Among Japan's industries, the manufacturing industry has been playing a leading role in the use of artificial intelligence (AI). AI has already improved a number of manufacturing and sales processes and reduced the cost of production and marketing activities. The industry applies image recognition technology powered by machine and deep learning—the core component of artificial intelligence—mainly to the inspection of machinery and equipment, detection of production line anomalies, failure prediction, quality inspection, and sensory inspection. Factors contributing to the increasing use of AI in the manufacturing industry include the declining birthrate and aging population, which make it difficult for the industry to find successors who can replace retiring veteran engineers.
AI has a wide range of applications not only in the manufacturing sector but also other areas, including medical care, drug discovery, and environment conservation monitoring systems. A particular focus of mine is the use of AI in urban design. The use of satellite data, which demonstrates how the land is used and illuminated at night, has enabled researchers to understand urban areas from both a wide and detailed perspective. We have already applied deep learning to our research and succeeded in predicting future land use at a higher level of accuracy compared to the past. This AI-based, new urban prediction is changing the field, where people traditionally argue about urban design based on individual land use and blueprints while referencing paper-based maps.
One of the key questions in urban design is the consideration of transportation networks. Public transportation and motorway networks have been constructed on the premise of increasing populations and public budgets. However, against the backdrop of a declining population, an aging society, and a labor shortage, their maintenance and repair costs are underfunded, forcing operators to reconsider their management policies. Automated driving technology holds significant potential in solving these issues.
In the future, cash-strapped local bus operators, who are currently relying on public subsidies, will reduce the number of routes and the frequency of bus service. Automated vehicles are expected to provide a means of transportation that can substitute for these local bus services. In particular, local municipal governments can maintain the level of service for vulnerable road users by utilizing automated driving vehicles as a means of door-to-door transportation for the elderly. These municipal governments are eager to stop their population from declining; if one such municipality has successful results, others are expected to follow suit, allowing automated driving technology to be adopted at a faster pace than in urban areas. Our analysis of the demand for automated driving shows that Japan's elderly have a greater demand for automated driving compared to their counterparts in other countries.
Adoption Cost Is a Hurdle for Dissemination
Automated driving technologies have evolved gradually over time. Some of these technologies, including the technology that enables a driver to follow the car in front of them on a highway without touching the steering wheel, have already been commercialized, and a number of them have been adopted in vehicles available in the market.
Furthermore, countries such as Japan, the United States, China, and Singapore have been conducting demonstration experiments using automated cars on public roads. In Switzerland, automated electric buses are already in service on public roads, although they are available for limited hours. Today, the technologies have advanced to the level at which buses drive themselves on regular routes and driverless taxis operate in limited areas.
However, even if automated driving becomes a reality, high adoption costs will discourage dissemination. What, then, is the adoption cost that will allow the technology to prevail?
Our research found that hybrid car users can afford additional costs in the range of the 190,000 yen level for Level 3 automated driving technology (conditional automation; the user is expected to respond appropriately to the system's request to intervene), and in the range of up to the 310,000 yen level for level 5 automated driving (full driving automation; the user is never expected to intervene). Electric vehicle users can afford a cost increase in the range of the 100,000 yen level for level 3 and up to the 220,000 yen level for level 5. On the other hand, most hybrid vehicles require users to bear a cost in the range of between 50,000 yen to 80,000 yen. This means that the persistently high additional cost which is necessary for the purchase of automated driving functionality may pose a challenge to dissemination.
The Japanese government has set a goal of achieving at least level 3 by around 2020 and level 5 in subsequent years. However, at present, it is technically unfeasible to reduce the additional cost to the level acceptable to consumers, i.e. 100,000 to 190,000 yen for level 3 and 220,000 to 310,000 yen for level 5 (in the case of hybrid and electric vehicles).
Given the current pace of technological advancement, it would not be possible to reduce the manufacturing cost to the price range we found consumers would be willing to pay. Therefore, either of the following two options will need to be chosen in the future: to subsidize the purchase of automated cars by reducing user's fuel cost to virtually a zero level at the government's expense; or to reduce the price through innovation created as a result of basic research, while selling automated vehicles at a low price at the expense of manufacturers and/or dealers. The former is the approach taken for electric cars and the latter is the one employed for fuel cell vehicles.
Negative Environmental Impacts of Automated Driving
However, even if the cost issues are solved, the following challenges will emerge: while the use of automated vehicles reduces the risk of accidents and driving fatigue, an increased in car use demand will result in higher fuel consumption and an increase in greenhouse gas emissions.
To deal with this, we estimated the impact of automated driving on the car use demand, using a survey conducted among consumer households. As a result, it was predicted that automated driving would increase annual car mileage per household by approximately 600 km to 3,300 km. In other words, we must consider that the possible increase in the consumption of fuels with the increase in automated driving will be accompanied with cons in addition to the pros. If mileage increases by approx. 600 km to 3,300 km, CO₂ (carbon dioxide) emission will grow by approx. 6.5 million to 33.82 million tons per annum.
Accordingly, in order to curb fuel consumption and greenhouse gas emissions, automated driving technologies must first be introduced to fuel-efficient vehicles such as hybrid, plug-in hybrid, and electric vehicles, rather than gasoline-powered vehicles. It is also necessary to pay close attention to the review of road and other transport infrastructure, as increased usage is expected to worsen traffic congestion.
The government has announced that it will use AI as the key driver of "Society 5.0," a concept of a future society that utilizes cutting-edge technologies to improve urban capabilities. In most cases, AI may play a substitute role for one-time, "single-point" technologies and developments, such as automated driving technologies and the deep learning-based land-use prediction discussed herein. However, considering the fact that technologies—just like those used in AI-based automated driving systems—have inherent pros and cons, it appears that we are now entering an era when the more technologies advance, the more humans are required to have the wisdom to connect the dots and find the fields in which AI will be able to best contribute to solving problems, like urban design and societal development.