Japanese-style AI coming up next.
Japan is about to launch a gigantic industry-government-academia project to develop artificial intelligence technology matching the nation’s specific needs.
The project will include one of the country’s biggest research centers, set up in the government-backed Riken research institute, and has a budget of 100 billion yen (around $974 million) over the next 10 years. Some 100 top researchers from participating companies, such as Toyota Motor and NEC, and research institutes will begin studies soon at a laboratory to be opened by the Riken Center for Advanced Integrated Intelligence Research in a high-rise building near Tokyo Station.
Positioned as a pillar of the Japanese government’s growth strategy, AI simulates the human brain to create “wise” computers. Some experts forecast the technology will surpass human intelligence by 2045.
Competition for the development of AI technology is about to heat up. Studies on AI began globally more than 50 years ago and gathered steam in the 1960s and 1980s, but programs for its practical application ended up in smoke in the absence of advanced computers capable of performing required functions.
In contrast, the current AI boom is widely considered to be the real thing. Practical application is ready to start now that three necessary technologies — automated processing of data, high-volume computing and processing of voice, image and other big data — are available, said Yuichiro Anzai, chairman of the Japanese government’s Artificial Intelligence Strategy Committee.
“Deep learning” is an AI technology that is drawing attention due to its ability to extract information from large amounts of image, voice and other data. University of Toronto researchers discovered the potential capacity of deep learning, a branch of machine learning, in 2012 during experiments to discern images.
In March, the AlphaGo computer program, developed by a Google subsidiary’s deep learning team, beat a world champion professional Go player, an achievement that had been considered impossible for another 10 years.
A computer science discipline called “question answering” is also contributing to advances in AI technology. At the University of Tokyo, IBM’s QA-based Watson cognitive system, which provides call center solutions to banks, has found treatment plans for cancer patients for whom doctors struggle to develop diagnoses.
The QA system quickly finds solutions from a huge number of dictionaries. It is finding its way into natural language processing, which is increasingly used to analyze the contents of social networking services and to facilitate machine translation.
Deep learning is also effective in searching information on the internet, driving cars automatically and controlling robots.
While large U.S. companies have ample big data for which deep learning technology can be utilized, no Japanese companies handle data on the same scale. The Japanese government therefore decided to launch the domestic AI project with the aim of addressing this weakness, seeking to develop new types of AI that can find optimal solutions from a small amount of data.
“Japan should promote AI studies in its own way,” said Masashi Sugiyama, head of the Riken Center for Advanced Integrated Intelligence Research.
The use of AI to find solutions from limited data also matches private-sector needs. As a result, the project has won support from Toyota, NEC and other companies.
Japan will make use of AI to address problems such as old infrastructure, the rapid aging of the population and soaring medical expenses. Nursing-care robots will be developed that can listen to users even in noisy environments. In the medical field, AI will be used to find treatment plans for each patient, even if medical records are insufficient.
In short, AI technologies that do not rely on big data are needed for Japan’s future.
Studies on AI that utilizes a limited amount of data are already emerging. Researchers at the Tokyo Institute of Technology have developed the Self-Organizing Incremental Neural Network, which predicts the flooding of rivers to compensate for a shortage of experience. Fujitsu Laboratories is working to find defective items among products at plants.
Knowledge of mathematics and statistics is key to the realization of new AI technologies. Cooperation between researchers well versed in AI theories and on-site experts is indispensable.
“In Japan, with no giant IT companies, it is important for multiple companies and universities to join hands,” said Junichi Tsujii, director of the Artificial Intelligence Research Center at the National Institute of Advanced Industrial Science and Technology.
In the 1980s, Japan failed in an initiative to develop a fifth-generation computer capable of making logical inferences to solve problems. Although Japan spent more than 50 billion yen on the effort for 10 years, a special computing machine developed by the project was too complicated to become practical.
Japan is still short on AI experts, although it has produced a considerable number of competent young people in the field. The nation will be tested by its attempt to develop useful AI the Japanese way.