Sixth Episode of the Automated Podcast. Check out the podcast episodes at https://automatedpodcast.org/
So I’ve been really interested in autonomous vehicles for several years. Especially as they are one if not the most visible automation technologies to the ordinary modern person and they seem to be nearly ready for mainstream use in many parts of the world. 2 weeks ago I even went to a workshop in Brussels where a project was presenting the results of a 3 year study showing the impact of jobs of autonomous vehicles over the next 30 years. Though it specifically focused on the impact within Europe, the points mentioned are relevant for most parts of the world so I’ll talk about that later in this episode.
What is an autonomous vehicle?
Though I’m sure many of us think of cars when we talk of autonomous vehicles as this is the most publicized application, it’s actually a bit more broad and includes trucks, planes, boats, and trains. Artificial Intelligence (AI) is the key technology in the development of autonomous or self-driving vehicles enabling it to make its own decisions without human intervention through the continuous collection and analysis of data. As mentioned in the previous podcast episode on artificial intelligence, the massive amount of Data that we can collect nowadays is key to enabling AI as well as self driving vehicles. Real-time video object detection is a key component in autonomous driving systems. To enable this a ton of sensor types are used, the main ones used are: cameras, radar, lidar (eye safe laser based radar creating a 3d image of the surroundings). But there are also Inertial Measurement Units (IMUs), GPS, Vehicle-to-Everything (V2X) communication, and high definition maps. All the data from these sensors are then united together to generate an understanding of the environment, and then though an AI based system, react accordingly with actions like steering or applying thrust. Needless to say, this is the most challenging for a normal road vehicle that has to interact with dozens of other vehicles, pedestrians, signs, speed limits, and accidents. This is why higher levels of autonomy exist in the other transport modes. In case you didn’t know, commercial passenger planes are already 90% automated, and both fully automated boats and trains are already in use in parts of the world. I lived in Toulouse, France a few years ago and on the first day went into the metro where there was nobody driving the metro. Though this is perhaps one of the easiest to automate as it is a controlled environment with predefined stops and times. Much like the control systems that replaced elevator operators.
5 levels of autonomous vehicles. Autonomous vehicles are classified by six levels, from zero to five. I’ll have a chart in the shownotes that is used to depict the 5 levels, but for the purposes of the podcast, lv 5 is where there doesn’t even need to be a steering wheel, while lv 4 is able to do most actions autonomously but a driver still needs to pay attention to the road.
History of autonomous vehicles
In 1939, General Motors, held an exhibit with the first self-driving car. It was electric and guided by radio-controlled electromagnetic fields generated with magnetized metal spikes embedded in the roadway. By 1958, this became a reality. The car’s front end was embedded with sensors called pick-up coils that could detect the current flowing through a wire embedded in the road. The current could be manipulated to tell the vehicle to move the steering wheel left or right. But this is more akin to a highway system. Many attempts since then occurred across the world, but it wasn’t until the early 21st century that autonomous vehicles gained mainstream attention through the DARPA grand challenge. A 1 million prize would be awarded to a team that would be able to complete a 200+km driving course through the Mojave desert. Darpa is research department for the US military and was hoping to use the challenge to drive autonomous vehicles for the US military. The first challenge held in 2004, had 15 teams compete. 0 succeeded. However in 2005, 5 of the 21 teams completed the course, and since then improvements in the technology have led to millions of km driven by self driving cars. Google’s self-driving car, now developed by Waymo, is leading the charge with several million KM driven by their vehicles.
Most commercial flights have used autopilot systems for at least a decade that reduces a pilots actual control to around 10% of the flight, usually relegated to the landing and takeoff. Essentially we have had lv 4 autonomous planes for years. And with Boeing’s new autonomous fighter jet set to hit the skies in 2020, and German researchers recently built a control system for small aircraft, it’s not hard to see the age of the flesh and bone pilot, is coming to an end.
This is the same for both boats and trains. In May of 2019 an autonomous boat, The 12m-long Uncrewed vessel SEA-KIT Maxlimer, already made a commercial run from the UK to Belgium, carrying a box of oysters. It safely navigated what is one of the busiest shipping lanes in the world.
In Western Australia June of this year (2019), the Rio Tinto completed the transition to full automatic driverless operation of its entire heavy-haul rail system, making it the world’s first fully-automated mainline rail network.
