Blog

Blog: A World of Robotaxis and (or?) Robodrivers


Early morning, time to start the day. A bipedal machine steps out of its charging receptacle and performs self-diagnostics on its visual sensors and decision-making capabilities, then heads downstairs. It gets into a classic car, one of the old models with pedals for acceleration and a manual steering wheel. It uses its mechanical legs to press on these pedals, steers to avoid walls, and pulls out of the garage to begin its drive. It has a human passenger to pick up, and it doesn’t want to be late.

Robodriver automates the human, not the car — can it be a cheaper and safer alternative?

This automated human is equipped with a full array of sensing technology, allowing it to process input at an extraordinarily rapid pace. It “sees” the road ahead, it knows where other cars are in relation to its own vehicle. It can follow road signs to the correct exit and can even account for weather delays. If a car suddenly swerves into its lane, the automated human can react quickly and appropriately. It isn’t very useful in other situations, but as far as driving is concerned, it’s pretty much an expert.

In another world, a biological human wakes up and gets ready for work. He brushes his teeth and combs his hair, puts on a nice shirt and goes downstairs. He opens an app on his phone and calls for a taxi service to pick him up at the curb. Less than a minute later his ride pulls up, a sleek little electric car that doesn’t require a human driver to perform any of its functions. He gets in, shuts the door, and the car hums silently onto the road.

The taxi is a fully autonomous vehicle equipped with an array of sensors, navigational technology, and networked intelligence. It knows where the road boundaries are, it can follow road signs to navigate to a destination, and it can react to unusual situations such as construction delays. It takes care of the driving so the human can sit back and relax.

Today, a person wakes up and gets ready for work. She eats breakfast and reads the news on her phone, then walks to the garage. She has a new car, one equipped with a number of advanced sensors and on-board computers, but she handles most of the control herself. She’s a human, which means she comes equipped with visual sensors backed by an incredibly powerful decision-making organ. The car has buttons and levers for its operation, all she needs to do is activate them at the right time and she can drive anywhere she wants.

What’s the difference between these three scenarios? Not much, as it turns out. In each one we want to transport a person from point A to point B. First, we use a human-shaped machine as a pilot. Next we tuck the same technology inside the vehicle itself. In the third we see what the first two are actually trying to accomplish: replicate a human driver’s capabilities. Take that effort out of the equation and it becomes apparent that vehicles themselves are ancillary to the whole project.

We speak of this technological evolution as automating the vehicle. Self-driving cars, not human-like-AI-piloted cars, right? But the shape of the technology shouldn’t dictate how we perceive it. We think of cars as relatively simple machines that have been around for over a century. Automating them is as easy as slapping on some sensors, writing code, and turning them loose, yeah? Because we’re replacing drivers, though, we should look at autonomous technology in a different light. It might sound easy to make a car self-driving, but what about a self-driving human? Not so simple, it turns out.

Let’s look at this from another angle. The driver with her modern car can operate nearly any vehicle in any part of the world. She can go on paved roads or dirt ones, drive in the rain or at night, navigate traffic jams or cruise the open freeway. To accomplish this all she needs are her innate capabilities paired with a valid license. Now apply this to the bipedal machine. Can it perfectly mimic these functions? Should it require certification before hitting the road? How about the robotaxi, can it do everything we can do, does it need a license?

Human driver can drive anywhere in the world — how about a fully autonomous vehicle? Will it need to be pre-programmed for every specific location?

It’s easy to get caught in the hype and think about self-driving cars as just another gadget. Making a computer that can think and act like a human, even in a restricted capacity, is not “just around the corner”, however. We are moving towards a future where the above scenarios are real possibilities. That future is a long way away, though, because automating a car is not the same as automating a driver.

It’s good that people are excited about automation. We are making progress towards it in small steps, things like driver assist and automatic braking. We can’t let that excitement cloud a realistic vision of the future, though. Gadgets are great, and technology is amazing. When it comes down to it, we create these things for other humans. Let the bipedal machines and robotaxis come, but only when they’re as good at their tasks as the best of us are.

Source: Artificial Intelligence on Medium

Leave a Reply

Your email address will not be published. Required fields are marked *

Back To Top