The Real Moral Dilemma of Self-Driving Cars
ฝัง
- เผยแพร่เมื่อ 18 ม.ค. 2017
- We talk about all the potentially challenging situations autonomous cars could get into but not about how human drivers are not very good. Tens of thousands die on the roads every year in collisions, most of which could be prevented by autonomous vehicles.
Sponsored by BMW
I wanted to make a video about autonomous cars for some time but I hadn't had the opportunity. The self-driving technology is already at a state where it can save lives if only it were more widely implemented.
Links to original clips:
TED-Ed • The ethical dilemma of...
BBC Newsnight: • The trolley problem an...
Music from www.epidemicsound.com "Ambient Electronic Groove," "Pet Animals 2," "The Long Ride."
Filmed by Raquel Nuno
Edited by Trevor Carlee
![CAMPปลิ้น | EP.76[1/2] โจโฉเปิดป่าโหดรับโอ๊ต-ป๊อบ รอบนี้พี่เอาจริง!](http://i.ytimg.com/vi/pNg5O7Pf-a0/mqdefault.jpg)
The REAL moral dilemma is that we aren't all using public transport and saving a fuckton of energy and emissions.
Capta Praelium public transport isn't something that makes sense for a lot of people. Take to much time to get from point A to point B. Sometimes as much as 3 or 4 times as long than via your personal car. Doesn't make sense, at least to me. Correct me if I'm wrong though.
Capta Praelium and the problem with that is not everyone can use public transportation. Or cares to double their commute time. Or works at the same place everyday. Or works at the same place all day. Mass transit is simply too inconvenient for most, hence mass car ownership.
That's all well and good if you live in japan, but if you've seen the london underground you'd know why people don't use public transport as much.
Capta Praelium Cause public transit is a JOKE in North America, especially in Canada.
No, public transit is not a joke in North America and Canada. You should come look at public transit here in South Africa. It's disgusting.
what the hell is the point of the buttonless button thing? other than "look at our cool sci-fi car"
so much more expensive and no real benefits other than looking pretty.
Why does there need to be any other point, when that one is so completely valid?
It's a concept car. It's a car to showcase concepts that allow people like you to give feedback. My guess is the acoustic feedback might be used in future projected keyboards or even VR.
If you need any other reason, then i will never understand you. Thing was cool.
Sure, it's cool, but did you wonder why there were 2 buttons at most at all times? This thing is inaccurate as hell and a holographic keyboard is useful only if you press one button at a time. A great concept nonetheless but still highly impractical.
At the end I thought it was the car that said "I hope you enjoyed the drive" and I thought "wow they even gave it a German accent!" 😂
lol me too
Ban the damn things
I would've been fine with that if GreatScott! did the voice; "Let's get started!".
I read your comment first and still thought it for a moment
frankly, they should use his voice to say that, sounded so pleasent.
I think a good solution for the autonomous vehicle is neither hitting the SUV nor the bike; it's doing a vertical takeoff & flying over the truck.
This problem was just SO bad!
1. Brake? Hit nothing? ewentually get your back bruised by the guy behind you?
2. Average human driver would absolutely panic and definitely not think about what is a better choice, possibly making way more damage than any autonomus vehicle would
3. This sort of situation is non existent in comparision to driver fault accidents
Chronologious so we’re in Speed Racer now? IM HYPED
@@ninoski4043 The correct solution is to maintain enough distance from the truck so that the self-driving car can safely stop without crashing into the falling bale of hay or swerving into other lanes. If the car behind is too close, it just means keeping even more distance from the truck in front.
Literally every single variant of the trolley problem with self-driving cars is resolved by maintaining safe speed and safe distance. It's a big fat non-issue brought up again and again by people who know nothing about engineering. The real problem everybody should be thinking about instead is that sensor data and the AI analyzing it are both unreliable. That pedestrian your self-driving car is trying to avoid? That might be a mere sensor glitch.
