That fucker isn’t going to spend a day in jail. He should be in prison for at least a decade. You don’t accidentally go 74 miles an hour down an off ramp.
I should restate I suppose: Criminal law determined each life ended was only worth $11,500 each in this instance. This precedent determines you can drive your Tesla 74mph into another vehicle if you can pay a $11,500 fine for each random person you slaughter.
The incentives in the United States strongly favor pedestrian fatalities and reckless driving. It is insane that killing two people in a car accident is treated with a fine instead of ANY jail time. There’s a reason why fatalities on roadways continue to increase in the US. This is one of them.
> insane that killing two people in a car accident is treated with a fine instead of ANY jail time
Which component of justice— retribution, incapacitation, deterrence, rehabilitation and reparation [1]—is served by jailing this person?
The only one I see is retribution. Given the families are pursuing damages, the reparation prerogative is better served with him out of jail (and not draining his assets fighting prison time).
What was this guy doing while the vehicle he was supposed to be controlling ran a red light at more than twice the posted speed limit? The article is lacking on details here.
If he was, say, watching a movie on his phone, then he should go to jail to deter others from watching movies on their phones.
Already failed to do due diligence and ensure system he was using was fully safe. Do you think removal of license would prevent anything? As far as I know not all cars yet have license readers before starting.
Removing a negligent driver’s license turns their next speeding ticket or red light camera into driving on a suspended or revoked license, which in some places carries the possibility of jail time.
So then all crimes should carry the death penalty for maximum deterrence? Or deterrence is impossible and therefore there should be no crimes? What point are you making here?
That’s incapacitation. Depending on the driver’s record and whether he was negligently delegating control to the car (e.g. taking a nap), this prerogative would be variably satisfied. Facts and circumstances matter; this is why we have judges versus mobs.
Retribution and deterrence certainly would work with reasonable penalty let's say 10 years per murdered person. Also that sort of sentence would work for incapacitation. As clearly 23k is not high enough to stop many people from repeating this crime...
> clearly 23k is not high enough to stop many people from repeating this crime
That’s not the penalty. It’s a criminal conviction, the fine and probation. I don’t know the terms of the probation, but I assume it means he goes to jail for traffic infractions. Also, he’s going to pay a lot more than $23k given the civil case is ongoing.
Don't you find 23k fine insanely low for in essence murdering two people? To me it is laughable amount. This sort of crime should have such a high fine that everyone including billionaires will be deterred to repeat it.
Even the language around automobiles is corrupted. Essentially all "accidents" aren't.
Owning and operating a multi-ton machine capable of travelling at high speed and rapidly accelerating is a huge responsibility. Being distracted isn't an accident, it's negligence. Not mechanically maintaining the vehicle is negligence. Not supervising your vehicle, even when you have some sort of digital aid enabled, is negligence. None of these are accidents.
Calling a system that is not fully self-driving "autopilot" while knowing exactly what the general public considers that term to mean is....something worse than negligence.
I don't understand what you're arguing at this point.
Given the rates of vehicle deaths in the US, the penalties are clearly not high enough to act as a deterrent.
The "default" for this charge absolutely involves at least a year of jail time (that is, in fact, the defining difference between a felony and a misdemeanor), and the plea deal in this case is an affront to justice.
How often do people actually pay civil damages from these sorts of cases? It seems like there is a whole cottage industry around finding loopholes to avoid paying.
I think drivers are fine with causing a certain amount of risk to pedestrians. Strong, predictable penalties would lower the typical risk level drivers were willing to subject pedestrians to.
It’s not binary, it’s a distribution and I do believe you can alter that distribution.
Deterrence: it would make a lot of inattentive drivers take driving a bit more seriously if jail time could be the result of irresponsibility behind the wheel.
Or give those who negligently kill with their cars a choice: serve the jail time for your negligent homicide, or give up your license for life. That would still have a deterrent effect while not being life-ruining (just life-inconveniencing)
Except that studies show over and over that increasing punishments doesn't work as a deterrence method. Increasing the likelihood of getting caught/facing punishment is far more effective than increasing the punishment. Unfortunately, in this case, you'd have to target behaviors contributing to pedestrian fatalities rather than the fatalities themselves since that's so infrequent and unlikely for a given driver to adjust their behavior.
Forget jail time, why can’t we take away driver’s licenses from people who’ve proven they’re a danger to everyone around them? So often when we see fatalities it’ll turn out that the perp had tens of thousands in unpaid tickets.
There will come a day when cars truly drive themselves and still get into accidents, and I’m willing to bet that car manufacturers will somehow have zero liability. They should be held accountable just like any driver would.
Note: the criminal penalties are a conviction, fine and probation. The victims’ “families have separately filed civil lawsuits against Aziz Riad and Tesla that are ongoing.”
The civil lawsuit cannot give suspended prison sentence i guess. Who does this person know? Or is it usual in the US? In my country, a killer will always get time, even if its "only" negligence: the time will often get suspended until the next infraction. This end after 20 years, so at least idiots like this will maybe be careful about other people lives.
France. It's not really deterrence (deterrence is mostly about other people not killing too, no?). This is taking into account that mistakes were made, and that maybe the person can improve and not be an idiot on the road.
