March 18, 2018 is a sad time for the 49 year old Elaine Herzberg and her family. The middle aged woman lives in a small city in central Arizona, and the city has been in the news in recent years because of its relaxed attitude towards autopilot. In fact, Uber’s autopilot crashed into a car driven by a human driver in March last year.
After completing this week’s shopping task at the local supermarket, she pushed her bike, appeared in the dark in front of a Uber Vovlvo XC90 car that was driving the autopilot, and then died, and the plastic bags of the supermarket logo were scattered.
The latest car accident video shows that the autopilot technology of the car has not succeeded in avoiding pedestrians, and expensive lidar is like decoration. Another important thing is that the car actually sits on a “test driver” (alternate driver). In theory, when Elaine appears in the field of vision, she should complete the task of emergency braking.
This is the first case in the history of autopilot, and the investigation around it will inevitably lead to protracted discussions under the interest of the stakeholders, the public’s attention and pressure. In March 26th, the Arizona government described the incident as “undeniable failure” and announced the suspension of Uber’s unmanned vehicle testing program in the state.
However, in addition to the truth and accountability of the incident itself, autopilot hit a pedestrian. What problems do we need to think about?
Reporters connected Silicon Valley famous investor, Fusion Fund founder Zhang Lu, and she discussed this matter for Silicon Valley and even the significance of automatic driving industry.
Question 1: is Uber collision a necessary mistake in automatic driving technology?
There are certainly some bug technologies for automatic driving, which can not be completely solved even in the foreseeable future. “For all technologies, this is true. People are also bug aggregates.” Zhang Lu said.
But in her view, the cause of the incident is not due to these unavoidable mistakes.
“In fact, it reflects a kind of fall — the degree of technology development now has a gap with everyone’s expectations.”
How to listen to these four words automatically means that the car can get rid of the driver’s hands and ride freely on the road. However, in fact, a lot of depot technology developed by autopilot is only “auxiliary driving”.
“There are still some technical deficiencies. In extreme cases, it is too late to respond.” The car itself has installed a large number of sensors, radars, laser radars and cameras at 360 degrees. “However, these things are not reliable when there is insufficient light or extreme weather.” Zhang Lu said.
Another problem is the signal delay.
“It may only take more than 0. seconds to respond, and 0. seconds will be enough for an accident.”
Therefore, in this test vehicle, in order to avoid this situation, a backup driver is responsible for manual intervention.
This does not necessarily solve the problem – in fact, not fully mature technology is put on the road and tends to be a habit of overreliance technology, which ignores the danger.
For example, the Tesla tram, the largest shipments. As a Tesla owner, a Silicon Valley entrepreneur is honest, he knows he needs both hands on the steering wheel, but sometimes when the car looks like a car, it’s almost unavoidable to leave the steering wheel when it looks like a car. “Even if I put my hands on the steering wheel, my eyes do not necessarily look at the road; even if I look at the road, my mind may be thinking about other things.”
Numb, he described his own state after entering the automatic driving mode. It is not enough to know that machines are driving or even driving instead of driving.
This is probably one of the reasons why the alternate driver was too late to respond to the accident. In the latest video, she looked down from time to time, reflecting the absent-minded normality of the driver during the test.
So rather than discussing the ethical and philosophical issues of autopilot, the most important discussion of this event may be about “how to determine the reliability of the autopilot, how it can be effectively conveyed to the driver and how to keep the test driver awake.”
Of course, this logical line is more than that, avoiding the tragedies brought by autopilot. What else can we do?
Question two: what else can we do to avoid the disaster caused by this kind of automatic driving technology?
First of all, it is an obvious answer to continue to promote technological progress.
If the laser radar, radar and camera can react more sensitively, cheaper and more observable, autopilot can be more “smart” and more confident in the response to emergencies.
Video shows that Elaine has crossed three lanes before hitting the driverless vehicle. The car is equipped with expensive laser radar, radar and camera, which can theoretically cover the 360 degree range around the body, and night vision should be even more sensitive than the day.
