Comment by perihelions | Hacker News Reader

Comment by perihelions | Hacker News Reader

- "blocked a San Francisco ambulance from getting a pedestrian hit by a vehicle"

I'm taking no "position" on either "side" of the debate, but I feel it's noteworthy that a *human-driven* car killed a pedestrian (seriously?) here, and that fact is utterly banal and completely unremarkable to the authors—and to us. AI cars are still highly flawed and kill people. Human-driven cars are highly flawed, probably more flawed, and kill way more people. This is sort of a trolley problem, and we suck at solving those.

Because: incompetent killer humans are a diffuse force of nature; incompetent killer AI's are a centralized, attributable malefactor. We humans love identifiable malefactors, evildoers; love identifying and spotlighting and ostracizing them. It's a honed evolutionary trait. We've no atavistic instinct for the opposite, utilitarianism—for minimizing harm through "rational" (?) number-crunching, even where it conflicts with "obvious" (?) morality. Should we? Or is the other side of the trolley switch the correct one?

Because why the person needed an ambulance is completely immaterial to whether the ambulance was impeded a self-driving vehicle or not.

It could have just as easily been a heart attack and the questions raised about self-driving vehicles would be no different.

GP is criticizing the narrative focus. A robotaxi causing a new, less serious problem is the focus over a commonplace very serious problem caused by conventional cars.

Humans have always lived around danger; and now, for as long as anyone can remember, cars have been possibly the most significant source danger while also being integral in modern life. I would argue reacting to new problems is the rational thing to do.

Sure, cars are dangerous, but they enable a social structure that is far safer than what we had(hospitals, specialized surgeons, first responders, advanced pharmaceuticals, etc) to say nothing of productivity.

> Sure, cars are dangerous, but they enable a social structure that is far safer than what we had(hospitals, specialized surgeons, first responders, advanced pharmaceuticals, etc) to say nothing of productivity.

The next time I go to the Amsterdam, I'll be sure to see how many less hospitals they have.

But the issues with driven cars are talked about a lot. This is a situation where we have to evaluate if self driving cars should be allowed to test in the public commons. All other discussions aren't relevant to this discussion (but are still important to have)

But the article isn’t about human drivers hitting pedestrians. There is already much written about that.

It was blocked by a self-driving vehicle and an empty cop car.

No I think OP has a valid point. There is almost no news coverage about the absolutely massive problem of human drivers killing people constantly. It's become an actual crisis in America that's spiraling out of control.

Meanwhile the media only focuses on weird stuff like scooter accidents and self driving cars which is statistically irrelevant.

There are lots and lots and lots of human drivers, and very few self driving cars.

I think it's legitimate to find it unremarkable, as an anecdata, that one of very many human drivers hit a pedestrian, then one of very few autonomous cars blocked an ambulance - something so obviously bad that even a drunken driver could avoid it.

Kind of hard to assign liability to some matrix algebra that unfortunately did some computations that were suboptimal given the conditions. If these costs add up, it turns into a tragedy of the commons type of problem. A more modern form of the "corporations can pollute but don't pay the costs".

This is why automated driving is untenable, just like why LLMs are untenable with copyright. Capitalism is so awkward that it requires these contradictory legal frameworks to give an impression of rationality or naturalistic logic. The contradictions come to the foreground more with new technology, but they have always been there.

It’s ‘kind of hard’ to assign liability in all sorts of current situations, but we do. There is a whole industry that does just this.

This isn't even a "trolley problem" it sounds like this is just "fake news":

> "Throughout the entire duration the AV is stopped, traffic remains unblocked and flowing to the right of the AV," a Cruise spokesperson said in a statement. "The ambulance behind the AV had a clear path to pass the AV as other vehicles, including another ambulance, proceeded to do."

> The video captured by Cruise showed that the ambulance parked behind the Cruise and did not attempt to pass the robotaxi in the rightmost unblocked lane. Instead, responders moved a firetruck to allow the ambulance to pass on the left. The video, which Cruise declined to share publicly, indicates that 90 seconds elapsed between the patient being put on the stretcher and the ambulance leaving the scene.

This is also the easiest "trolley problem" in the world since it was the vehicle driven by the human that struck the pedestrian and gave them a life threatening injury. I know humans have a metric fuckton of perceptual biases, but I have no idea why that isn't the intellectual equivalent of stabbing-you-in-the-face-obvious.

If the story was centered on the driver of the vehicle and if they were some rich narcissist we wouldn't have any difficulty understanding that blaming the vehicle that blocked the ambulance for 90 seconds was blame shifting.

I think the fact that we as a society allow for cars to be driven at all, given the number of deaths that inevitably result, proves that there is at least some utilitarian instinct. Because of course someone could always argue that cars themselves are a "centralized, attributable malefactor" (similar to, say, cigarettes). For instance we could imagine a world in which the only cars allowed on the road were emergency vehicles. In that world, your point might be stronger.

And also, AI cars are on the roads, aren't they? Sure, not all of humanity is instantly going to be ok with them, but we'll probably get there (assuming the technology gets there).

This is an excellent point. There are two independent causes to this person's death.

1. A human driver grievously injured a pedestrian

2. An AI driver delayed medical care for the injured pedestrian

That person is potentially still alive if 2 doesn't happen. They are certainly still alive if 1 doesn't happen. Therefore, 1 deserves more of the blame here.

Considering these are two independent causes, we can still be upset that 2 happened and work to stop it from happening again. But we also shouldn't get so concerned with 2 that it causes us to increase the occurrences of 1 (e.g. by overzealously legislating away a future in which AI drivers are drastically safer than humans).

I disagree that 1 deserves more of the blame here—a blocked ambulance is blocked from attending to all medical emergencies, not just medical emergencies caused by (human-powered) car accidents.