If the person can't improve, and is caught reckless driving or drunk driving (or worse, killing/hurting someone else), then the 'suspended' sentence transform into a full sentence, to protect other people from the idiot.
I don't believe that partially automated driving, as shown by Autopilot is fit for human use. People will rely on it as if it's fully automated and more accidents like this will happen.
We see this same problem with face recognition software that says an innocent man's face matches a criminal. The cops just accept what the computer says and arrest people because the computer says it's a match.
We can't have the machine do stuff for us unless it's superior to human judgment, or the stakes are much lower than ruining lives.
> I don't believe that partially automated driving, as shown by Autopilot is fit for human use. People will rely on it as if it's fully automated and more accidents like this will happen.
Keep in mind 125 people are killed on the roads every single day in the US.
If all vehicles magically switched to this system overnight and the system only killed 100 people per day, that would be a huge win.
The fact some people will be killed by automated driving systems is not enough to ban them. If it was, we would ban all cars.
The only important statistic is if those systems kill less people than the current state of affairs.
Continuing your analysis - vehicles magically switched to this system overnight and the system only killed 100 people per day. Huge win. But then you find out that the drivers of those 100 daily accidents were all browsing their phone when the accidents happened. Would we ban phone usage while using autopilot? And if we did ban phone usage while using autopilot, and someone was using their phone… should they go to jail?
I expect the 125 people who are killed today have a high factor of alcohol being involved, and our society doesn’t take drunk driving lightly. Those drivers would certainly go to jail. Why would phone use be any more accepted if the death rate suddenly dropped to 124, or, as in your hypothetical situation, 100?
What? Phone usage is already banned while driving, period. Also, along with other Autopilot distractions, it is also detected by the cabin camera and it can kick you from using it.
So I guess the answer to your questions are yes and yes?
If the goal is full autonomy then people using their phones while the automobile is driving have no impact whatsoever on the accident when that goal is reached, if laws banning distraction suffice for you to solve problems then your goal is not full autonomy, in which case what's the point of any of it? When full autonomy is reached, I expect that 100% of accidents will occur when passengers are busy letting the car do the driving, so that correlation would be meaningless.
I used to agree with this, I no longer do. There was an interesting read on HN yesterday coincidentally that articulates a part of my reasoning, https://www.devever.net/~hl/ruthlessness
The difference is that machines have no agency. Imagine a scenario where on roads with 100% human driving, the most of the people that are killed are PCP addled people running naked across freeways where they aren't expected to be, but with self driving you have 50% grandmas walking with babies in a stroller across crosswalks. We don't have to say that one life is more valuable than another to come to the conclusion that the latter is worse, all we have to do is acknowledge the pedestrian's portion of the responsibility for the outcome.
With that in mind, your last sentence is false, the aggregate death count is not all that matters, what matters is the situations in which they occur.
Now let's think about the link in this comment. We begin to form the rules of crossing a street around the behaviors of machines. Not unreasonable. But then it's just a small step from there to "you should've known how the machine was going to behave because it's behavior is simple and predictable." We wind up in a situation where we absolve the self driving car of all liability when an accident occurs because the only entity involved with agency was the pedestrian. We do this already with train tracks.
So with that in mind, the machine systems killing less people is not enough, it has to be so good that the only deaths are suicides, unbelievably bad behavior by those killed, or freak accidents so unlikely that they genuinely shock everyone who learns of them. That's the only way to transition from humans to software operating the machines and be able to justify the ruthlessness with which machines operate themselves.
> Imagine a scenario where on roads with 100% human driving, the most of the people that are killed are PCP addled people running naked across freeways where they aren't expected to be
I don't think it's helpful to introduce this scenario that is entirely not true in the real world and therefore doesn't teach us anything. You're saying that a huge number of road deaths are people that were asking for it anyway.
Dozens and dozens of people were killed on the road today who were not drunk, who were not on drugs and were not breaking a single road rule.
This is the reality of the world. Good people are being killed every day by the way things are now.
If we can do anything to reduce that number, we should.
...
Another way I like to think about this is from the ad in Victoria where a guy has to pick which of his family members will be killed on the roads [1]. Obviously none of us want any of our family to be killed, we all choose zero.
It's a thought experiment. A spherical cow that demonstrates that simply measuring the aggregate death count is lossy, it assumes that the proportion of pedestrian deaths where the pedestrian is at fault remains the same in both situations which is not true, so I draw a picture which makes that glaring distinction. We need to know the details of these deaths to know which scenario is preferable. Here, reverse my scenario, the autonomous vehicles only ever kill people who jump out in front of them. That's preferable, no? Then why is the inverse not worth addressing?
Current situation: 125 people are killed every single day
Future autonomous situation: 100 people are killed every single day
I don't care who they are, what they were doing or why they were killed.
The details of their deaths are no important. Saving 25 lives per day is clearly the right thing to do.
Hint: Imagine all of these people are your family and friends. Do you want to keep having 125 of them die every day, or only 100? Does it matter if they're being run over as pedestrians or slammed into by drunk drivers or simply sliding off the road on ice? Of course it doesn't, saving 25 of them every day is the only thing that matters.
> The details of their deaths are no important. Saving 25 lives per day is clearly the right thing to do.
