At the center of research field which is now called artificial intelligence or AI, there is deep learning. With the advent of this deep learning, it is certain that the current AI boom occurred. But in AI research, deep learning is only one technology. AI, ” artificial intelligence ” is an attempt to artificially create human intelligence. The realization of artificial intelligence that Turing and Minsky also dreamed about are still being vigorously studied. On the contrary, thanks to the great success of deep learning, in the United States, Europe, China and Japan, the government led the enormous research expenditures and various research projects are under way.
MICRONS of the United States Defense Advanced Research Projects Agency says that the neural network is still only a technology inspired by the brain and is not actually a reproduction of the brain and the difference between the current AI algorithm and the actual mechanism and action of the brain It is aimed at filling in. The European Human Brain Project aims to introduce a billion dollars and simulate the human brain completely computer by 10 years.
There is also a strange brain science project. The Russian millionaire nonprofit project aims to develop a technology that enables individuals to transfer individual personality to a more sophisticated and abiotic tray and to realize a life extending life with an eye to elderly immortality . Research on uploading such human consciousness to computers and research on mind clones has already been gathering huge amounts of funds in the United States and research has already proceeded.
In Japan, AIST serves as the research and development base for “next generation artificial intelligence and robot core technology development”, and is trying to apply AI technology to the industrial world. RIKEN also runs many AI research projects. But Japanese projects are generally small, mainly application of AI. In Japan, once the bubble became gorgeous, there was an ambitious AI project in the world of the fifth generation computer development plan. Although the final evaluation of this project was the conclusion that it was “failure”, it should evaluate its challenging spirit.
Still, due to the rapid progress of deep learning research in recent years, realization of( ) gradually in Japan seems to become not a dream. The word “general purpose artificial intelligence” means “all-purpose intelligence that can deal with every problem”. In the researcher organization “Whole Brain Architecture Initiative (WBAI)” for establishing general purpose artificial intelligence founded in Japan in 2015, “acquire themselves multifaceted problem solving skills in various problem areas themselves, Artificial intelligence that can solve problems beyond the assumption “has been set.
This is set as a target for “specialized artificial intelligence” such as image recognition and speech recognition. In other words, if there is only data, it aims at intelligence that can acquire a wide range of problem solving skills by learning from it self.
In October 2017 AlphaGoZero won 100 games against the first AlphaGo who broke Lee Se-dul Nine-tiered just by playing self-match for only three days after being taught only the rules of Go . In other words, without learning any previous game record, I discovered the definite procedure that I was born in a long history of Go, and I also created an original fossil stone. Furthermore, according to the paper at the end of November, he was able to exceed human beings by self-learning to any of the three board games of chess, go and shogi. Before shocking fact that one program can learn three different esoteric games at a high level, general-purpose artificial intelligence gradually becomes reality.
This “whole brain architecture” is an ambitious research to approach from both aspects of artificial intelligence as engineering and neuroscience as science, to acquire intelligence beyond the brain by learning the human brain. To that end, we need two research and development: (1) “Developing each organ of the brain as a machine learning module” and (2) “Integrating these multiple machine learning modules on the cognitive brain architecture” .
(1 ), HTM ( Hierarchical Temporal memory ), which is an algorithm modeling cerebral neocortex, is considered to be strong. And also ② is also accumulation of knowledge in neuroscience, the connection style of the whole brain is being clarified. For this reason, in the past few years, what was once a “whole-brain architecture” in the SF world had been raised as a real research goal even in Japan.
Realization of this general purpose artificial intelligence is, of course, not easy. However, even if you do not reach the ideal general-purpose artificial intelligence, it is certain that various knowledge and technologies generated in the development process up to that will have a big impact on society. WBAI, an NPO corporation, aims to build this general-purpose artificial intelligence with the goal of 2030. And for that purpose, we mention that it is helpful in advancing science and technology and solving human problems of humanity. It seems to be a basic researcher, intellectual curiosity is the abstract purpose with top priority.
The future of AI is not to be expected. What the human race is like AI, what I imagined is AI. The AI that the researchers of the world expect, but the dream AI should be the future AI. Even if whatever future is waiting ahead. Because Alan Kay is not the only one who said that the future is inventing, not anticipating.