However, it did not respond, and even the basic automatic driving scheme could slow down, but directly hit pedestrians at a speed of 40 miles per hour (about 60 km / h). According to the e-mail content of Professor USC Bryant Smith, the lidar and radar scheme carried by the unmanned vehicle will definitely detect the appearance of pedestrians. The emergence of this accident may mean that the existing technical solutions on this vehicle have major technical flaws.
And some of the more “cheap” sensor options, which may include fewer lidars and more cameras, mean less security?
“In the current technical conditions, the laser radar, radar, camera must be used combined, and good fusion, to provide an effective data collection, processing, analysis and feedback systems.” Zhang Lu said.
Of course, there are other possibilities. In the CNN report, MIT AgeLab scientist Bryan Reimer said this could be a “edge case” and that a machine has never been trained to deal with it.
Second, the formation of the Internet of vehicles and smart city will also make the automatic driving more insurance and more reliable.
“No matter how the automatic driving of these sensors and vehicles progresses, they are still around a car, and the interaction between cars and cars, cars and signals, cars and cities can solve the problem in a higher dimension.” Zhang Lu said.
Imagine that when Elaine came to the roadside at night and was ready to cross the road, it was detected by the signal and conveyed the information to a nearby vehicle, so that the vehicle could have slowed down before she stepped on the road. The road test data under these extreme conditions can also help autopilot ecosystem to become more mature.
“We began to see and invest in the direction of the Internet, because auto driving eventually coexisted with people and cars to just cars and cars on the road. By then, the intelligent transportation ecosystem will become even more important. ” In Zhang Lu’s view, the competition of the vehicle or autopilot system has been very fierce, and the future investment opportunities are contained in the related technology of car networking, traffic network intelligence and 3D precision map.
Finally, from the policy level, this incident may also give the authorities a waking up.
In fact, this year, California is about to open the automatic driving road test qualification without test driver on the vehicle. In the New York Times, it was also mentioned in the article of the New York Times that Arizona opened no road test opportunities for drivers to attract high tech companies and bring vitality to the economy of the central state capital, which has no natural resources and has no advantage in geographical location.
“The local government has to make sure that it has the ability to assess what kind of unmanned car is on the way, which is a warning to them again, and that’s where I will remind them when I talk to some members of the Silicon Valley and the mayor.” In addition to the construction of science and technology, investment industry, Zhang Lu and the Silicon Valley business business cooperation and communication is very close, just by Silicon Valley business weekly as the Silicon Valley is one of the 100 influential women. She told reporters that in fact, the government’s policies and laws are bound to lag behind the pace of technological development, and such caution also has its meaning.
It is not difficult to imagine that the corresponding changes in legal liability, sentencing, and vehicle insurance will follow, which is the only way for the development of the auto driving industry. “This is not necessarily a bad thing. It will promote us to be more prudent, keep thinking and avoid future problems.”
Question three: so in the end, does this matter affect the enthusiasm of all walks of life for autopilot?
As the world economic growth slows down, capital is looking for the next growth point. Automatic driving with complex industrial chain and huge market seems to be a good target. In the past three years, the total investment of autopilot and related industries has reached US $80 billion.
The consequence of capital attention is that many start-ups are forced to accelerate. There are some unprepared autopilot enterprises, which should first solve the basic technology, but are forced to do the whole car, do the bottom technology, and do L4 automatic driving…
“This is actually a kind of” growth and development. ” Zhang Lu thought for a moment, and found a word that can describe the state of today’s automatic driving industry.
This incident is actually a tragedy in the background. “But it is impossible for the whole industry to develop because of this. It is more of a reminder that practitioners and policymakers will pay more attention to how to assess the security and the qualifications of the road, and it is likely to promote more rational thinking and capital investment in the market.” Zhang Lu concluded.
With the influx of capital, the auto driving enterprises start to “jump”. In the face of the first homicide case of automatic driving that shocked the world, the most important industry wave in twenty-first Century was surging. The wheels of autopilot industry are constantly rolling. What we can do is try to remain calm, and look forward to the development of science and technology at the same time to confirm their starting point – Safety first.