I think it's clear that if we were talking about saving 125 people 90 years old or older or 100 children, the clear choice is to save the children, from almost any rational point of view. Even just from a societal health or utility standpoint, children are worth more than people that are statistically close to death anyway at a societal level.
If that premise is acceptable, and if we accept that switching to self driving might change the demographics of who dies in accidents, then it's important to quantify not just whether it saves more people or not, but what people, in case that shifts matters.
> Hint: Imagine all of these people are your family and friends. Do you want to keep having 125 of them die every day, or only 100?
I would rather lose both my parents than one of my children. I don't know a single parent that would feel otherwise.
Additionally, there's also the possibility that like adding extra lanes of traffic to a freeway can make traffic jams worse, self driving cars could increase deaths per capita while reducing deaths per trip, by promoting a lot more car travel (by either allowing current non drivers to drive or increasing time current drivers spend in the car).
> I think it's clear that if we were talking about saving 125 people 90 years old or older or 100 children, the clear choice is to save the children, from almost any rational point of view. Even just from a societal health or utility standpoint, children are worth more than people that are statistically close to death anyway at a societal level.
That assumes that the people currently being killed on the roads are "less valuable" than the people that would be killed on the roads by autonomous systems.
There is no evidence to prove that, it's purely hypothetical.
Yes, there is no evidence. The point I was trying to make is that gathering that evidence and knowing what is happening is important. It's not enough to just assume any drop in deaths is automatically better than previously, when what we're talking about is a drastic change in how the overall system operates.
Im not arguing that autonomous cars will be worse, I'm just arguing there is the possibility for them to be worse in non-obvious ways, and you prior arguments seemed to ignore that. We should not be so blinded by a single metric that we fail to understand what's really going on. To are many cases of that in technological advances, legislative change, and societal shifts, that we aren't aware of until long after the additional damage has been done.
That's an absurd concern. Children are already the most at-risk on our roads right now, with humans behind the wheel. We're already living your nightmare. The rise of autonomous vehicles isn't going to increase the likelihood of basketballs accidentally bouncing into the street. Nor will it cause a wild swing in the demographic mix of people using pedestrian crossings.
By the time they're widely deployed, autonomous vehicles aren't going to be a little bit safer than the average human driver. They will have to be superior by at least one order of magnitude. As in, the number of deaths per million miles at least 10X lower than the human average.
I think one order of magnitude is the least we should demand. I would consider it a rather low bar. I suspect that it'd need to be more like two orders of magnitude in order to retain sufficient community consent.
So if that's the threshold of acceptability, then we'd be talking about slashing deaths by 90% or possibly even 99%. Either way, autonomous vehicles could kill children exclusively and they'd still be safer around children than human drivers.
> I think one order of magnitude is the least we should demand. I would consider it a rather low bar. I suspect that it'd need to be more like two orders of magnitude in order to retain sufficient community consent.
So right now we have 125 deaths every single day on the roads in the US. Same today. Same tomorrow. 45,625 every single year. Year after year.
You're saying it would need to be < 12.5 per day to be acceptable, but really we should aim for < ~2 per day.
So that means if we invent some system that reduces the number of deaths PER DAY from 125 down to, let's say, 30, you think we SHOULD NOT implement it, because that's not enough of an improvement?
Let's say that system is the best we can ever make, but we decide to never turn it on, because it's not a whole order of magnitude better. You're choosing to kill an additional 34,675 people per year, because a reduction in deaths from 125 per day down to 30 per day is "not good enough".
I’m saying that 10x should be a low expectation for a system that doesn’t get distracted, doesn’t get drunk, doesn’t get drowsy, doesn’t get angry, doesn’t get a bruised ego, doesn’t have a TikTok account, and can look in every direction at every moment.
But if the best we can do is a 9x improvement of course we should still deploy it. The question is whether society will accept it. I’m not sure it would. I’m not sure it would accept a system that reduced deaths by two orders of magnitude, or 100x.
Imagine for a moment, Tesla fully cracks the FSD nut, and every car on the road is replaced with a self driving Tesla, 100 times safer than humans. “One person is dying at the hands of the pedestrian butcher Elon Musk every single day!” the headlines will cry.
It will still matter. Unpopular new laws aren't guaranteed to survive swings of political power. Local/state governments could permit it, and a politically motivated national/federal government could amend the manslaughter laws in a way that would make autonomous vehicles untenable.
(It's sad that society is obsessed with politicising everything and everyone. I don't agree with some of Musk's political views, but so what? I never expected to be in perfect alignment with him or anyone else anyway. And it would be the height of arrogance to assume that my views are consistently better than his, rather than a different order of priorities resulting in a different set of trade-offs.)
Nobody is choosing to kill anybody, we are talking about ideal scenarios, you should quit trying to moralize the discussion to impune those you disagree with. Address the points they're making rather than harping on the fact that they don't agree with yours.
I think you misunderstood the point of my comment.
If indeed it is an order of magnitude less deaths, then it probably doesn't matter if the demographics shift, but we actually have to see that change and know the demographics to actually know that. I'm just saying that we should pay attention, as we should with anything that has the capability to change per-capita figures for adverse actions, as to what is actually happening and not assume.
That said, since we're already seeing an increase in these cars and the use of their autonomous options which match "self-driving" to greater or lesser degrees, I doubt we're going to see a 10x decrease before they are inthe wild, because to some degree they already are.
