The tech industry has been big on pushing Autonomous Driving technology for the past few years. But how close are we really to seeing it in reality and having these self-driving cars in available to buy and to be on the streets of Toronto? In this article we’ll try and find out.
What’s The Definition?
Self Driving cars are already on the roads of Toronto, depending how you define Self driving anyway. Many new cars have small automated features like Highway control that could be stretched to say that they are self driving, however those applications are circumstantial and we’re not seeing any cars being built without a steering wheel yet. And certainly no automated driving around the city streets.
The definition of a self-driving car we’re using is a car with no human control required. You tell the car where to go and it gets you there. What’s known in the industry as “Level 5 Autonomy”. If that’s our benchmark, how far away are we?
The Next 5 Years
In 2016 the American Secretary for transportation boldly claimed that Self-Driving Cars would be on the road by 2021! A prediction that obviously didn’t amount to much. However, in the years since we often hear a similar line from Self-Driving enthusiasts, Self-Driving cars are always right around the corner, within the next 5 years minimum.
In 2020, Towards Data Science predicted we’d have Self Driving cars by 2025 and we’ve yet to see if that prediction holds but based on current progress it doesn’t seem likely to us.
Why does this keep happening? Why are such strong predictions being made that don’t come to pass? What exactly is the hold up?
The Hold Up
The answer to all of those questions is the same. Machine Learning. Relatively speaking AI research is still a brand new field and a lot of detail continues to be unknown. Driving is an extraordinarily complicated process that you trained for months to be able to do, but the training is only step one. You can train a machine to parallel park and follow the speed limit. What’s much harder is teaching the car spontaneity.
Driving requires hundreds of split second decisions every time you do it, but a human driver has instincts and knows to slow down when they see a pedestrian who looks like they’re about to run across the road or a car in front that changes lane too early.
There’s a million examples and that’s exactly the problem. It’s impossible to program a reaction for every single possible instance, the only realistic way to do it is for the machine to start learning how to deal with these situations independently.
This is the final hurdle, it’s why people keep claiming we’re so close to autonomy because we’ve been at this last hurdle for about a decade now. The cars are built, they have the sensors and they know how to drive on an empty road. They just need to be able to think.
It looks like we’re so close to self driving cars in Toronto, but as the years go on it’s clear that machine learning is going to be the majority of the problem when it comes to a self-driving future. And unfortunately that means we really have no idea when self-driving cars will become a reality. They are reliant on a breakthrough that hasn’t happened yet.