Hacker News Reader

Comment by spankalee | Hacker News Reader

We already accept 40,000 vehicle deaths a year, 7,500 pedestrian deaths a year, 1,100 by police, 260 by train, and almost 50,000 deaths from falls.

Nothing is completely safe, and it's just factually correct to say that there will be a fatal crash at some point that's the fault of an autonomous car (there actually have been already), but that Waymo's at least are already far safer than human drivers, and if that remains true, then society will "accept" it.

Also, the question was literally "Will society accept a death potentially caused by a robot?" which is not great, IMO. What does it mean for "society" to "accept" a death? There will be lawsuits, regulations, etc. Is the question whether self-driving cars will be banned everywhere after one fatality?

View on HN

A big difference is that you can't put a robot in jail. Even though the person driving the car is usually not harmed by hitting somebody on foot, it's still a life-or-death level situation, they might be looking at multiple years in prison, a large portion of their life gone.

The same even stakes are not there with robots killing humans. For one side it's a life-or-death situation, while for the other side it's profit margins and numbers. Companies are usually very happy to increase yearly profit for something as minor as a decimal percentage rise in human deaths spread across society, that's not even controversial.

Heck, not only can you not put a robot in jail, you can't even stop it from driving the next day as if nothing happened because it's duplicate running the exact same software and hardware is all over society.

I still think robot cars is a good thing though, because they will have a lot less accidents than us humans who love to drink and drive, or speed for no good reason. Still, it will raise some big important questions.

> A big difference is that you can't put a robot in jail

We rarely put human drivers in jail even when they're clearly at fault. We often don't even take their license away.

We have put CEOs in prison in the past. We could do so again. If a company really operates with blatant disregard for safety we should. Waymo's CEO makes it clear that she thinks their cars are better than normal humans, so long as that is really the case and she isn't ignoring issues she shouldn't go to prison for deaths their cars cost, but it is (or should be) an option if the company isn't careful.

The point of failure is probably not the CEO, though. They are rarely technical people directly supervising the taxis. If it can be proven that management skimped on quality, then, by all means, jail them. Otherwise, it becomes the fault of the people monitoring the systems.

At least in most parts of the US, hitting and killing a pedestrian does not usually result in jail time for the driver - unless the driver was driving under the influence or was driving recklessly. Most times their license isn't even taken away.

Older article that I remember but still remains true based on news reports I read https://revealnews.org/article/bay-area-drivers-who-kill-ped...

A friend of mine works in a bank. He said, "Did you know, the optimal amount of fraud is actually not zero? Because first of all, you can't get zero, and second, if you try, you screw up the whole bank."

The optimal amount of fraud is close to zero, but it's not zero. (He didn't say if they knew the actual number, but they probably do.)

I think the same logic applies to a lot of other things as well. You want to get as close to 0 as you can, without breaking everything else.

In real life tradeoffs always have to be made, and some of them are tough for the public to know about.

The classic case is road construction. Road design A will cost $120M and 3 people per decade will die. Design B costs $180M and kills 2 people per decade. Which one do you build?

The USDOT uses $13.7M as the "Valuation of a Statistical Life": https://www.transportation.gov/office-policy/transportation-...

> optimal amount of fraud

That sounds like the amount of fraud is the thing being optimized for.

> but that Waymo's at least are already far safer than human drivers

Is that really true? the only data I can find is published by Waymo or others who are obviously not independent of Waymo. Is there any independent data or investigators in this? Until then I need to be cautious. The big question is are they really safer, or are they just safer if you include a few outliers (alcohol) that among humans are significantly worse.

I'm not doubting the claim, and I've always though it is a matter of time before autonomous cars are statistically safer.

From what I can find, those 7500 are 7500 pedestrians killed by vehicle a year, so we might as well round it up de 50k vehicular death a year and be done with it.

Seems more honest.

Reply on news.ycombinator.com