I just want us to know what that means by looking at as much info as we can rather than a single metric.
While I agree with you, my hypothetical scenario above was deliberately constructed to avoid ascribing worth to distinct people, and only measuring how much of the outcome was the result of their negligent or deliberate behavior. We can come to the same conclusions while still assuming all lives have equal worth. All we have to do is determine where the fault lies. If a person jumps out in front of my car because they want to die, that's not my fault. If I run a stop sign and t bone a person it is. The same if I'm a computer in the trunk of a Tesla, and we want to be sure that if we are going to construct an environment where machines call the shots that they're less dangerous, so we should be sure that at the very least they're causing less accidents than people, not simply that they're participating in less accidents. Just the number of deaths doesn't give us this measure, it doesn't actually tell us how safe the machine is. What we ultimately want is primarily that the machine is less often at fault than a human driver, secondarily but very importantly that the machine avoids harming others even when those others are negligent and at fault, and also can avoid harming people in situations deliberately caused by others, particularly if those harmed weren't those causing the situation.
Effectively it's a version of the trolley problem. Do you do nothing and 125 people die per day or do you pull a lever and 100 people die per day? Would you pull that lever?
I would want to understand the differences in who it would kill. If not pulling the lever, 70 of those are suicides, and if pulling it only 10 are, you're actually killing more people by pulling the lever, and potentially subjecting unwilling participants to continued suffering.
But my main point is that it is a machine pulling the lever or not. Once you deploy such a machine we are all subject to it's cold calculation. We should go the extra mile to make sure the world we live in winds up being better and not just assume less deaths = better outcome.
To be clear, my use of 125 and 100 is only to mirror the post I was replying to. Expressed as a trolley problem I would choose that trade-off. As a real world proposition no society would, and certainly no autonomous vehicle operator would — for precisely the reason you state. We don't like robots which kill people.
As I've said elsewhere, I don't expect any autonomous driving system to be anything less than 10X safer than human drivers, or one order of magnitude. And I don't think there could ever be a durable social warrant for autonomous vehicles unless they're at least 100X safer, or two orders of magnitude.[0]
Given a minimum expectation of being 10X safer, and given that drivers don't get to choose the demographics of people who place themselves in harm's way near our roads, it is statistically unfathomable that autonomous vehicles could ever be net more dangerous for any demographic.
--
[0] If anything I feel that's too conservative and three orders of magnitude might be closer to the threshold at which societies would tolerate autonomous cars. Road deaths in the USA are in the order of 40,000 annually. Three orders of magnitude still means 40 deaths per year. That's plenty enough to keep the press aroused and erect.
You can reject my premise, but it is sound. If self driving cars only ever killed people that jumped in front of them deliberately, that's a resounding success, that means they never accidentally kill anyone. So if self driving cars swerve more efficiently to avoid hitting suicides but are worse at stopping at a crosswalk for children, thats an abysmal failure, they accidentally kill people more than humans do. Your take - that any amount of decrease in death is sufficient - is excessively reductionist because it doesn't actually measure how much safer the car is than a human driver. For it to be a reliable measure we have to accept the premise that the same proportion of pedestrian deaths will be due to deliberate action or negligence of the pedestrian, an unfounded premise, much more absurd than mine, and one that ignores the ruthlessness with which machines execute their instructions. Your hint attempting to emotionally sensationalize your argument does nothing to make it more convincing.
Autopilot. This automatically accelerates, decelerates, and steers on roads. This works on all roads with lane markings, not just limited access freeways. It also obeys speed limit signs.
Enhanced autopilot. This is autopilot but will also self drive from on-ramp to off-ramp, changing lanes to pass (and exit the passing lane), taking interchanges to other freeways, and exiting the offramp.
Full self driving beta. This will drive on pretty much any street. Early on it was purely a gimmick that required constant intervention, but lately it’s gotten pretty good. My main complaint is that it drives too conservatively, causing impatient drivers to behave erratically around you. Imagine someone driving as if they're taking a driving test. Because of that I tend to drive manually unless I’m on a freeway.
> Most new cars are sold with features roughly in parity with Autopilot.
Really? My understanding is that many new vehicles have lane keep assist and dynamic cruise control, but that they are much more limited than Autopilot. For example, lots of cars have much tighter limits on the curves they will follow. They also don't have the UI that Autopilot has, which shows what cars it detects. This gives more of a sense of full autonomy (although I find it a bit creepy/worrisome when vehicles appear and disappear on the screen for no reason — one would think this would give people pause).
There are definitely more expensive vehicles (Cadillac, Mercedes) that have more robust systems, but my sense is that your typical Toyota, Subaru, etc. isn't close to Autopilot.
Autopilot really is just adaptive cruise control and lane centering. It can't change lanes by itself. It can't take an exit. It won't stop at an intersection. It can't turn corners. These things are blindingly obvious within a few minutes of use. Yes it's more slick than the cruise assist features on your typical Toyota, Subaru etc, but not dramatically so.
Some people argue that Tesla customers think they've been sold autonomy. I can't see how. Tesla goes to pains to make you aware of the $12k you would need to spend to get the system that "self drives". You are painfully aware of the $12k you didn't spend.
