- cross-posted to:
- technology@lemmy.ml
- cross-posted to:
- technology@lemmy.ml
cross-posted from: https://derp.foo/post/81940
There is a discussion on Hacker News, but feel free to comment here as well.
cross-posted from: https://derp.foo/post/81940
There is a discussion on Hacker News, but feel free to comment here as well.
So I hate Elon Musk and I think Tesla is way overhyped, but I do want to point out that singular anecdotes like this don’t mean anything.
Human drivers run red lights and crash cars all the time. It’s not a question of whether a self-driving car runs a light or gets in a crash, it’s whether they do it more often than a human driver. What are the statistics for red lights run per mile driven?
I’d hazard a guess that it’s lower, but regardless shouldn’t be available to the consumer yet if this is what they found in preliminary testing. Trying to hide it is quite disingenuous though - of course it’s going to make mistakes while in testing and even after, trying to hide those mistakes and act as if they don’t exist is not how you treat your potential customer base.
Okay but now there is a tragic accountability scenario to deal with. You forget you are discussing human beings and not statistics
When we’re talking about public safety, it should be entirely about statistics. Basing public safety policy on feelings and emotions is how you get 3 hour long TSA checkpoints at airports to prevent exactly zero attempted hijackings in the last 22 years.
Personally -and this is why you simply aren’t getting your way, because it’s a broad and legitimate concern, - but I’m not interested in letting some corporate special interest plow through the regulatory processes required to manage things such as accountability and justice frameworks with having lethal robots have sovereignty on our public roads. Even if they are less lethal than people - which happen to have pesky rights afforded to them that aren’t afforded to autonomous vehicles. The data is great and hopeful for a good future - the implementation matters. TSA is a false-equivalence, the situations are not the same.
I am not saying we should exempt autonomous vehicle manufacturers from regulation. I’m actually saying the opposite: that we need to base any decision on a rigorous analysis of safety data for these vehicles, which means the manufacturers should be required to provide said data to regulatory agencies.
It’s kind of like how people are more likely to die in a car accident on their way to the airport than they are to die in a plane. Yet people are more likely to be afraid of flying.
There’s a sense of control that people tend to gravitate towards. In the aggregate, Tesla might run fewer red lights than humans. At the individual level there will certainly be “safe drivers” who run red lights at a lower rate, who end up dying because of the Tesla. It’s a hard pill to swallow.
I’ll be honest here, I hate cars and the car-centered culture of the USA. I care way more about the victims of bad/careless/drunk/distracted drivers than I do about the bad/careless/drunk/distracted drivers themselves.
If me being in a self-driving car means other people around me are more safe, then it’s not even a question.
I think what it really boils down to is that the vast majority of drivers who run red lights choose to do so out of stupidity, where as someone trusting Tesla’s claims about their new “self-driving” car might not have the chance to stop the vehicle as it hurtles itself through a red light. So yes, in terms of raw numbers it will cause less accidents in some cases, but that it can happen at all when the average trusting consumer/user would expect to never do that compared to a normal car should be a huge issue.
Also, as far as liability goes, I’m horrified to think about what the future of vehicle injury lawsuits will look like in the US when the driver can blame the software and the company providing the software is run by a grifter asshole.
Your concern seems to be for the pilot of the car that causes the accident. What about the victims? They don’t care if the car was being driven by a person or a computer, only that they were struck by it.
A car is a giant metal death machine, and by choosing to drive one, you are responsible not only for yourself, but also the people around you. If self-driving cars can substantially reduce the number of victims, then as a potential victim, I don’t care if you feel safer as the driver. I want to feel less threatened by the cars around me.
You’ve managed to make up an angle to this that I made absolutely zero stance on, and then get mad about it, congrats.
Everyone on the road near these stupid things is at risk when Tesla pretends like it’s road-safe and the moronic drivers trust it to obey traffic laws. The concern is obviously for everyone involved, not sure why you’re pretending I said otherwise.
If I know I’m mostly surrounded by humans who on average don’t accelerate through red lights, I can make certain assumptions when I’m on the road. Yes, the car next to me could randomly swerve into my lane, but on average you can assume they won’t unless you also observe something happening farther ahead in their lane. When you start adding in the combination of bad-faith company and terminally naive driver I described above, you drastically increase uncertainty and risk to everyone within range of the nearest Tesla. The unknown in that equation is always going to be the stupid fucking Tesla.
So then you’ve just circled back around to what I originally said: is it actually true that you’re at more risk near a Tesla than you are near a human driver? Do you have any evidence for this assertion? Random anecdotes about a Tesla running a light don’t mean anything because humans also run red lights all the time. Human drivers are a constant unknown. I have never and will never trust a human driver.
I would actually say yes, the Tesla poses more risk. Driving safety is all about anticipating what the other drivers are going to do. After commuting in Houston for 2-3 years, I actually became quite good at identifying scenarios where something dangerous could happen. I wasn’t always right if they were actually going to happen, but I was always prepared to take action in case it was. For instance, if the positioning is right for someone to suddenly cut you off, you can hang back and see if they’ll actually do it. If a larger car is next to you and you’re both making a turn, you can be wary of it spilling into your lane. I avoided a collision today actually because of that.