If the robotaxi had blocked the ambulance from reaching someone who'd just had a stroke, blame would rest 100% on the robotaxi. And there's no telling that this won't happen again.

>If the robotaxi had blocked the ambulance from reaching someone who'd just had a stroke, blame would rest 100% on the robotaxi.

Yes, but that isn't what happened here. You are no longer assigning blame for a real incident. You are theorizing about the potential damage that AI drivers can cause in similar situations. But once you do that, you also have to consider the tradeoffs between human and AI drivers and all the deaths that human drivers cause. Otherwise, you aren't actually prioritizing the overall safety of our communities.

> You are no longer assigning blame for a real incident. You are theorizing about the potential damage that AI drivers can cause in similar situations.

I am! Because I think that's a useful thing to do.

> But once you do that, you also have to consider the tradeoffs between human and AI drivers and all the deaths that human drivers cause.

Also yes—and one factor right now is that we have a robust EMS system designed to respond to medical emergencies, including road accidents. If the near-future tradeoff is that robotaxis (1) coexist alongside human-piloted cars, rather than replacing them outright, which means road accidents will continue to happen, and (2) robotaxis hinder ambulances from reaching humans in need, including (but not limited to) humans who were struck by cars piloted by other humans, I think that's a poor tradeoff.

If it were possible to jump past the messy middle part and immediately get to the "all cars are autopiloted, except ambulances, which aren't hindered by autopiloted vehicles in any way," then of course that sounds great! But it's precisely the messy middle part that worries me.

For better or worse, human drivers aren't going away anytime soon. And you rightfully point out that human drivers are prone to accidents. Which is all the more reason that EMS services are critical in the near-term.

(But also, counterfactually speaking, it is absolutely the robotaxi's fault that the person died. Had the ambulance been unhindered, they would have had a chance at life after receiving proper medical care; since the ambulance was hindered, they didn't get that care and died as a result.)

If you are talking about these in general terms about a hypothetical future, I don't think one example of medical care being delayed is very strong data compared to the 42,939 motor vehicle deaths the IIHS reported in 2021.

>(But also, counterfactually speaking, it is absolutely the robotaxi's fault that the person died. Had the ambulance been unhindered, they would have had a chance at life after receiving proper medical care; since the ambulance was hindered, they didn't get that care and died as a result.)

I'll take the certainty of life for that pedestrian if they weren't hit by a human driver over the "chance at life" that would have come from quicker medical care.

But they were hit by a human driver. 42,939 people were struck and killed by human drivers. So in these present circumstances, delaying ambulance care (which is the immediate, not-hypothetical question here) is inexcusable.

The incident being discussed here is the blocking of an ambulance. The root cause for the ambulance being dispatched is immaterial to the argument.

The position you have outlined creates a second argument, and its being called out as a distraction. Its akin to "Children eat less fruit when the cupboard is full of sugar" and then someone countering "I don't like pasta, why aren't we discussing that?".

They seem related because they are about food preferences, but they are not because the original study was discussing the presence of sugary foods and children bearing on health, not your personal preferences as an adult. In this case the discussion is about an automatic vehicle failing to move aside for an ambulance, and it is not about why that ambulance was attending a job, or about what that job is. Changing the topic does a disservice to the issue.

The ambulance could have been blocked by normal well behaving traffic too. Seems like the optimal solution would be for most roads to have a separate bus/bike lane which normal traffic can not ever use, then in emergencies it can be used as a fast lane around traffic.

Equal blame to both.

Both of them caused death. One by causing the potential death and the other stopping the remediation of death.

The difference is you can sentence a person. A robot, not such.

>The difference is you can sentence a person to court. A robot, not such.

People generally don't end up facing much legal trouble when killing a pedestrian in an accident. It usually requires negligence. I don't know enough about this accident (or the Cruise vehicle's behavior) to say that anyone was negligent.

I don't think OP is talking about assigning blame in this particular case: The point is that a huge and arguably bigger problem (to the tune of ~40K deaths per year in the US) is buried in one sentence in the middle of the article.

The USA has kind of thrown up its collective hands and declared the danger of car accidents to be unsolvable. We treat it like an inevitable fact of life, now. The most the USA does, regulation-wise, are mild safety regulations on car manufacturers with tons of loopholes. There's relatively little penalty for drivers who accidentally injure or kill with their cars, and almost no liability on the car manufacturer. How does the saying go? If you want to kill someone and minimize your punishment, your best bet is to "accidentally" hit them with your car.

>I don't think OP is talking about assigning blame in this particular case

To be clear, I'm only talking about this specific case in these terms because the coincidence of it being a human driver which caused the medical emergency makes this a great example to highlight all the points you and OP are talking about. We are in agreement on all those points.

On a tangent of trolley like moral dilemmas, people tend to feel there's a moral difference between throwing a bomb onto a person versus throwing a person onto a bomb.

Will Cruise be taking centralized responsibility for the issues here? Will the person who caused the initial accident be held responsible?

I think it's very worthwhile to think of the overall improvements to safety while still acknowledging the arbitrage a corporation brings.

AI driven cars are not highly flawed and very, very rarely kill people. Humans driving cars are often reckless or intoxicated and kill many, many people every day.

Follow the money, politics, and known knowns and you’ll find the reasons on why articles like this are published.

News stories are about things that are new/novel -- that's why people read and discuss the news.

This is the thing I keep trying to remind people about when people complain about how much energy Bitcoin uses. Just think about how much energy is used by the U.S. government to maintain fiat.

It's wild that people want to focus the problems of completely unproven new technology whose adherents are betting on it taking over.

Reply on news.ycombinator.com