What’s confusing to me is I consider “autopilot” to be what happens when I press down on the stalk once, which doesn’t do lane centering and only enables adaptive cruise control. Reading the current marketing descriptions of the features (https://www.autopilotreview.com/tesla-autopilot-vs-enhanced-... Autopilot – this was,and Navigate-on-Autopilot) doesn’t help. If you don’t have FSD beta purchased or enhanced autopilot purchased, does pressing down on the stalk twice enable lane centering?
I wouldn't say "most". I've tested Toyota and Nissan, and they can't even drive on a normal road and are barely useful on a motorway. Every time there is a merge lane on the right they will both creep into that lane once the lane markings disappear. Nissan's ProPilot can't handle a regular meandering road, at the first sign of a tight corner it will give up, but not before pretending that everything is fine leaving you with 2 seconds to stop it from driving off the road. I think that Tesla is about 20 years ahead of Toyota and 50 years ahead of Nissan.
I would like the NHTSA to gather and publish robust data on:
• Crashes per mile on each manufacturer's "self-driving" systems
• Crashes per mile on each manufacturer's LKAS (lane keeping assistant) systems
• Crashes per mile on each manufacturer's TACC (traffic aware cruise control)
• Crashes per mile on each manufacturer's regular cruise control
• Crashes per mile under manual control
Break each out by highway, suburban, city, and car model year. Also include whether the crash resulted in injuries and deaths, and whether those deaths were occupants or not.
Then we can actually determine where the problem is, and the scope of it.
I suspect age correlates with vehicle choice and know it correlates with accident rates, so I'm not sure those figures would tell you what you think they're telling you.
I see you claiming it's not difficult, but I don't see you making it happen. If it's so simple, take your developed plan to the authorities and make it happen. Show them what data they need to collect and how they can force manufacturers to provide it.
> Then we can actually determine where the problem is, and the scope of it.
Unpopular opinion: there probably isn't a problem? None of these system is going to be perfectly safe, but there's quite a bit of evidence at this point that they're all safer, as actually used (which, yes, includes some misuse) than a human driver is at doing the same task.
No one wants to hear that though, so instead we take "where the problem is" as a prior and continue arguments like this ad nauseum.
Full disclosure: my car is going to drive me to Idaho in the morning.
There's plenty of data already though (of which a very clear LACK of signal should be a pretty obvious one). People, likely you included, discount that stuff because it doesn't confirm their priors. And so they demand new data that they believe will. And if it doesn't? Goalposts are movable for a reason.
At the end of the day you either believe or you don't. But if there was an actual problem, we'd know. There are millions of these things on the street now.
You're elsewhere in this subthread dismissing existing evidence (insurance rates) because it doesn't confirm your priors. I think we both know what would happen if this study magically appeared. "Demanding" data is a cheap evasion. The requirement for rational discourse is accepting the data you have.
lmao, nobody presented evidence based on insurance rates. I just suggested there are many other compounding factors that make insurance rates difficult to use for this purpose.
It's also absolutely hilarious to me you're assuming I think Autopilot is unsafe.
You could look at liability insurance rates. If the same driver has a lower insurance rate for a car with a certain automated system, then that system probably makes driving safer.
Seriously. Tesla loyalists will say there’s net fewer accidents because of Autopilot, and maybe there are, but it’s still Tesla’s responsibility to prevent the system from being misused or entering dangerous situations, whether through negligence or ignorance. Hopefully the recent recall was a wake up call.
I'm not sure what this has to do with what I said. I want to know if there actually are fewer accidents because of Autopilot versus other situations. People are suspect of Tesla's own published data, but we can just easily answer it. For some reason, nobody advocates for finding out what's happening, just complaining one way or the other.
More numbers would be great they always parrot Teslas statistics comparing the number of miles per accident that people felt comfortable enabling AP (easiest highway miles, more comparable to when other vehicles have cruise control enabled) vs all miles.
This is a silly article. The numbers Tesla releases themselves say "we count any crash in which Autopilot was deactivated within 5 seconds before impact", so Autopilot aborting less than one second before a crash would still be counted as an Autopilot crash. (I think this happens because AEB takes over.) But sure, this should obviously be accounted for in some well-run government study.
It’s generally unintentional. It’s rarely the case that the collision would have occurred had the driver been in compliance with the law. Road design is often a factor because the biggest thing which makes it a fatality is speed, and we have many roads where the speed limit is supposed to be 20-25mph but the design is like a much faster highway, causing almost everyone to go over the limit and thus have less time to react to anyone unexpected and hit them at higher speeds.
This recent blog post was making the rounds based on Washington state data showing the speeding rate _in school zones_ was over 70%, which I think is what you’d find almost anywhere in the United States:
> Despite facing more than seven years behind bars, a judge sentenced him to probation in June.
I assume this is the usual issue where the news looks at the maximum sentence and the legal system uses a standard formula that almost never applies the maximum?
> It was among a series of deadly crashes investigated by the National Highway Traffic Safety Administration that led to this week’s recall.
Assuming the sentence actually is lighter than typical, maybe this is part of why?
No time, even suspended time? You don't have suspended prison time in California?
I mean, i understand it's the US, but you got to at least remove the driving licence for like 5 years and add a 5-10 year suspended prison time (until his next road infraction, seems a good way to prevent him from killing people by arrogance and negligence again).
