In a recent tweet, Lex Fridman, a research scientist working on human-centered artificial intelligence at MIT, said: “all of us working in autonomous vehicle research want nothing more than to save lives.”
While AI scientists and engineers share this noble goal, they often differ dramatically in the path they take to reaching it.
Companies such as Uber’s Advanced Technology Group, Tesla and General Motors’s Cruise take an approach many will consider maverick, perhaps even cavalier: they deploy vehicles equipped with newly-developed autonomous operation capabilities on public roads and test them under real-world conditions, driving millions of actual miles and many more virtual miles using advanced simulation software. Design engineers monitor the performance of the robotic software and update the algorithms inside the vehicle, sometimes remotely and in near real time.
This strategy continues to grab headlines that trumpet the impressive abilities of these vehicles to navigate extremely challenging traffic situations. But this approach has also resulted in a number of highly advertised crashes and fatalities that generated equally loud headlines, calling for increased regulations and limit what some describe as Silicon Valley brash approach to product testing. And federal and state laws governing self-driving cars continue to be in flux.
Japanese automakers, on the other hand, have maintained a much lower profile.
Japanese OEMs are adding ADAS feature to production vehicles and working on all aspects of autonomous operations technology, but there’s little physical evidence of any of these, especially when compared to the American technology companies who are very quick to release innovation from the lab to the wild.
Is the lack of publicity an indication that the Japanese technology is lagging behind its American competitors? Unlikely.
Japan has a long tradition of theoretical and applied research in robotics, motion control and AI. I remember the alarms sounding in the West in 1983 when details about Japan’s Fifth Generation Computer project were revealed in an article by of Ehud Shapiro. The project’s goal was to leapfrog Western computer expertise and create an entirely new computing architecture that would give AI engineers unparalleled computing power and highly potent AI languages (I actually attended one of Shapiro’s doomsday lectures).
So perhaps there are no significant differences in the state of autonomous car technology development between Japan and U.S., but each has a different attitude and roadmap to realizing their vision?
Japanese OEMs and government bodies are taking a long-term goal-oriented approach, focusing on the potential value autonomous vehicles will bring to the country’s rapidly aging population. AI researches are developing commercial applications such as long-haul trucking, which has a greater business value than the personal urban mobility focus of US companies.
In a push to boost its economy and ready driverless cars for the 2020 Tokyo Summer Olympics, Japan’s National Police Agency drafted rules for testing driverless vehicles on public roads while the vehicles are being monitored remotely.
At the end of the day, could Japan’s seemingly more pragmatic approach to technology testing and articulating market value lead to a smoother and faster market adoption of driverless cars? Is Japan’s silent strategy for driverless cars the tortoise to America’s hare?
Beyond the fascinating debate which long-term market strategy will win the race, there are many questions about the execution of the strategy and the path to success. Here are a few to get you thinking about the subject:
- What regulatory mechanisms and best practices are needed to conduct road testing safely and protecting citizens while not stifling innovation?
- Are companies developing autonomous driving capabilities focusing on the right problems? Is urban personal mobility a sensible goal or should companies work on mobility solutions for the elderly and people with disabilities? Are companies better off seeking potentially simpler yet more profitable commercial transportation?
- Can machine learning algorithms learn without driving millions of miles on public roads? Should researchers make more use of advanced simulation, which will make testing safer, cheaper and faster?
Join the Debate
SAE Connected and Automated Vehicle Conference Israel will host a live debate on the topic. On January 17th, a Battle of the Minds panel will debate the merits and risks of two opposing strategies. One side will argue for the need to adopt a comprehensive goal oriented and risk-averse approach that relies on tried-and-true engineering, simulation and validation methods. The opposing side will maintain that the only way to mature safe autonomous driving capabilities is by driving more miles under real-world conditions.