We have a sense of what human drivers might do. We don’t have that sense for self driving cars. I can’t adequately predict when I need to take defensive actions, because their behavior is totally foreign to me. They may run a red light well after it’s turned red, while I would expect a human to only do that if it had recently changed. It’s very rare for someone to run a red when they pull up to a light that they’ve only seen as red.
This same concept is why you can’t make a 100% safe self driving car. Driving safety is a function of everyone on the road. You could drive as safely as possible, but you’re still at the mercy of everyone else’s decisions. Introducing a system that people aren’t familiar with will create a disruption, and disruptions cause accidents.
Everyone has to adopt self driving technology at about the same time. When it’s mostly self driving cars, it can be incredibly safe. But that in between where it isn’t fully adopted is an increase in risk.
Again, we don’t need a 100% safe self driving car, we just need a self driving car that’s at least as safe as a human driver.
I disagree with the premise that humans are entirely predictable on the road, and I also disagree that self driving cars are less predictable. Computers are pretty much the very definition of predictable: they follow the rules and don’t ever make last minute decisions (unless their programming is faulty), and they can be trained to always err on the side of caution.
You’re still missing the point. It’s not about how much the drivers around the Tesla should “feel safer” (and they absolutely shouldn’t), it’s about the misguided trust the Tesla driver has in it’s capability to operate autonomously. Their assumptions about what the car can or will do without the need for human intervention makes them an insane risk to everyone around them.
Also, the vast majority of Tesla owners are weird fanboys who deny every issue and critique, do you really think this is an anecdotal edge case? They wouldn’t be caught dead admitting buyers remorse every time their early access car software messes up. We’re lucky the person in the article was annoyed enough to actually record the incident.
I would never trust a machine to operate a moving vehicle fully, to pretend it’s any less of an unknown is absurd. Anecdotal fanboying about how great the tech “should be” or “will be someday” also don’t mean anything.
Do you have statistics to back this up? Are Teslas actually more likely to get into accidents and cause damage/injury compared to a human driver?
I mean, maybe they are. My point is not that Teslas are safer, only that you can’t determine that based on a few videos. People like to post these videos of Teslas running a light, or getting into an accident, but it doesn’t prove anything. The criteria for self-driving cars to be allowed on the road shouldn’t be that they are 100% safe, only that they are as safe or safer than human drivers. Because human drivers are really, really bad, and get into accidents all the time.
This is where our complete disconnect is. IMO when you put something on the road that has the capacity to remove control from the driver it absolutely needs to be 100% reliable. To me, there is no justifiable percentage of acceptable losses for this kind of stuff. It either needs to be fully compliant or not allowed on the road around other drivers at all. Humans more likely to cause accidents and requiring automated systems to not endanger the lives of those in / around the vehicle are not mutually exclusive concepts.
If 100% safety is your criteria, then humans shouldn’t be allowed to drive. Humans suck at it. We’re really, really bad at driving. We get in accidents all the time. Tens of thousands of people die every year, and hundreds of thousands are seriously injured. You are holding self-driving cars to standards that human drivers could never hope to meet.
This is exactly the issue. The driver isn’t at fault, because they’re not even driving the thing. The program is. But are we going to prosecute a programmer who genuinely tried their best to make a good product?
Unless we hold corporations overall liable for this, there is no recourse. And we should hold them liable. If they can be sued for accidents caused by self driving cars, they’re sure as hell going to make them as safe as technologically possible.
I don’t think we should prosecute the programmer, but I do think the manufacturing company should be liable.
The whole argument for self driving cars is they don’t make stupid mistakes. Just because they do it less than humans isn’t a good argument.
I am putting my trust in the technology. It needs to work otherwise I’m not going to use it. It working 99% time is not good enough especially when it failed because it wasn’t good enough.
Beeing better that a human driver is exactly what they need to be. Currently, they fail at somewhat different things compared to human drivers. So if we combine both we get the least amount of mistakes. The car never sleeps, the human never mistakes a bridge column for the road.
I don’t think exceeding 99% is realistic with self-driving cars… There’s always going to be a margin of error with anything.
Probably. But we can do way better then what we have now
That’s not the argument for self-driving cars at all. The argument for self-driving cars is that people hate driving because it’s a huge stressful time sink. An additional benefit of self-driving cars is that computers have better reaction times than humans and don’t stare at a phone screen while flying down the freeway at 70 mph.
If we find that SDC get in, say, 50% fewer serious accidents per 100 miles than human drivers, that would mean tens of thousands fewer deaths and hundreds of thousands fewer injuries. Your objection to that is that it’s not good enough because you demand zero serious accidents? That’s preposterous.
Is there any progress on the question who is responsible for such accidents? Of course less accidents are desirable, but if the manufacturer suddenly is responsible for deaths and not the human behind the wheel, then there is a really big incentive to have zero serious accidents. If the system is not perfect from the start, someday the industry and the government have to decide how to handle this.
I believe the manufacturer should be liable for damage caused by their product due to manufacturing defects and faulty software. This incentivizes manufacturers to make the safest product possible to reduct their liability. If it turns out that it’s not possible for manufacturers to make these cars safe enough to be profitable, then so be it.
The real question is “who goes to jail when that self driving truck flattens a family of four?”
Because if the answer is nobody, then we shouldn’t have full self driving.
This is a corporation breaking the law as a business decision, it should be noted if we let small errors slide they will lose incentive to be accountable .