Scope of maritime and rail industry
Though we are probably roughly connected to the scope of both car and plane usage, I’ll take a quick second to sketch the size of rail and maritime usage. While still a huge, highly profitable business, railroads move less freight than they did 10 years ago, and their biggest customer — the coal industry — is in long-term decline. Worldwide, Almost $700 billion in cargo moved by train in 2017, but trucks to put this in perspective trucks still carried over $12 trillion. But this is nothing compared to Maritime transport as over 90% of the world’s trade is carried by sea and it is, by far, the most cost-effective way to move goods and raw materials around the world.
Impact on jobs
The main concern with all forms of self driving vehicles is the number of people that directly depend on driving vehicles for subsistence. Worldwide, there are some estimated 6 million truck drivers, 18 million taxi drivers, under 1 million pilots, train/rail conductors, and boat captains. This number is roughly doubled when we talk about those indirectly employed (repair and maintenance, food and service, etc). Though the jobs have of course changed, they have existed for hundreds of years with not so drastic changes. A horse and buggy driver does need a different set of skills compared to a taxi driver but the essentially, steering, object recognition, geographical awareness etc, have not changed significantly and have enabled low skilled individuals an opportunity to make a living with minimal training requirements. If these primary jobs (drivers) are eliminated which seems to be the trend that we are moving towards then it seems that a more profound change is going to happen in job availability and skill requirements which will hit low skilled individuals more substantially. I’m not professing a solution to this but the workshop in Brussels that I attended did identify this as a main cause for concern. The workshop also identified the overall trend that low skilled jobs and specific tasks (like attaching a boat to a dock) will be replaced by automation solutions. Jobs that do remain will move more towards monitoring and supervision like remotely operating a boat or truck. The first autonomous truck actually hit the Nevada highways back in 2015 but with a back up driver still in the truck. What is interesting is that, just like the remote drone operators, it is possible to have remote truck drivers, operating perhaps as many as a dozen trucks, and only taking control of the wheel in the case of an emergency. New jobs will of course emerge, like automation and robotics experts, data scientists, and legal and privacy experts were some examples brought up in the workshop. However, it really is unclear as to whether the new and emerging jobs will be able to make up for the quantity of jobs that will be eliminated. One other question that I personally have is connected to the difference in the educational or training time required. Driving a taxi requires far less training time than a data scientist. So as these higher skilled jobs become more common, questions of educational support might have to be brought up, as there might be less jobs available when you take on a 4 or 7 year educational program. I was a projectionist for a movie theatre during my university days. I had to memorize the intricate loops that the film needed to take through the projector, but this was learned over the course of a week, and if you messed up, the entire movie could be stuck on a single frame until you re-looped the film. However, years later when I went back to the theatre I found out that the projectionist position no longer existed as the system was all digital and all someone had to do was press the ‘start’ button.
Speed of transition
Going back to episode 2 where I discussed the transition from horse and buggy to cars, it took nearly 30 years for the transition to happen, allowing people to shift from one career to another. Many horse carriage manufacturers were able to become car or car part manufacturers, and had their employees learn new skills to do so, the shift to metal parts wasn’t that big of a shift, and many of the old skills were still relevant to the new form of work. However, it is argued that this might not be the case today as autonomous vehicles require no driver, but do require data scientists, AI engineers, risk analysis consultants, lawyers to deal with new laws, which are radically different types of work. We also have an advanced distribution and logistics industry compared to 100 years ago so I’m a bit skeptical as to whether this transition will take 30 years once the technology starts being really implemented. But just because a technology exists doesn’t mean it will be implemented. Here in Barcelona we have had numerous taxi strikes that have successfully shut Uber out from the city. However, I’m not sure if there has ever been a successful long-term strike against a technology that provides better service, is more reliable, and is ultimately safer.. If you know of any let me know.
Random ending thoughts. Social issues for automation
Cargo boats and planes are probably the first ones to be automated as no human passengers — just cargo makes it easier to accept. Even though there used to be 3 pilots in every cockpit, the move from 3–2 was acceptable. But from 2 or 1 to 0 will be an order of magnitude more challenging for the public to swallow.
First unveiled in 2016 at the consumer electronics show — no commercial ones available yet, still in prototype or bogged down with legal and safety testing. Automated personal transport drone in the sky.
Legal issues. Who is at fault in an accident with a self-driving car? The driver, the car manufacturer, the software programmer? For a more in depth look at this I have a link for a paper in the shownotes
For next week I’m still undecided on either looking at the internet of things or blockchain first. Which one do you want to hear about first? Let me know over twitter or Linkedin.
As always go to Itunes or wherever you listen to your podcast and leave a rating and a review.