I read HN as a single HTML page in a text-only browser. For example, it's about 7MB today.
To read the flagged submissions I do something like
sed -n '/\[flagged\]/p' yc.htm > 1.htm
links -no-connect 1.htm
This excludes the comments so I can evaluate the submission myself without the noise.
Or, if I'm reading yc.htm, I might search for the pattern "[flagged]" in the page. In the browser I use, searching is done with a single keypress "/" and then "n" for next match an "p" for previous.
Popular browsers might require more complex keypresses or possibly mouse clicks; they might be sluggish with a 7MB HTML file. Not sure.
Nice. I've thought it would be nice to be able to view posts that made it to the front page, and perhaps surpassed some level of commenting, and were then flagged off.
112 comments
[ 3.1 ms ] story [ 206 ms ] threadI should restate I suppose: Criminal law determined each life ended was only worth $11,500 each in this instance. This precedent determines you can drive your Tesla 74mph into another vehicle if you can pay a $11,500 fine for each random person you slaughter.
Which component of justice— retribution, incapacitation, deterrence, rehabilitation and reparation [1]—is served by jailing this person?
The only one I see is retribution. Given the families are pursuing damages, the reparation prerogative is better served with him out of jail (and not draining his assets fighting prison time).
[1] https://www.unodc.org/e4j/en/crime-prevention-criminal-justi...
If he was, say, watching a movie on his phone, then he should go to jail to deter others from watching movies on their phones.
Sure. Nobody was murdered.
> clearly 23k is not high enough to stop many people from repeating this crime
That’s not the penalty. It’s a criminal conviction, the fine and probation. I don’t know the terms of the probation, but I assume it means he goes to jail for traffic infractions. Also, he’s going to pay a lot more than $23k given the civil case is ongoing.
Even the language around automobiles is corrupted. Essentially all "accidents" aren't.
Owning and operating a multi-ton machine capable of travelling at high speed and rapidly accelerating is a huge responsibility. Being distracted isn't an accident, it's negligence. Not mechanically maintaining the vehicle is negligence. Not supervising your vehicle, even when you have some sort of digital aid enabled, is negligence. None of these are accidents.
Calling a system that is not fully self-driving "autopilot" while knowing exactly what the general public considers that term to mean is....something worse than negligence.
The driver was convicted of a felony.
> Calling a system that is not fully self-driving "autopilot"...is....something worse than negligence
So we agree that the driver isn't, by default, subject to a minimum penalty of jail time.
Given the rates of vehicle deaths in the US, the penalties are clearly not high enough to act as a deterrent.
The "default" for this charge absolutely involves at least a year of jail time (that is, in fact, the defining difference between a felony and a misdemeanor), and the plea deal in this case is an affront to justice.
Do you expect Giuliani to pay a cent?
It’s not binary, it’s a distribution and I do believe you can alter that distribution.
Or give those who negligently kill with their cars a choice: serve the jail time for your negligent homicide, or give up your license for life. That would still have a deterrent effect while not being life-ruining (just life-inconveniencing)
Is this not default? They have clearly proved that are not fit to participate in road traffic.
For a lot of people $23k is minimal disruption to their life.
Which country? Because study after study shows deterence is weak at best for such crimes.
If the person can't improve, and is caught reckless driving or drunk driving (or worse, killing/hurting someone else), then the 'suspended' sentence transform into a full sentence, to protect other people from the idiot.
We see this same problem with face recognition software that says an innocent man's face matches a criminal. The cops just accept what the computer says and arrest people because the computer says it's a match.
We can't have the machine do stuff for us unless it's superior to human judgment, or the stakes are much lower than ruining lives.
Keep in mind 125 people are killed on the roads every single day in the US. If all vehicles magically switched to this system overnight and the system only killed 100 people per day, that would be a huge win.
The fact some people will be killed by automated driving systems is not enough to ban them. If it was, we would ban all cars.
The only important statistic is if those systems kill less people than the current state of affairs.
I expect the 125 people who are killed today have a high factor of alcohol being involved, and our society doesn’t take drunk driving lightly. Those drivers would certainly go to jail. Why would phone use be any more accepted if the death rate suddenly dropped to 124, or, as in your hypothetical situation, 100?
So I guess the answer to your questions are yes and yes?
The difference is that machines have no agency. Imagine a scenario where on roads with 100% human driving, the most of the people that are killed are PCP addled people running naked across freeways where they aren't expected to be, but with self driving you have 50% grandmas walking with babies in a stroller across crosswalks. We don't have to say that one life is more valuable than another to come to the conclusion that the latter is worse, all we have to do is acknowledge the pedestrian's portion of the responsibility for the outcome.
With that in mind, your last sentence is false, the aggregate death count is not all that matters, what matters is the situations in which they occur.
Now let's think about the link in this comment. We begin to form the rules of crossing a street around the behaviors of machines. Not unreasonable. But then it's just a small step from there to "you should've known how the machine was going to behave because it's behavior is simple and predictable." We wind up in a situation where we absolve the self driving car of all liability when an accident occurs because the only entity involved with agency was the pedestrian. We do this already with train tracks.
