Hacker News Reader

Comment by cogman10 | Hacker News Reader

Yeah, I've brought this point up in other locations.

It does not matter that any autonomous driving tech is safer than human drivers. They MUST be perfect for the general public to accept them. The only accidents they'd be allowed to get into are ones that are beyond their control.

Algorithmic accidents, no matter how rare they are, won't be tolerated by the general public. Nobody will accept a self driving car running over a cat or rear ending a bus even if regular humans do that all day long.

The expectation for self driving cars is a perfectly attentive driver making correct decisions. Because, that's what you theoretically have. The computer's mind doesn't "wander" and it can't be distracted. There's no excuse for it to drive worse than the best human driver.

View on HN

Imagine if algorithmic accidents had biases. For example, let's say a car tended to crash into children (maybe they are harder to detect with cameras), more often than adults. This type of algorithmic bias would be unacceptable no matter how safe FSD were on aggregate.

So you're right, the only bar to reach is perfection (which is impossible), because algorithmic errors have biases that will likely deviate from human biases.

Call me an optimist, but I don't think it's impossible.

That said, there are going to be a lot of dead small animals due to autonomous vehicles. I'd hope that whoever develops the system has some good training data to stop it from hitting children.

The issue will be that it's going to be real hard to make a system that can tell the difference between a plastic bag and a poodle.

> let's say a car tended to crash into children (maybe they are harder to detect with cameras), more often than adults

This is already true today of human drivers because of the tall SUVs that are so popular. Do you think matching biases will be acceptable?

That's not true.

First, the vast majority of pedestrian deaths are adults. In 2018, a total of 206 age 15 or younger were killed by cars. Compare that to 5,965 killed who were age 16 or older.[1] Both in absolute numbers and relative to population, children are far less likely to be run over and killed than adults.

Second, while light trucks (vans, SUVs, & pickups) are 1.45x more deadly to pedestrians than cars, buses are far more dangerous than either. Motorcycles (which have excellent visibility) are particularly deadly to child pedestrians. From United States pedestrian fatality rates by vehicle type[2]:

> Compared with cars, the RR of killing a pedestrian per vehicle mile was 7.97 (95% CI 6.33 to 10.04) for buses; 1.93 (95% CI 1.30 to 2.86) for motorcycles; 1.45 (95% CI 1.37 to 1.55) for light trucks, and 0.96 (95% CI 0.79 to 1.18) for heavy trucks. Compared with cars, buses were 11.85 times (95% CI 6.07 to 23.12) and motorcycles were 3.77 times (95% CI 1.40 to 10.20) more likely per mile to kill children 0–14 years old. Buses were 16.70 times (95% CI 7.30 to 38.19) more likely to kill adults age 85 or older than were cars. The risk of killing a pedestrian per vehicle mile traveled in an urban area was 1.57 times (95% CI 1.47 to 1.67) the risk in a rural area.

All else equal, being hit by a larger vehicle does increase the risk of severe injury or death, but all else isn't equal. Larger vehicles tend to be more visible, louder, and slower than their smaller counterparts. Different types of vehicles are driven in different environments with different propensities for mingling with pedestrians. If vehicle mass and blind spots were the main factors in pedestrian deaths, we should have seen deaths skyrocket over the past 40 years (as cars got bigger and bulkier for greater passenger safety). Instead we saw pedestrian deaths decrease.

1. https://docs.google.com/spreadsheets/d/e/2PACX-1vRqGqodKkWkS...

2. https://injuryprevention.bmj.com/content/11/4/232

Would you accept it? Would you be ok if a car without a driver ran over your kid, even if they were playing in the local street?

I'd say, absolutely not. The only way we'd accept that is if the kid darted out before the vehicle could slow down, and even then we'd expect super human braking to (hopefully) avoid serious injury.

Also, self driving cars have and advantage that they can put cameras in places typical drivers eyes aren't. They should be able to see a lot more than you can from the driver's seat.

>They MUST be perfect for the general public to accept them

No they don't, its far from perfect right now yet its available and you can use it right now provided you have money to buy it.

Incorrect. What you can buy now is driver assist. I'm talking about actual "no driver at the wheel" driving.

AFAIK, the closest there is WAYMO, but they are geo-fenced. There's also some fixed route low speed buses out there. However, there's no self driving you can purchase which allows you to, for example, nap while the car is driving.

To get to the so called "no driver at the wheel" driving you have to go through what you called "driver assist" stage, its a continuous improvement.

So are you saying people accept it in the far from perfect driver stage now but won't accept it when it become so much improved "no driver at the wheel" stage ?

Correct.

I'm saying that the "no driver at the wheel" stage must be flawless. I've seen some claims that it's "Good enough if it's x times better than a human driver" or "can drive n number of miles without an accident". However, both of those are not the metrics to measure.

A "no driver at the wheel" package is good enough when it doesn't run over the neighbor's cat or Timmy on the street. Humans today make that mistake all the time, but that's not a mistake a driverless car can make and get away with. Consider, for example, Uber's killed pedestrian. Sure, it was a tough scenario and the driver wasn't paying attention. But the fact of the matter is everyone had the expectation that the Uber car would not hit that pedestrian even if they were really hard to see.

Until that happens, nobody will accept "no driver at the wheel" cars... except maybe in tightly controlled routes/geofences. Otherwise, SDC will require a driver to be attentive at the wheel at all times. That is, telsa's FSD is a very long way away from being able to hit that "telsa taxi service" that elon has pitched. I doubt it can make it with the current sensor set (not enough redundancy).

Well see, I personally accept and support far from perfect no driver at wheel car. I doubt I'm a minority.

You say that now, will you say that when (not if) a driverless car kills a kid or animal that a human wouldn't? Were you ok with Uber's killing?

Human kills already, and yet certainly do not want human to be banned for driving.

I too expect and accept that some accident caused by driverless car that kills kid/animal (that human wouldn't do) will happen (if not already happen). To expect otherwise will be unrealistic.

Reply on news.ycombinator.com