Hacker News new | ask | show | jobs
by HDBaseT 28 days ago
A human has to pass a test to be able to drive. A human (for the most part) doesn't just unknowingly drive into floods.

Why aren't we holding computers to AT LEAST the same expectation as humans.

3 comments

I love when people bring this up. When was the last time anyone here had to pass an actual driving test, like where you have to physically drive a car? For me, age 16, which was more years ago than I like to count.

How many licenses do we revoke for violating traffic laws? Getting into a car crash? Injuring someone in a car crash? Killing someone in a car crash? Not nearly enough! We are so lax about driving it's insane. But you want to hold these robot cars to some much higher standard? I mean, ok, but how much higher? It's a really freaking low bar right now

Less than 5 years ago, what's your point?

> "You want to hold these robot cars to some much higher standard"

No, I want to hold these cars to the same standard as humans. Humans very rarely, accidentally, drive into flooded water.

You have humans attempt to get through flooded roads all the time, that's another issue though.

Are you suggesting every DL holder knows all the driving conditions?

Quick, what should one do when the car starts drifting in ice? How about aqua planing?

If it is just taking a regular DL test, then waymo, Tesla and others would be driving all across the US by now. They already have a higher standard

Those are completely different scenario though.

We're talking about a piece of software being unable to determine if something is flooded or not. 99% of humans presented with the same visual ques would be able to determine its flooded.

This is what I'm waiting parity with. I don't see how that's an unfair ask, especially considering Atlanta floods fairly regularly, its not entirely uncommon.

> Quick, what should one do when the car starts drifting in ice? How about aqua planing?

If your car starts sliding, let off the gas, don’t hit the brakes, and countersteer into the direction of the slide to recover.

If you start hydroplaning you simply remove your foot from the gas pedal.

> Are you suggesting every DL holder knows all the driving conditions?

In my country at least: Yes.

Hydroplaning and driving on ice is part of the compulsory training, including driving on simulated ice on a special course.

That's great. Would you your country's test covers 100% of the situations a driver might encounter?

Even without knowing the details, I can confidently tell you they don't.

Does it teach you how to recover the car when the tires blow out? How about it is raining? How to react when a car is coming straight at you in the wrong way? How about when a dog jumps out?

Does Waymo know how to handle itself if the tires blow out?
A Waymo is already a dramatically safer driver than a human, and it isn’t even close.

There have been, and will continue to be, many cases drive into flood zones and die.

>A Waymo is already a dramatically safer driver than a human, and it isn’t even close.

Driving safe is not always about having faster reaction speed.

It's about many things, including reaction speed, visual awareness, specific expertise and informed decision making wrt braking or acceleration power. All of these are better in a modern self-driving car (I do not know whether Tesla falls into this category) than in a human.

https://waymo.com/safety/impact/

Over a given driving distance, compared to humans, Waymos produce a 90% reduction in serious injury, 90% reduction in pedestrian strikes, 83% reduction in airbag deployments, 85% reduction in cyclist strikes.

Reaction speed does not matter if the driver can anticipate things before they happen. Actually I think what makes a good driver is ones ability to anticipate.

I don't think computers are anywhere near a human in that regard.

>visual awareness

A point cloud and some computer vision is not "visual awareness". Your statistics is also biased is of its source.

But in very controlled environments and for sedentary pace of driving, yes, self driving cars could be better than average drivers.

80-90% reduction, over the course of 170 million miles driven on the famously very controlled city streets of LA, SF, Austin and Phoenix.

On average, I wouldn't expect the regulatory agencies to be very friendly toward outright fraudulent reporting from Waymo. On the very outside, maybe these 80-90% reductions are optimistic roundups from 50-65% reductions. Or do you believe that Waymo is secretly running people down and scooping corpses into their trunks?

What is a sedentary pace of driving?

Can you tell me if this 80% reduction is comparing waymo with human taxi drivers?