In many fields, artificial intelligence (AI) is attracting public attention. Since the beginning of the 2010s, AI has rapidly developed due to breakthroughs in deep learning. Together with IoT and big data, AI, which is enabled by the highly sophisticated processing capacity of computers and software technologies, is expected to greatly transform our lifestyles. The age has now arrived in which AI is truly spreading. How will AI change our daily life, industry, and society?
[Fujitsu Forum 2017 Conference Report]
Current and Future Trends in AI
During the first half of the conference, experts from various fields gave lectures on three subjects: (1) an outline of "narrow AI," which can handle only specialized tasks, and "evolving AI"; (2) initiatives for AI research and development; and (3) future healthcare that makes the most of AI.
Today, Humans Stand at a Crossroads to Make the Most of AI
The AI developed thus far is "narrow AI." AI that has been developed to play chess or shogi can handle only specialized tasks related to chess or shogi. Even narrow AI, however, will be used for a wide range of applications in society by 2020. This is because smart machines such as autonomous cars, drones, and tractors will start to spread. In 2025, AI that understands language will appear, and when installed in robots, the AI will serve customers or work as butlers. In 2030, artificial general intelligence is expected to appear. Like humans, this AI will be able to apply its intelligence to various situations. Humans have versatile intelligence, which enables them to carry out diverse tasks; in the future, AI will be created to manage these general tasks.
It is also said that in 2030, the fourth Industrial Revolution will occur, bringing about dramatic changes to production structures. This means that machinery will replace humans in production activities. From 2030, large gaps in economic growth rates are likely to become apparent between countries that raise production automation to a high level by introducing advanced technology, including artificial general intelligence, and those that do not. In other words, we stand at a crossroads today where in various aspects, disparities among businesses and among individuals will widen depending on whether or not they make effective use of AI.
What, then, should we do? First, we must further advance automation. Right now, we are in the middle of the third Industrial Revolution. We must make greater use of IT, and meanwhile, we must increase not only the number of AI researchers but also human resources in the humanities who can put AI to practical use. After that, businesses should keep in mind that humans should focus on creative jobs that only humans can do.
Leading-edge Research on Machine Learning by Riken Center for AIP
In FY2016, the Ministry of Education, Culture, Sports, Science, and Technology launched the Advanced Intelligence Project (AIP). This integrated project involves AI, big data, IoT, and cyber security, and to carry out this undertaking, Riken established the Center for Advanced Intelligence Project. The AIP Center is implementing five projects: establishment of next-generation fundamental technology, scientific advancement, contribution to social implementation, coping with ethical and social challenges, and human resource development.
The battle for AI research and development is fought on a global scale, but when it comes to major applied research projects, the scale of budget granted is already deciding the winners. By contrast, it appears that the battle for basic research is still fought on an individual basis. Director Sugiyama emphasized that it was important to focus on research topics that would bear fruit ten years later rather than pursue those that are currently enjoying a boom.
Though deep learning is attracting much public attention nowadays, there are still many problems that cannot be solved using today's technology. Director Sugiyama went on to say that the AIP Center would continue to concentrate its energies on basic research in deep learning to further improve AI performance and efficiency.
"Big data + deep learning" is actively being studied, but the high cost to label the data needed for analysis has become a practical problem. Therefore, Riken researchers are trying out various methods such as classification using positive and unlabeled examples. In the future, it will become even more important in machine learning research to develop methods with low labeling cost and a high learning accuracy.
AI and Supercomputers Opening up the Future for Healthcare and Drug Discovery
In his laboratory, which focuses on developing innovative technology for precision medicine and drug discovery, Prof. Okuno specifically works to advance healthcare and drug discovery by making free use of simulation technology and AI. In precision medicine, it is becoming possible to easily obtain information on the human genome, and research on using such information to choose more effective, safe treatment methods based on individuals' physical makeup is underway.
If information on an individual's genome sequence is available, for example, doctors can determine which diseases he or she is likely to develop and which medicines may readily take effect. If this information can be obtained per patient, doctors can provide precision medicine. For this reason, genomic medicine or precision medicine is advancing around the world, and AI is attracting much public attention as it can solve problems in this area.
One example of an AI for genomic medicine is IBM's Watson. In Japan, Kyoto University and Fujitsu worked together to launch a project to develop a Japanese version of Watson in an effort to create an AI properly using data on Japanese genomes.
Watson is well-known for its application in cancer treatment, but the Kyoto University-Fujitsu project, which aims to build AI for various types of genomic medicine, not just cancer, will develop a next-generation AI that can perform prediction.
In pharmaceutical development, two major problems are development costs and time. It is estimated that it costs about ¥120 billion to develop one drug. The development time is estimated to take at least ten years. Advanced healthcare is needed to sophisticate healthcare, but it takes much money to introduce such advanced healthcare. Given these circumstances, Prof. Okuno aims to introduce supercomputer and AI technologies for drug discovery to dramatically reduce pharmaceutical development costs. For example, he intends to shorten the development period by four years and to reduce the industry's overall development costs by around ¥1.2 trillion.
