Jakob, Online English teacher at Acadsoc.
Want to know more about Acadsoc?
Want to be an online English teacher and write for us?
Apply via acadsoc.ph !
‘In a future suburban development, a homeowner will order an autonomous car, via an app, from a remote solar-charging lot. As a car approaches, it will “talk” to a home: Lights and other utilities are activated or shut off for greater energy efficiency. Because these suburban homes will not have driveways or garages, front yards can be bigger, devoted to ecological functions or recreational activities’.[i]
That’s the life Millennials in the US are dreaming for, moving to a suburb from big cities but can still enjoy full convenience brought by high technology. There are several keywords in their dream, ‘autonomous’, ‘talk’, ‘no driveway or garages’. This indicates the trend, the future of our cars, being driverless, smart and safe.
It doesn’t matter if you are in suburb or downtown. Firstly, if autonomous cars take up the road one day, you are less likely to be caught in a traffic jam. Secondly, even if there is a traffic jam outside, you can still focus on your smartphone or laptop when ‘driving’.
Many of you may feel incredible after Google released a new driverless prototype vehicle or heard another state of America just passed the act to let such cars be on the road. However, the footprint of driverless cars has been almost 100 years’ long when a radio control company in the US called Houdina Radio Control released the first prototype, a wireless vehicle controlled by radio in 1925. What really makes people exclaim today are both driverless and smart—the real autonomous cars, which can finish Dynamic Driving Tasks (DDT) independently with very few or no human interference under the help of Artificial Intelligence.
If you have something to share for Artificial Intelligence, feel free to join the discussion of Artificial Intelligence.
Or you may want to discuss AI’s position in online English classes?
In 1940, Norman Bel Geddes wrote in his book Magic Motorways, under the chapter ‘Eliminate the Human Factor in Driving’:
But these cars of 1960 and the highway on which they drive will have in them devices which will correct the faults of human beings as drivers. They will prevent a driver from committing errors. They will prevent his turning out into traffic except when he should. They will aid him in passing through intersections without slowing down or causing anyone else to do so…[ii]
It may be too ambitious for Norman to dream of an autonomous car, the one that can entirely eliminate the human factor in driving. What he planned for 1960 is more likely a combination of automated driving aid/correction system with highways standardised and armed with highly sensitive sensors.
It is, still a magnificent imagination back then when no one knows what AI is. And what Norman wouldn’t know because he died in 1958, that his solutions led the direction of driverless cars from the 1950s till 1970s.
Many automotive manufacturers and universities, including GM and Ohio State University, tried driverless cars navigated by electronic devices equipped with roadways. Such projects were once given expectations to help solve traffic jam, avoid car accidents and increase the capacity of roads. However, to install sensors and computers on roadways proved to be costly and difficult. Like any other public infrastructure projects, no private companies were willing to pay that much before selling their first driverless car.
Respectively in 1977 and 1979, researchers from Tsukuba Engineering Research Laboratory (Japan) and Stanford University started to try navigating cars based on video processing. As eyes to a vehicle, the cameras installed provides real-time pictures to the central computer, as the brain will give the order to avoid obstacles on the road.
From 1985-1988, Defense Advanced Research Projects Agency (DARPA) launched a projected called Autonomous Land Vehicle (VLA) coded with Carnegie Mellon University and Stanford University. Revolutionary, this program has been move beyond to test some newest technology back then in the 1980s, including many fundamental techniques for Artificial Intelligence. Visual Modeling (discovery of general models to represent objects and natural terrain), Dynamic Image Interpretation (discovery of dynamic environment), Object Recognition and tracking (improved object recognition by self-learning), Obstacle Avoidance (distinguish/evaluate obstacles in the path to avoid)…These technologies are the base for the driverless cars nowadays and many other AI-based products. And finally, in a point-to-point traverse of ALV in 1988, it travelled 20km with a speed of 20km/hr and successfully applied landmark recognition as navigation avoid obstacles based on map and off-road manoeuvring.[iii]
From then on, driverless vehicles are tight with the development of AI. In 1989, Carnegie Mellon University firstly deployed neural processing system for navigation of driverless cars. And in 1995, the university initialised Navlab Plan, aid by PANS platform (Portable Advanced Navigation Support)[iv] carried in a laptop. They used a Pontiac Minivan for a 3,000 (5000 km) drive across the USA from Pittsburgh to San Diego where it was under computer control for 98% of the way[v]. Though Navlab was not fully autonomous when a human driver had to control accelerator and brake or interface from time to time, it was the first time that computer, instead of human beings, try to manage Dynamic Driving Tasks (DDT) in such a long journey.
Actually, the development of driverless cars after 2000 till now still inherited mostly from the research results in last century. Universities and national labs in the US had cooperated to test many of Artificial Intelligence that we frequently hear today.
The difference is many private companies, such as Google, Tesla and Uber, become new players in this territory. Like their predecessors who tried to commercialise driverless cars in the 1950s and 1960s, once again the profit-chasers step into this market, with the rapid development in AI this time.
Formerly led by Sebastian Thrun, the Director of Stanford AI Lab, Google Engineer and one of the creators of Google Street View, Google Driverless Car Department finally separated from Google in 2016 and found Waymo, another subsidiary (like Google) under Alphabet. In 2012, Google Driverless Car started joining the races held by National Association of Stock Car Auto Racing (NASCAR). And in 2016, Google Self-Driving Car won the 9th consecutive NASCAR Race[vi]. And Waymo has a clear advantage compared to its competitor, GM, as a long-history company in the automotive industry, according to the report from California’s Department of Motor Vehicles. Waymon’s self-driving car only reported 63 times of disengagements in a total of 352,545 miles driving in 2017 while GM reported 105 times disengagements in 131,676 miles[vii]. This fact explicates that backed by Google, as a leading company in software and AI development, Waymo seemingly gets the closest position to commercialised driverless cars.
However, there is criticised voice pointing out that even Waymo is far from ready to sell their first autonomous product. Unlike GM, who manufacture vehicles for years, or Uber, who has had a lion’s share in the grab-a-car industry, there is no existed path or customer base for Waymo to promote their products. What’s worse, it seems Waymo has been trapped in a dilemma that they failed to significantly improve the performance of their cars. Disengage for unwanted manoeuvre of the vehicle (30 reported in 2016 and 19 reported in 2017) and Disengage for a perception discrepancy (20 reported in 2016 and 16 reported in 2017) were almost in the same level. While they have entirely avoided the disengagement caused by software discrepancy in 2017 while 50 reported in 2016[viii][ix]. It may imply a difficulty for Google to handle something they are unfamiliar with before, such as CPU, dynamic camera and visual recognition.
As discussed previously, AI cannot entirely replace human teachers, even for online English teachers in very long time because of the complexity of students and their demand. The same thing may happen in driverless cars as well. In the coming decade, there will be more and more states in the US or other countries such as China allowing an autonomous car to be on the road. However, there will be more and more restrictions, such as the requirement of a human driver standing by for emergency when a computer is driving the car, limitation of roads & highways that driverless cars can go, regular report of car performance, etc. And of course, besides legal and technical problems, all AI-based products will no doubt facing the pressure from morality. Who should be responsible if an autonomous car crash a passerby when in autonomous driving mode? The human driver, the company, the law…There will be endless argument and discussion for such topics, but the technology, as I believe, will not stop from moving forward in all cases.
Please ask your kids and ask yourself: Are you ready for the era of AI?
[ii] Bel Geddes, Norman (1940). Magic motorways, RANDOM HOUSE press, 20 E 57 Street NEW YORK, p. 56