@@nextghost "The real problem everybody should be thinking about instead is that sensor data and the AI analyzing it are both unreliable"
Which should bring people to try to improve both of those rather than trying to eliminate the possibility of being replaced as a driver. Even if these are unreliable there is a point at which they are more reliable than the alternative - a human. Humans are insanely unreliable aswell, apart from wrong behaviour, their sensory data and the brain interpreting it isn't reliable either.
Not only do many humans have flaws in their sensory systems (wearing glasses for example) but even with a completely healthy body and mind, humans are susceptible to wrong interpretations. Hell, there is a whole industry built upon the fact that human sensory data and interpretation isn't reliable: magic/illusions.
As soon as the error of self-driving cars is smaller than that of humans it's irresponsible - not - to replace humans.
There is also another point that's flawed with the whole "dilemma" shown in the video: It assumes that a self-driving car has to drive among a majority of human-driven cars, which again would be irresponsible as it is the same as argument 1.
If every car on the street was autonomous and communicating with each other over a network, then the surrounding cars could act accordingly given the situation shown in the video. While humans generally - don't - think that much in such situations, let alone think about how - other - drivers react, resulting in even more damage if the decisions of several humans don't align (one could also consider that reaction and calculation time of humans horrendous). A network would pick - one - path and perform it with all necessary cars, hopefully the one that has the least damage for all participants.
But it's not even necessary to pick the path of least damage, it's only necessary to pick the path of - less (or equal) - damage compared to the one a human would pick.
@@victos-vertex Imagine a sun sneezing driver after each of a serie of tunnels...
I feel the biggest question is who you should sue if the car hits you.
But this is already settled. Who do you sue if a vehicle on cruise control rear-ends you, or a brake failure results in a collision that could otherwise have been avoided, or if an airbag deploys and injures a passenger who is a child. These legal questions have already been tested in the court, and the type and level of technology at play will not make a difference. Each State in the U.S. has different rules, but it basically comes down to contributory liability.
"vehicle on cruise control rear-ends you": a human is still in charge of the cruise control.
"a brake failure results in a collision that could otherwise have been avoided" OEM if its a factory fault, which is what it would be if the autonomous software made a bad decision. Human if it was negligence in maintaining the car.
"if an airbag deploys and injures a passenger who is a child": Human had to put the kid in front of the airbag. If the car had an "autonomous airbag" that was supposed to tell who was in the passenger seat, then it WOULD be the OEM at fault.
And this bring up another issue;
Who is at fault if there is a software update to correct a discovered fault in the OEM autonomous software, but the owner of the car has neglected to install the update prior to the accident?
@RyDaddy I assume you are an attorney with lots of experience litigating these issues across the Country in order to provide such an absolute response. My point is that the question asked by the OP and so many others is posed as if we have no idea on how to answer it (and, somehow by extension, getting these answers is necessary before moving to self-driving cars). But the truth is that very similar questions are answered many times a year in the various states, and that the advanced technology of self-driving automobiles doesn't pose some new, unanswerable liability questions that our current system can't handle with ease.
You sue the person who owns the car but that isn't a big deal. Insurance on a self driving car should be pretty reasonable, since they are a lot safer than human drivers. I would expect when there are a lot of people using them, it will be a fraction of what insurance costs today.
It like what CGP Grey said:
"It doesn't have to be good, it just has to be better than us. "
yeah
It doesn't matter how good it is. Ultimately, people need somebody or something to blame. The bigger moral problem isn't that of "shall I kill the biker or the family in the SUV?" It is, "If I kill the family in the SUV, who is ultimately responsible for it? The driver? The programmer? The car company? And who is financially responsible for it?" Everything boils down to money.
@@TTaM581 the most important question is can a machine decide who to live and who should die.
@@dmitrizaslavski8480 can a human do either? (in this hypothetical situation we don't know if the suv driver is a murder or something like that)
@@Gamer-uf1kl People believe so, at least. When I did research work at IBM, in medical department, we succeed to reach about 93% of CT correct interpretations. The best physicist had a score of 94% and average 91%. All experienced ones. But you can't license such tool for automatic diagnosis as only human can be morally accountable for life changing decisions if machine is not "perfect".