So with that in mind, the machine systems killing less people is not enough, it has to be so good that the only deaths are suicides, unbelievably bad behavior by those killed, or freak accidents so unlikely that they genuinely shock everyone who learns of them. That's the only way to transition from humans to software operating the machines and be able to justify the ruthlessness with which machines operate themselves.
I don't think it's helpful to introduce this scenario that is entirely not true in the real world and therefore doesn't teach us anything. You're saying that a huge number of road deaths are people that were asking for it anyway.
Dozens and dozens of people were killed on the road today who were not drunk, who were not on drugs and were not breaking a single road rule.
This is the reality of the world. Good people are being killed every day by the way things are now.
If we can do anything to reduce that number, we should.
...
Another way I like to think about this is from the ad in Victoria where a guy has to pick which of his family members will be killed on the roads [1]. Obviously none of us want any of our family to be killed, we all choose zero.
Closer to zero is better than not closer to zero.
[1] https://www.youtube.com/watch?v=k2tOye9DKdQ
Current situation: 125 people are killed every single day
Future autonomous situation: 100 people are killed every single day
I don't care who they are, what they were doing or why they were killed.
The details of their deaths are no important. Saving 25 lives per day is clearly the right thing to do.
Hint: Imagine all of these people are your family and friends. Do you want to keep having 125 of them die every day, or only 100? Does it matter if they're being run over as pedestrians or slammed into by drunk drivers or simply sliding off the road on ice? Of course it doesn't, saving 25 of them every day is the only thing that matters.
I think it's clear that if we were talking about saving 125 people 90 years old or older or 100 children, the clear choice is to save the children, from almost any rational point of view. Even just from a societal health or utility standpoint, children are worth more than people that are statistically close to death anyway at a societal level.
If that premise is acceptable, and if we accept that switching to self driving might change the demographics of who dies in accidents, then it's important to quantify not just whether it saves more people or not, but what people, in case that shifts matters.
> Hint: Imagine all of these people are your family and friends. Do you want to keep having 125 of them die every day, or only 100?
I would rather lose both my parents than one of my children. I don't know a single parent that would feel otherwise.
Additionally, there's also the possibility that like adding extra lanes of traffic to a freeway can make traffic jams worse, self driving cars could increase deaths per capita while reducing deaths per trip, by promoting a lot more car travel (by either allowing current non drivers to drive or increasing time current drivers spend in the car).
That assumes that the people currently being killed on the roads are "less valuable" than the people that would be killed on the roads by autonomous systems.
There is no evidence to prove that, it's purely hypothetical.
Im not arguing that autonomous cars will be worse, I'm just arguing there is the possibility for them to be worse in non-obvious ways, and you prior arguments seemed to ignore that. We should not be so blinded by a single metric that we fail to understand what's really going on. To are many cases of that in technological advances, legislative change, and societal shifts, that we aren't aware of until long after the additional damage has been done.
By the time they're widely deployed, autonomous vehicles aren't going to be a little bit safer than the average human driver. They will have to be superior by at least one order of magnitude. As in, the number of deaths per million miles at least 10X lower than the human average.
I think one order of magnitude is the least we should demand. I would consider it a rather low bar. I suspect that it'd need to be more like two orders of magnitude in order to retain sufficient community consent.
So if that's the threshold of acceptability, then we'd be talking about slashing deaths by 90% or possibly even 99%. Either way, autonomous vehicles could kill children exclusively and they'd still be safer around children than human drivers.
So right now we have 125 deaths every single day on the roads in the US. Same today. Same tomorrow. 45,625 every single year. Year after year.
You're saying it would need to be < 12.5 per day to be acceptable, but really we should aim for < ~2 per day.
So that means if we invent some system that reduces the number of deaths PER DAY from 125 down to, let's say, 30, you think we SHOULD NOT implement it, because that's not enough of an improvement?
Let's say that system is the best we can ever make, but we decide to never turn it on, because it's not a whole order of magnitude better. You're choosing to kill an additional 34,675 people per year, because a reduction in deaths from 125 per day down to 30 per day is "not good enough".
But if the best we can do is a 9x improvement of course we should still deploy it. The question is whether society will accept it. I’m not sure it would. I’m not sure it would accept a system that reduced deaths by two orders of magnitude, or 100x.
Imagine for a moment, Tesla fully cracks the FSD nut, and every car on the road is replaced with a self driving Tesla, 100 times safer than humans. “One person is dying at the hands of the pedestrian butcher Elon Musk every single day!” the headlines will cry.
You're absolutely right, but it won't make any difference if the laws are passed.
(It's sad that society is obsessed with politicising everything and everyone. I don't agree with some of Musk's political views, but so what? I never expected to be in perfect alignment with him or anyone else anyway. And it would be the height of arrogance to assume that my views are consistently better than his, rather than a different order of priorities resulting in a different set of trade-offs.)
If indeed it is an order of magnitude less deaths, then it probably doesn't matter if the demographics shift, but we actually have to see that change and know the demographics to actually know that. I'm just saying that we should pay attention, as we should with anything that has the capability to change per-capita figures for adverse actions, as to what is actually happening and not assume.
That said, since we're already seeing an increase in these cars and the use of their autonomous options which match "self-driving" to greater or lesser degrees, I doubt we're going to see a 10x decrease before they are inthe wild, because to some degree they already are.
