Self-driving cars. They are marketed by tech companies and car manufacturers alike as if they are the one true solution to eliminating human error on the road. And that’s completely fine, because these companies may very well be correct in that regard. However, the truth hurts a lot of the time, and this is one of those times. Self-driving cars are nowhere near ready for public use.

But before we get to the reasons behind that statement, we first need to understand what we’re really trying to aim for when we say “fully autonomous”, and that can be best described by these levels of automation:

How Does a Car Drive Itself?

Before we can even say that a particular car is self-driving, we need to define what exactly that entails. And to do that, there are five levels that represent the degree to which a car can self-regulate without human intervention. These levels are as follows:

Level 1 – No automation; driver performs all critical functions involved in driving such as steering, accelerating, braking, and much more. With no automation, the driver is required to give full attention to the task at hand in order to ensure safety.

Level 2 – The vehicle is able to assist the driver with all previous tasks. The car is equipped with sensory arrays which allow it to alert the driver of any potential dangers. A good example of this are parking sensors and blind spot sensors. While the driver may take attention off of certain tasks, the driver is still primarily in control of critical functions.

Level 3 – The vehicle is able to monitor the environment through the use of various sensor arrays like LiDAR. The driver’s attention is still critical at this level, but because the car is able to make small adjustments, the driver can now disengage from critical functions (like during highway driving) and allow the technology to control the car under safe and limited conditions. This is the category where safety features like cruise control and auto braking fall under, wherein the car is able to adjust speed depending on the data gathered.

Level 4 – The autonomous driving system would first notify the driver when conditions are safe to relinquish control to the car, and only then does the driver switch the vehicle into autonomous mode. Despite this, the car cannot determine between dynamic driving situations like traffic conditions, and a driver is still required for situations when the car is unable to react properly. Most self-driving car manufacturers are currently aiming for this level of autonomy.

Level 5 – This is the level at which absolutely no human intervention is required. The car is able to completely self-regulate and will function autonomously without the need for the driver to be engaged manually.

Google Waymo

While there are many car manufacturers that are working (racing, rather) to develop a car that has level 4 or even level 5 autonomy, Google Waymo is the forerunner in this endeavour. Waymo’s cars have travelled 10 million miles without a car wreck or even so much as an accident. A human driver would have had about 53 accidents in that distance. So, it would seem that self-driving cars should be just about ready to enter mass production.

But here’s where we hit a huge wall.

Human Capital (Data Gathering and Interpretation)

Google Waymo is only capable of functioning in perfect weather and it’s only being tested as a closed beta service in a small community. It is only able to operate at level 4 autonomy, therefore there are severe limitations in terms of the car’s reliance on how much data field personnel are able to gather. The car still cannot assimilate data on its own. It has to be fed the data in order for it to be able to become operational. And this is a lot of data we’re talking about.

You can liken this innovation to a baby who hasn’t figured out how to eat yet. While self-driving cars are, without a doubt, pieces of disruptive technology, it’s still a long way before we’re going to see them out on the roads anytime soon.