A recent study by the University of Michigan Transportation Research Institute found that 96.2 percent of people polled (actually only about 500 people) said they wanted “to have a steering wheel plus gas and brake pedals (or some other controls) available in completely self-driving vehicles.” You know what all those people have in common? They’re wrong.
In fact, pretty much everybody is wrong, except, in this case at least, Google. As much as I love to drive, I think when it comes to autonomous cars, the idea that a human must be present in the driver’s seat is a ridiculous requirement based on fear and a lack of trust.
Both feelings, of course, are absolutely understandable and have some pretty big repercussions, but they’re still fundamentally the wrong way to go. If we’re really going to with autonomous cars, we need to go all in.
Sure, there should be some sort of manual override controls for autonomous cars, if for no other reason so you can get your ass home when you realize you uploaded a virus to your car’s brain because you can’t stop looking at that weird porno site on your connected phone, but the idea that a human driver needs to be in place, at a full set of driving controls, ready to leap into action when the robo-car goes bonkers, I think is absurd.
And, for once, it looks like Google agrees with me, at least according to this New York Times interview with Sergei Brin:
Google engineers realized that asking a human passenger — who could be reading or daydreaming or even sleeping — to take over in an emergency won’t work.
I mean, just think about — with as much brutal honesty as you can muster — exactly how alert and ready to spring into action you’d be on a long, boring highway trip at night when the car is doing all the driving for you? I know how my reaction time would be, and aside from some very public screw-ups, I try to be a pretty alert driver: shitty.
It’s not just 500 randos who’ve never been in a robotic car that don’t feel comfortable taking the driver out of the equation — it’s car companies, too, even ones that pride themselves on being the technological vanguard, like Tesla. Here’s what they had to say about their self-driving features:
“We’re not getting rid of the pilot. This is about releasing the driver from tedious tasks so they can focus and provide better input.”
That concept right there — releasing the driver from the tedious stuff so they can provide ‘better input’ — I think is fundamentally flawed, because neither driving or people work that way.
Let’s take the driving part first. The idea that you can somehow remove big parts of the task of driving and then expect ‘better input’ seems absurd. How, exactly, would that work? A huge part of ‘providing input’ (or, as humans talk, ‘driving’) means that you have an overall, visceral feel for the car and its surroundings. When you’re driving, even during the “tedious tasks” you’re getting a constant stream of information about how the car feels on the road, how responsive the brakes and throttle and steering are, how the weight shifts and moves — all sorts of usually sub-conscious information, whether you’re aware of it or not.
Being removed from that stream of sensory data and then just coming in to provide ‘better input’ at choice, non-tedious times seems like a bunch of happy marketing horseshit.
And, of course, if you’re expected to react faster than the car’s computers for emergency or safety reasons, good fucking luck. In fact, if you are able to react faster than the car, my friend Agent Decker would like to see you so you can take this quick little Voight-Kampff test. It’ll just take a minute or two.
We’re not going to react faster than an autonomous car to dangers. Hell, I don’t even see why anyone’s expecting we’d react at all — in a moving car we’re not driving, anyone who says they’re not either asleep or playing on their phone or reading or daydreaming about their sex-jetpack is lying to you.
Once people trust that these autonomous cars work, the ramp to us completely giving in to laziness and letting the cars drive completely will be faster than an autonomous GT-R. I’m not saying we need to relinquish all control — a big, obvious ALL STOP button should be required for every autonomous car — but I think the act of slamming a mustard-covered hand down on a big red button is about the limit of what we can reliably count on a human passenger to do.
Autonomous cars are by no means infallible — they still get stymied by rain and snow and all sorts of other chaotic, natural things, so we’ll still have to take over sometimes. At least for the near future. But that could be handled by simply not letting a car enter autonomous mode if weather conditions are incompatible, or via other restrictions.
Of course, everything is much easier if we all pretend we need people behind that wheel — a recent Forbes article does a good job of outlying how much that one little rule will help keep the business of building and selling cars close to what it is today, but in this case I think Google’s more dramatic approach makes more sense.
I think going half-in on autonomous cars is just going to make cars that are terrible to actually drive and are hamstrung when it comes to their full autonomous potential. We need to accept that these won’t really be cars anymore — they’re transportation robots — and they can be designed without many of the current constraints we have on cars.
I’m not saying people need to stop driving, not by a long shot. I always plan to have a car that needs me to drive it (and, most likely, shift it, roll its windows, and probably push it sometimes) and one of my greatest fears is that widespread autonomous car acceptance will make things harder for those of us that do want to drive.
But pretending we’re needed behind the wheel at all times in a full autonomous car, like most states currently require, just doesn’t make sense. Eventually I want to send my dog to the vet alone in a car, or send my car to pick up gyros without me. Because if I’m going to give up driving (even boring utility driving), it better damn well be worth it.
Contact the author at firstname.lastname@example.org.