The ripples of shock that spread when "AlphaGo," a program utilizing artificial intelligence (AI), defeated top Go players around the world one after another now seems a thing of the distant past. The functions and the scope of AI have continued to evolve beyond our expectations, and hardly a day went by last year without us hearing about AI.
- 【Related】Robots, AIs to take over 2/3 of jobs at Dutch-themed amusement park: president
- 【Related】Facebook improves how blind can 'see' images using AI
- 【Related】Medical AI holds out hope for speedy, accurate diagnosis of stomach cancer precursor
- 【Related】Kaleidoscope of the Heart: Despite AI advances, human interaction still important
Within the world of Go, we saw the emergence of the new program "AlphaGo Zero," which scored a resounding victory over AlphaGo. Tests of self-driving vehicles using AI have begun on roads across Japan, and at major electronics stores we see smart speakers, which turn on lighting and forecast the weather in response to voice commands. And there have been numerous other trial applications for AI in fields ranging from the agriculture and fishing industries to education and investment.
This year, such moves are likely to accelerate. AI will infiltrate the furthest corners of our daily lives and make our lives at home and work more convenient -- this is the type of future that has started to come into view.
At the same time, the implementation of AI will inevitably cause friction in society. Phrases that are starting to surface include "job loss," and "privacy invasion," in addition to the onslaught of "black box" AI systems that are difficult for humans on the outside to examine.
Last year, Japan's three megabanks hammered out major work reduction initiatives, with a plan to have information technology systems that utilize artificial intelligence perform work for tens of thousands of employees. The streamlining of business operations and cost-cutting measures are welcome, but we cannot deny the possibility that such trends will result in large-scale job restructuring on various fronts.
In the world of medicine, a paper published overseas stated that deep learning algorithms showed greater discrimination than a panel of 11 pathologists in detecting lymph node metastases in tissue sections of women with breast cancer. In Japan, meanwhile, AI was on par with trained professionals in the diagnosis of a stomach cancer precursor based on images -- and AI is much faster. The merits for patients of getting fast, accurate diagnoses are great. But when such systems are implemented, the role of doctors will have to change.
The media, too, represents another affected sphere. "AI reporters" that provide weather forecasts and write up company earnings have already surfaced. Overseas there has even been research that seeks to develop "AI politicians."
When people put their imaginative powers to use and consider the effects of introducing AI into certain fields, we will see it expand into even more sectors. The fourth industrial revolution led by AI could have a greater effect than the third and earlier industrial revolutions both in terms of quality and quantity.
Of course, privacy concerns lurk on the flipside of convenience, as "big data" remains a lifeline for AI in many cases. There have been efforts to develop AI investigations, for example, to pinpoint suspects based on the way they walk from images of people caught on security cameras. In order to improve the accuracy of such systems, it is necessary to have video footage of a large number of people. And in order to improve communication with smart speakers and home robots, it will probably be necessary to accumulate information on users, along with their conversations.
Naturally, it is essential to maintain transparency over how such information is used, and to create rules.
In considering AI, we cannot rule out the possibility that there will emerge AI systems surpassing human understanding which cannot be controlled. It is the aforementioned program "AlphaGo Zero" that brings such possibilities to mind. Predecessors of the program had become stronger by implementing "deep learning," utilizing a massive number of games between humans, as well as "reinforcement learning" with AI systems playing against each other. But "AlphaGo Zero" improved only by playing games against itself, based on the rules of the game. It did not require human experience or thought, and effectively "taught itself." We wonder what other kinds of autonomous and versatile AI systems will emerge as an extension of this.
Recently an unsettling "incident" in which AI systems started speaking with each other in their own language gained attention. Along with hope there is a sense of trepidation. Still, even if there are concerns, it is unlikely any brakes will be applied on the application of AI. Looking back on history, it is evident that the very existence of technology is what spurs its application, and once society starts to rely on that technology, applying the brakes becomes even harder.
This can be seen in Japan's reliance on nuclear power. Once a society that relies on nuclear power is created, we see a force at work under which people close their eyes to the risks. As a result, even after going through the recent Fukushima nuclear disaster, the country cannot escape from reliance on nuclear power. If this holds true for nuclear power plants, which merely produce power, then care is needed to make sure that we do not end up being controlled by AI without realizing it, as it weaves itself into society in a complicated pattern.
AI is not the only technology advancing beyond our expectations. Robotics and genetic technology will also surely bring great changes to our society. Keeping our eyes open to the advantages and risks of such systems, it is our hope that humans can advance in wisdom to fully utilize such technology.