In this initiative, supercomputer and AI technologies are an extremely important weapon. In this sense, the professor places high hopes on IT businesses in Japan.
How Will AI Transform Our Future?
With the moderator joining the three experts who had taken the podium earlier, the second half of the conference featured a panel discussion about the "present state and future of AI" from three perspectives: society, business, and ethics.
The Society and Business Brought About by AI
The moderator, Hirotaka Hara, set three themes: AI and society, AI and business, and AI and ethics. Regarding the first theme, "AI and society," Riken's Mr. Sugiyama explained that there was still a large gap between narrow AI, which is dedicated to a particular task, and artificial general intelligence, which has intelligence like that of humans. He stated that it is unnecessary to worry about the emergence of AI that will displace humans, which people currently fear.
Kyoto University's Prof. Okuno described health professionals' expectations for AI and its involvement in the healthcare sector. Healthcare, his specialty, is one very important social issue facing Japan. While there are challenges such as the aging of the population and soaring medical expenses which AI can help mitigate, there are concerns about diagnostic errors made by AI. Prof. Okuno explained: "There are concerns about what doctors will do if an AI misdiagnoses, but AI-based diagnoses will not lead directly to treatment. Rather, assisted by AI, doctors will make the final decision. We are also discussing this matter with the Ministry of Health, Labour, and Welfare." He further emphasized that even if AI spreads, doctors' involvement will be necessary.
He also described the healthcare sector's great expectations for AI: "Even if a doctor is not a specialist in a particular field, he/she can administer safe treatment using AI with a certain degree of medical knowledge."
Associate Prof. Inoue of Komazawa University stated that the spread of AI would change trends in the distribution of workers and that the number of people engaged in work of the mind, such as creation and management, was likely to increase considerably.
Will Current Jobs Disappear If AI Spreads?
Under the next theme, "AI and business," the discussion focused on "Will AI eliminate our jobs?" If current jobs disappear, what should we do?
In response to this question, Associate Prof. Inoue disclosed a scenario in which only about 10% of the population will be working around 2045. He predicted the future: "Even today, the shift to IT is changing the nature of jobs and work styles, but if we consider AI as a more intelligent form of IT, AI will have even greater effects on our lifestyles."
Referring to the emergence of such general-purpose AI, Mr. Sugiyama commented: "From a scientist's perspective, I cannot predict exactly when it will occur, but it will not happen in the near future. Before it happens, we must achieve a considerable number of breakthroughs in succession."
There was another question: "Will AI deprive people in all industries of jobs?" Prof. Okuno responded: "Healthcare consists of interactions among people; for example, the way patients are cured varies depending on how nurses treat them. That is not what AI does." He continued: "Health professionals expect an environment in which the heavy burdens imposed on them are reduced mainly by leaving what machines can do up to AI."
Mr. Sugiyama explained that in this respect, "The machine learning we are researching represents the brain of AI, but when AI systems are actually introduced into society, research on interfaces with humans will become important." He added that if basic research on AI developed into something considerably clever, though mental work might disappear in a sense, real interfaces with humans would always remain. Thus, he expressed his opinion that if it lacks in hospitality, the development of AI alone will not bring dramatic changes to society or work environments.
AI's Development Requires New Ethical Standards
There is no doubt that AI will permeate society in the future. When that happens, however, ethical issues will be raised. For example, how far can autonomous driving and diagnosis/treatment be left up to AI? If problems arise, who takes responsibility? To conclude the panel discussion, participants exchanged views about ethics for AI.
Prof. Okuno expressed the opinion that the responsibility for using AI lay with the user. For example, he said, if a doctor has something to check, today he or she searches for information on Google, but if an AI did the job on the doctor's behalf, it would be very helpful as the doctor has little time to read theses. However, in the end, how much the doctor leaves such searching up to the AI and how the doctor evaluates the information produced by the AI depends on the doctor's skill.
In addition, Mr. Sugiyama presented concerns about AI usage. One example he gave was that it would be relatively easy to use AI to screen candidates for new employees, but who would take responsibility if there were defects in the screening process that did not involve humans?
The moderator, Hara, concluded the panel discussion by stating that there was still a mountain of problems that must be solved for AI to spread and be used safely. He went on to say that the introduction of AI was a global trend, and it is necessary to establish systems and continue to invest so that Japan is not left behind in this competition.
Yasushi Okuno Professor, Graduate School of Medicine, Kyoto University
Masashi Sugiyama Director, Riken Center for Advanced Intelligence Project
Tomohiro Inoue Associate Professor, Faculty of Economics, Komazawa University
Hirotaka Hara Corporate Executive Officer, Fujitsu Limited