I don't fancy myself a predictor. I just muse, in a particularly earnest and relentless way, spiked by the clarity of 50 years of yoga and meditation and seasoned with my experience in seven distinct careers, coming up with ideas, theories, connections, perspectives, explanations, and, sometimes, predictions. To me it's all of a piece. I'm not aiming to be Nostradamus, I'm just trying to penetrate my own confused disorientation.
But here's one I got flat wrong: self driving cars seem to work.
My issue was always driver/pedestrian interaction. As a native New Yorker, I'm aware of the stalemate in the war between pedestrians and drivers. It's a virtuous stalemate, because if either had the upper hand, traffic would never flow.
Ideally each would simply use its allotted right-of-way (that's what the traffic lights and stop signs are for), but you can only go so far with legislation, restriction, and organization. People are sloppy, inattentive, selfish, and batshit crazy. So even if 90% diligently played by the rules, the remaining 10% is reigned in solely by fear of violent death.
Their own violent death, for pedestrians, and the death of some litigious stranger, for drivers. And there's nothing like a car sweeping through an intersection with verve to shut off the "maybe I'll try to cross real quick" impulse.
But self-driving cars will smash on their brakes, and all parties know it. This rational caution can be gamed and leveraged to a degree that cars driven by drunken crazy randos can't. Stalemate lost, pedestrians win, traffic no longer works. If one wave of an umbrella at an oncoming car makes it screech to a halt, everyone with an umbrella (or a baby carriage or an arm) will become the High God of Traffic and cross whenever they bloody well want to. And isn't that the pedestrian dream? Can't we sense (at least in east coast urban centers) that this is only tenuously restrained by a wholesome violent death fear?
The only solution would be to separate roadways from pedestrians (raise them, lower them, etc), but that would cost trillions in city centers. So I predicted self-driving would be impossible.
But that scenario turns out to be an edge case within an edge case.
First, "belligerent pedestrian encounters in East Coast urban centers" represent a tiny fraction of driving scenarios, overall (though it's not for nothing that Waymo only operates on the West Coast). Edge case!
Second, so long as there's still one single car out there being driven by a red-blooded human, pedestrians can't count on mercy. 10% self-driving, 50% self-driving, or even 95% self-driving means pedestrians must watch their asses or potentially die. We're far, far away from 100% self-driving, and by that point things might be different in any number of ways. So this is the edgiest of edge cases within the greater edge case, not the impassible hindrance I'd imagined.
There are other edge case considerations with self driving cars.
Watching this video of self-driving in Manhattan, I kept squirming as the car did things I wouldn't do. Like barrelling down 7th avenue at 30 mph, just inches from a long row of school buses parked to the right. I knew—but the algorithm doesn't know— that kids sometimes venture into the road from between buses to, say, chase a ball. So you slow down, you hug the left side of your lane to give extra room (or move left to another lane). You put your attention there, and if you can't, you slow down even more.
I've assumed I'm superior to algorithms because I entertain a wider range of scenarios. But two things occur:
1. Most other drivers (in fact, probably 90%) would blithely barrell down 7th avenue at 30 mph just inches from a long row of school buses parked to the right. They wouldn't imagine the edge case. So the algorithm is behaving perfectly normally within the bell curve.
2. In that same video, a bicycle whips by obliquely from out of the driver's blind spot, and the self-driving noticed and compensated. Me? I did not see him coming. I'd have had a much nearer-miss, and my startled over-compensation might have gotten me rear-ended even if I'd avoided the bike.
So my vast experience and diligent carefulness protect against extreme edge cases. But my human perceptual limitations make me vulnerable to common dangers.
That's why self-driving wins. I was wrong.
Their own violent death, for pedestrians, and the death of some litigious stranger, for drivers. And there's nothing like a car sweeping through an intersection with verve to shut off the "maybe I'll try to cross real quick" impulse.
But self-driving cars will smash on their brakes, and all parties know it. This rational caution can be gamed and leveraged to a degree that cars driven by drunken crazy randos can't. Stalemate lost, pedestrians win, traffic no longer works. If one wave of an umbrella at an oncoming car makes it screech to a halt, everyone with an umbrella (or a baby carriage or an arm) will become the High God of Traffic and cross whenever they bloody well want to. And isn't that the pedestrian dream? Can't we sense (at least in east coast urban centers) that this is only tenuously restrained by a wholesome violent death fear?
The only solution would be to separate roadways from pedestrians (raise them, lower them, etc), but that would cost trillions in city centers. So I predicted self-driving would be impossible.
But that scenario turns out to be an edge case within an edge case.
First, "belligerent pedestrian encounters in East Coast urban centers" represent a tiny fraction of driving scenarios, overall (though it's not for nothing that Waymo only operates on the West Coast). Edge case!
Second, so long as there's still one single car out there being driven by a red-blooded human, pedestrians can't count on mercy. 10% self-driving, 50% self-driving, or even 95% self-driving means pedestrians must watch their asses or potentially die. We're far, far away from 100% self-driving, and by that point things might be different in any number of ways. So this is the edgiest of edge cases within the greater edge case, not the impassible hindrance I'd imagined.
There are other edge case considerations with self driving cars.
Watching this video of self-driving in Manhattan, I kept squirming as the car did things I wouldn't do. Like barrelling down 7th avenue at 30 mph, just inches from a long row of school buses parked to the right. I knew—but the algorithm doesn't know— that kids sometimes venture into the road from between buses to, say, chase a ball. So you slow down, you hug the left side of your lane to give extra room (or move left to another lane). You put your attention there, and if you can't, you slow down even more.
I've assumed I'm superior to algorithms because I entertain a wider range of scenarios. But two things occur:
1. Most other drivers (in fact, probably 90%) would blithely barrell down 7th avenue at 30 mph just inches from a long row of school buses parked to the right. They wouldn't imagine the edge case. So the algorithm is behaving perfectly normally within the bell curve.
2. In that same video, a bicycle whips by obliquely from out of the driver's blind spot, and the self-driving noticed and compensated. Me? I did not see him coming. I'd have had a much nearer-miss, and my startled over-compensation might have gotten me rear-ended even if I'd avoided the bike.
So my vast experience and diligent carefulness protect against extreme edge cases. But my human perceptual limitations make me vulnerable to common dangers.
That's why self-driving wins. I was wrong.