I just want us to know what that means by looking at as much info as we can rather than a single metric.
I would.
But my main point is that it is a machine pulling the lever or not. Once you deploy such a machine we are all subject to it's cold calculation. We should go the extra mile to make sure the world we live in winds up being better and not just assume less deaths = better outcome.
As I've said elsewhere, I don't expect any autonomous driving system to be anything less than 10X safer than human drivers, or one order of magnitude. And I don't think there could ever be a durable social warrant for autonomous vehicles unless they're at least 100X safer, or two orders of magnitude.[0]
Given a minimum expectation of being 10X safer, and given that drivers don't get to choose the demographics of people who place themselves in harm's way near our roads, it is statistically unfathomable that autonomous vehicles could ever be net more dangerous for any demographic.
--
[0] If anything I feel that's too conservative and three orders of magnitude might be closer to the threshold at which societies would tolerate autonomous cars. Road deaths in the USA are in the order of 40,000 annually. Three orders of magnitude still means 40 deaths per year. That's plenty enough to keep the press aroused and erect.
(You may have a point with respect to systems like Supercruise and Tesla FSD, but that’s not the topic of this article.)
Autopilot. This automatically accelerates, decelerates, and steers on roads. This works on all roads with lane markings, not just limited access freeways. It also obeys speed limit signs.
Enhanced autopilot. This is autopilot but will also self drive from on-ramp to off-ramp, changing lanes to pass (and exit the passing lane), taking interchanges to other freeways, and exiting the offramp.
Full self driving beta. This will drive on pretty much any street. Early on it was purely a gimmick that required constant intervention, but lately it’s gotten pretty good. My main complaint is that it drives too conservatively, causing impatient drivers to behave erratically around you. Imagine someone driving as if they're taking a driving test. Because of that I tend to drive manually unless I’m on a freeway.
Really? My understanding is that many new vehicles have lane keep assist and dynamic cruise control, but that they are much more limited than Autopilot. For example, lots of cars have much tighter limits on the curves they will follow. They also don't have the UI that Autopilot has, which shows what cars it detects. This gives more of a sense of full autonomy (although I find it a bit creepy/worrisome when vehicles appear and disappear on the screen for no reason — one would think this would give people pause).
There are definitely more expensive vehicles (Cadillac, Mercedes) that have more robust systems, but my sense is that your typical Toyota, Subaru, etc. isn't close to Autopilot.
Some people argue that Tesla customers think they've been sold autonomy. I can't see how. Tesla goes to pains to make you aware of the $12k you would need to spend to get the system that "self drives". You are painfully aware of the $12k you didn't spend.
• Crashes per mile on each manufacturer's "self-driving" systems
• Crashes per mile on each manufacturer's LKAS (lane keeping assistant) systems
• Crashes per mile on each manufacturer's TACC (traffic aware cruise control)
• Crashes per mile on each manufacturer's regular cruise control
• Crashes per mile under manual control
Break each out by highway, suburban, city, and car model year. Also include whether the crash resulted in injuries and deaths, and whether those deaths were occupants or not.
Then we can actually determine where the problem is, and the scope of it.
Unpopular opinion: there probably isn't a problem? None of these system is going to be perfectly safe, but there's quite a bit of evidence at this point that they're all safer, as actually used (which, yes, includes some misuse) than a human driver is at doing the same task.
No one wants to hear that though, so instead we take "where the problem is" as a prior and continue arguments like this ad nauseum.
Full disclosure: my car is going to drive me to Idaho in the morning.
At the end of the day you either believe or you don't. But if there was an actual problem, we'd know. There are millions of these things on the street now.
It's also absolutely hilarious to me you're assuming I think Autopilot is unsafe.
...counting miles where self-driving was activated including the 5 minutes afterwards — https://www.motortrend.com/news/nhtsa-tesla-autopilot-invest...
(Honi soit qui mal y pense.)
This recent blog post was making the rounds based on Washington state data showing the speeding rate _in school zones_ was over 70%, which I think is what you’d find almost anywhere in the United States:
https://www.theurbanist.org/2023/12/12/drivers-are-speeding-...
I assume this is the usual issue where the news looks at the maximum sentence and the legal system uses a standard formula that almost never applies the maximum?
> It was among a series of deadly crashes investigated by the National Highway Traffic Safety Administration that led to this week’s recall.
Assuming the sentence actually is lighter than typical, maybe this is part of why?
I mean, i understand it's the US, but you got to at least remove the driving licence for like 5 years and add a 5-10 year suspended prison time (until his next road infraction, seems a good way to prevent him from killing people by arrogance and negligence again).
Also thanks for flagging these too:
https://edition.cnn.com/2023/12/13/tech/tesla-recall-autopil...
https://arstechnica.com/cars/2023/12/more-than-2-million-tes...
https://www.bbc.co.uk/news/technology-67693935
I might have missed them otherwise.
To read the flagged submissions I do something like
This excludes the comments so I can evaluate the submission myself without the noise.Or, if I'm reading yc.htm, I might search for the pattern "[flagged]" in the page. In the browser I use, searching is done with a single keypress "/" and then "n" for next match an "p" for previous.
Popular browsers might require more complex keypresses or possibly mouse clicks; they might be sluggish with a 7MB HTML file. Not sure.