet “Discovery One” to control the ship’s operations and the HAL was able to chat with crew in natural language, play chess and even sing songs. For the details, I suggest you to see this great masterpiece movie and to pay attention to the changes of behavior of HAL 9000.
In the movie, HAL9000 was born in Jan 12th, 1992. However, in our real world such computer with intelligence like HAL9000 has not been born. It might be the world of SciFi will be ahead of our real life, mightn’t be.
Research activities in Artificial Intelligence are paid attention and conversely experienced so-called “winter age” several time. Twenty years has been passed since the birth of HAL9000 in the movie, there are a lot of application domain of Artificial Intelligence and become reality.
[Computer behavior is determined by data ?]
Professor Marvin Minsky of Massachusetts Institute of Technology (MIT) created the foundation of neural network analysis with Seymour Papert, known as computer scientist as well as development psychologist and creator of LOGO, programming language. Neural network is a network consisted of nerve circuit, creates a structure of mutually connected nerve mesh and distributes data to process. Neural network was mimicking a part of brain system and was applied to the tasks like inferences and recognitions which was considered difficult for the computer of that days.
These days Deep Learning as a part of foundation in recent Artificial Intelligence is being a popular topic. This deep learning is also known as “深層学習 (Shinso Gakushu)” in Japanese and was extended the concept of neural network by Professor Geoffrey Hinton at the University of Toronto around the year of 2006.
With the following wind of dramatically increasing computing power of computers, it becomes feasible to process vast volume of data which has never been possible. With another following wind that data becomes available to obtain easier than ever, it becomes possible to make a behavior of computer by data which is totally opposite approach from the past.
Deep Learning has provided tremendous impact to researchers who pursued traditional approaches. For example, in the domain of image recognition and character recognition that computer classifies images to right category, this Deep Learning approach showed the result that exceeds the result of traditional approach which gradually improved over time.
In addition, the research which using this Deep Learning was conducted by the researchers who has not been involved in the research field of image recognition and character recognition, and it is reported that many researchers who are top-runner in the research field was received a huge shock form the fact.
[Finally computer won against Go professional]
One of well-known application of Artificial Intelligence is Chess, Shogi (Japanese Chess) and Go between human vs Artificial Intelligence. In the Chess game, in 1997 at second encounter, IBM’s Deep Blue won Kasparov, the human Chess Champion by 2 win 1 lose 3 draw.
For Shogi or Japanese Chess, after the research was started around 1970s, the matches between computers were made in 1980s and then the matches between computers and human was performed as well. Computer Shogi called “Den-no Shogi” as well was being publicly known since “Shogi Den-Oh Sen (将棋電王戦) “ was regularly held since 2012 and was received an attention from various media.
In the first “Shogi Den-no Sen”, retired professional Shogi player, Kunio Yonenaga, fight against computer Shogi and Computer Shogi won. After that, it became regular event as group in various form such as group match.
For Go, the news was spread around in Feb 2016 that a UK company “DeepMind” who is acquired by Google in 2014 won the Go professional in Europe with Computer Go software “AlphaGo”. This “AlphaGo” does not learn the rule of Go, instead using Deep Learning technology previously mentioned and the AlphaGo is considered that it learned 30millions of patterns from Go move archives and found the best move in the match.
Among Chess, Shogi (Japanese Chess) and Go, Go is most complex because of the variation of the patterns. The combination of patterns are relatively equal to the number of atoms exist in the universe therefore it requires the volumes of computations that will not finish within the realistic period of time if you use the conventional method.
(Continues to Part 2 of 3)