AI development is moving ahead at a pace most industries are unready to adapt to, with looming promises of workers being replaced with deep-learning systems and automation with human oversight. Many public and private sectors are unready for these changes. Yet, as small a step as autonomous vehicles could lead to effects felt outside of the tech industry and may impact the world’s economy as a whole.
Self-driving vehicles and the AI era
While we may still be leaps and bounds away from streets filled with nothing but cars that drive themselves without any driver intervention, companies are already rolling out self-driving delivery pilots while companies like Tesla test their consumer-grade autonomous vehicle technologies.
As it stands, these vehicles still require a human being to be behind the wheel due to a combination of current laws and the need to have intelligent responses to situations that AI cannot currently react to appropriately. Autonomy, as it stands now, is less like self-driving vehicles and more like a form of cruise control; take a few steps beyond what is expected. Small startups are flocking to the AI sector to the point where we may be harboring an incorrect vision of the sector.
There’s little to no chance of seeing big companies unloading proprietary information for smaller companies to scavenge. Yet, some of those same smaller companies are taking an open-source approach to modernizing autonomy. For example, companies like Comma.ai already offer a sort of self-driving beta program where drivers can opt into assisted driving systems for certain makes and models of vehicles at no cost, with optional purchases for data tracking capabilities to help further Comma.ai’s programs.
Suppose the world at large opts into smaller programs without the necessity of purchasing expensive vehicles or driving assistance gadgets. In that case, the autonomous revolution could come earlier than anyone is ready to adapt to at a reasonable rate.
How do self-driving vehicles work?
In the past, organizations have adopted clever ways to describe self-driving vehicles. For example, Nissan used Star Wars as a medium to express its vehicles in collaboration with Lucasfilm.
Self-driving vehicles try to act as a human’s natural thinking ability and act rationally with the help of artificial intelligence systems. They make decisions based on their knowledge and by sensing the environment.
There are three key components to self-driving cars:
1. Radar – detect the direction and speed of the object
Radar is integral to many safety technologies such as automatic emergency braking, recording the speed of objects despite the weather, and adaptive cruise control.
This is why radar only records the direction and speed of objects as it fails to identify the object. For example, radar alone will not have the ability to distinguish a person or a bicycle from a car.
2. Cameras – for object identification
In keeping with human vision, self-driving cars use multiple cameras to recognize an object. They have a great resolution that provides a good definition of the surrounding object but only capture bright objects and detect weather.
As we mentioned above, radar doesn’t just recognize the object, so cameras are used in conjunction with machine learning. A vehicle can be designed to see pedestrians, lane lines, speed signs, etc., allowing it to respond to hazards.
3. Lidar – object position
An important final component is an autonomous vehicle in lidar, which automatically maps surrounding objects. Using millions of infrared light pulses per second, the leader scans the surroundings and calculates how long the reflection will take to reach it.
After obtaining different responses for wavelengths and response times, it produces a 3D representation.
Lidar creates a 3D representation of the environment by only sensing surrounding vehicles and does not have the same resolution as a camera.
Safety of autonomous vehicles
Sadly, autonomous vehicles are not as perfect as manually controlled vehicles and can cause accidents. However, self-driving vehicles have an attractive safety potential.
Incidents
Accidents involving self-driving cars are not uncommon, but not rare, with fatal incidents increasing since 2016. Their companies face problems during manufacturing, and Uber and Tesla in the US have had fatal accidents due to software problems.
The trolley problem
Along with dealing with issues like safety, self-driving cars also face the trolley problem. That’s a problem in which the vehicle has to choose between five or one death during a collision.
These moral decisions are complex and difficult, but manufacturers have used public decision-making in ML processes such as the MIT Moral Machine.
Defense potential
In 2018, 160,597 road traffic accidents were recorded in the UK, and the majority were attributed to driver error.
As a model driver, self-driving cars can help reduce this drawback. Because they detect their surroundings correctly, travel at a constant speed, and stay in the lane. When manually controlled vehicles and driverless cars share the road, unpredictable movements can cause problems.
The slow development realities
Autonomy isn’t coming at the sort of breakneck pace expected as little as a decade ago in what some analysts are referring to as the AI winter. But, like any new industry or technology, the initial wave of development and excitement surrounding those developments will wear off and show some of the more realistic projections that might be lost in the first wave of consumer excitement. For example, blockchain was once the fastest-growing technology globally, only to slow considerably as market uncertainty and future implementation concerns struck.
Autonomous vehicles have the distinct disadvantage of being a very transparent market to the average consumer. While not everyone may understand what goes into blockchain technology or even why AI works as it does, anyone can understand the ramifications of a self-driving car being involved in an auto accident. New legal precedents must be set, and the basis for road laws must be reworked in nearly every country. Even if the technology is ready, the world may not be.
With the potential to generate $800 billion in annual benefits when autonomy becomes the standard, one can only expect the world to catch up quickly. Nothing changes laws and regulatory fundamentals quite like the promise of profit.
We may still be several years off from a self-driving revolution. It’s not the sort of technology that will simply come and go or even quietly integrate into our lives without causing a ruckus, no matter how smoothly the transition period is. Expect industries to shift their focus toward autonomy and deep learning if they wish to remain relevant, and expect those changes to start happening soon.
By Andrej Kovacevic
Updated on 2nd August 2022