Cyborg Drive

The way we interface with our cars is changing rapidly.

A few companies have already launched fully autonomous test cars with no user interface at all in some limited capacity. Meanwhile in the consumer auto market, the average level of automation in cars has noticeably increased. Functionality like lane-keeping assist, automatic emergency braking, and lane departure avoidance is now table stakes for a modern car. Beyond that, many cars you can buy today have levels of automation we’ve never seen in consumer products before.

At some point in the future, computers will drive almost all of the time, but even today it is rare to see a human driving a modern car without any help from a computer. Even if you have basic safety features like automatic emergency braking on your car there’s a computer watching your driving, ready to act if you slip up. If you have some of the most advanced vehicle automation available today, your car can potentially do entire drives without any human control if you ask your car nicely (and get lucky).

When you drive a Tesla, for example, the Autopilot software is running all the time and processing what it sees around the car to build a 3D model of the world. It uses that model and its knowledge of driving to help keep you safe. When a Tesla is driven manually, the human is primarily in control with the software acting as a backup to help avert a collision or other undesirable situation should the human make a mistake. When a Tesla is driven on Autopilot, that exact same software that was running in the background is now in primary control, and the human is the backup in case the software makes a mistake. Either way, both the human and the software are monitoring the environment and ready to act if the other makes a mistake. The only difference is whether the human chooses to take primary responsibility for controlling the vehicle or whether they choose to defer to the computer.

This is pretty interesting if you think about it. In any modern car, control of the vehicle will always be shared between the human driver and the car’s software. The human driver will be able to switch between having primary control and giving the computer primary control, and they will also be able to configure which active safety features are enabled while they’re driving manually. Long term, the amount of time the car spends in primary control of the vehicle will increase and increase until humans are rarely ever in control at all.

Rather than modeling consumer usage of automated vehicles as a step change from the day autonomy is “solved”, it may make sense to think about it as an exponential function with t = 0 being today. A key takeaway here is that cars will start to be judged much more heavily on their software than the manual driving experience.

Leave a Reply