Waiting for the first "no human-driven vehicle allowed past this point" sign
sounds like a discrimination lawsuit in the making.
@@Historyfan476AD Not really, you don't allow cars on railroad either.
@@Rig0r_M0rtis Railroad tracks are different from the public road.
@@Historyfan476AD Robot car only roads are also different from normal public roads
@@Rig0r_M0rtis but the difference is that there were never cars on railroads
I can't wait to drive to work and then send my car off to find a parking space on it's own
This. I would be fine with stuff like this that improves our lives, but we definitely need to have optional manual for the rest of the time.
More likely scenario: You won't own the car at all, you'll summon a car to drive you to work by itself, then you get out and it goes off to serve someone else's journey. That way, driverless cars will also solve the problem of parking, which, when you think about it, is a ridiculous problem to exist in the first place.
@@JesusJuenger the upper class will likely have their own personal vehicles to shuttle them immediately on demand, while the average Joe will probably summon a public vehicle to transport them, much like a taxi. That gets rid of the need to own a car, the need to license/register a car, and the need to attain a license to operate the car.
*its
@@brettwarren5976 You're quick
Is this the end of Russian road rage videos?
The only legitimate counter reason to having driverless
?? those people in those videos need these driverless cars
Only if you ban Alcohol!
Mønkey went to check it out, Russian rage is awesome, and no that won't stop because they're morons
I am live in Russia , Russia its a stupid country
I can picture my future children asking me what it was like to drive a car.
And you can tell them no one ever died to computer failures.
+First Last And tell them that thousands of people died because we were too stupid to actually drive and instead checked a post on facebook about what someone was eating for lunch.
No, that would be silly.
if(!(computerErrors > humanErrors)) {
std::cout
if (computerDriver.hacked) {
std::cout
1:38 I love how it being an autonomous BMW is doubleparking right there :D
Pretentious jackass attitude built right in to the AI!
it wasn't done, though and it wasn't for sure parking, they also have a memory reverse feature that could've been being showed to people
@@kingacrisius maybe they were all bmw's :).
the other car isn't parked correctly, so it's actually more intelligent than just following the lines and crashing into the falsly parked car ;)
Just like a normal BWM driver would
I get an aneurysm every time someone brings up the "trolley problem" for self-driving cars. I one-up you with the META TROLLEY PROBLEM: Should an engineer devote time fixing some incredibly esoteric situation involving a little girl in a tiny alleyway that might happen twice per year in the world? Or should they be focused on fixing the kinds of situations that happen thousands of times per year and cost thousands of lives, where the safe decision is obvious and there's no moral dilemma?
Plus, if self-driving cars can communicate with one another, they can orchestrate traffic in such a way that surrounding cars swerving away from the girl saver. Take the example in the video. The car behind the truck detects a danger forming. It can orchestrate with the vehicles around it to make room on the left for the car directly behind the truck by speeding up the car to the right, and slowing down the car behind the car to the right. Up to 100m behind the obstacle gets cleared of cars by merging that lane into the lanes next to it, and a virtual road block is put in place until it has been cleaned.
But even if there is no communication, go into the car on the left. Hitting the motorcycle will cause serious harm to the driver. The car has only 1 confirmed person, and they are on the other side of the car, so there is a lot of crumple zone. And even if someone is in the right side of the car, they'll probably be better of being in a car and getting hit, then being on a motorcycle getting hit.
@@mike140298 Great comment. Exactly this. Communication is so amazing and so fast.
Engineers only need to make it just a little safer than normal people driving it.
If engineers make it even 1% safer: let's say there's 100 deaths per 10000 vehicles per year in a country. If engineers make their self-driving cars to only result in 99 deaths per 10000 vehicles per year in a country, then all cars should already be self-driving because they're already safer than people driving it. And it will only get better in the future with electronics and software updates.
Dilemmas don't matter, only the actual end results.
@@KulaGGin But society is stupid AF, so even though autonomous cars are already _much_ safer than human drivers and are quickly getting safer, getting them to be legally allowed and implemented on a large scale will take quite some time and effort.
@@leandrog2785 Exactly. Because people were scared of fire, and they still love to boycott things. They like to assign political arguments to something that could be so beneficial.
People who say theyd never drive in a self driving car, enjoy the insurance spikes once self driving cars become mainstreamed.
This is the same BS people were saying about VR. Where exactly is it now.......? Even if at some point they do become mainstram,it will be dozens of generations later assuming we haven't gone extinct.
Trully a dreadful future.
Are you implying that what people say has a direct implication in what will happen?
Kronguard dozens of generations? Self driving cars already work, google has the out on real roads every single day. You can go buy a Tesla right now that is almost fully autonomous while you're on a freeway. All of this is progress that has only been made in the last few years, it's a matter of a few years before they become reliable enough to mass produce for general consumer.
As for VR, it hasn't gone anywhere. In fact the business is doing just fine, you're just not hearing about it every single day because there isn't a story about it for the media to cover. "Business continues to grow reliably" is hardly a headline. Go do some googling about it if you don't believe me.
Kron, you might actually be living under a rock. Do some research, self-driving cars are coming faster than you think.
Was expecting a more in depth video, this was cool but clearly an ad.
Santiago Ferrari i listened to it as i was doing stuff. Judging by your spelling I'm clearly not as clever as you
Santiago Ferrari 'includes' not IS
so? ads can be informative too, you know
+Troy Tribbiani yes they can but the video just reached the interesting point and cut off. At least a more in-depth interview with the guy from BMW could have been included. And other specs of the car, what is it made of, how durable is it in case of accidents, etc.
ye, that woulda been cool
2:53
Isn't this because pilots only take controls on the most dangerous part of the flights(liftoff/landing)?
Yes, planes are safer when autopilot is on because autopilot is only on during the safest times during a flight and if anything bad was happening the pilots would immediately take control from the autopilot so there really is almost no opportunity for a plane to crash while autopilot is on assuming the pilots are conscious and can take control.
The autopilot can also land the aircraft but its only done if its very poor visibility, and like always, like the guy above me said, its important the pilots are conscious and have an eye on everything
Yes, it's the same problem with Tesla Autopilot safety reports - Autopilot is used mostly on highways and in good conditions; although that's slowly changing with software updates. And it's good that they also include their cars without autopilot and active safety features, as something like a control group... even though they usually compare Autopilot to the general average to get a bigger difference.
I'm sure that is a big part of it. Pilots are always engaged in takeoff/landing.
Autopilot can land a plane easily there are reports of people crashing planes cause they took over and failed the landing.
Man I thought this was depressing until I heard that happy royalty free music, phew I almost had to experience an emotion.
Damn good stuff, sir. 🥇
I liked the way TH-cam showed it was a paid promotion, it was apparent, but not obtrusive.
Damn, it actually got me too. I'm a BMW driver, and as I was watching it I was like "If I get a self driving car, i'd like it to be BMW"
What about the aspect of blame? People really aren't comfortable with the idea that they could die in a car crash caused by a machine, even if the machine is much less likely to cause a crash than they are.
James Morgan hence humans are stupid we care more about how we feel than reality
I'll be equally pissed about dying in a crash no matter if it caused by person or the machine.
What about the aspect of blame? People really aren't comfortable with the idea that they could die in a car crash caused by a person, even if the person is much less likely to cause a crash than they are.
The problem is the same for a better driver.
what about people who die due to other machine error today? For example air bags can fail, elevators can get stuck, etc.
that happens but not often failure of machines or engineered product or "machine error " are very rare how many stories of people dying from a failing car engine or failing brakes? rarely and surely not as much as accidents caused by DUIs or human error also have you seen the way people dive?
I think the premise that automation is one of the best ways to make the roads safer ignores the obvious solution (one car companies hate)... and that is LESS CARS! I'm pretty sure every form of public transit has a safety track record that is far better than driving (perhaps even if driverless). Sure it's not applicable in every situation, but it could be utilized/leveraged more.
Cars allow for more flexibility. And public transportation at a large scale only works in cities. In towns or suburbs it really doesn’t work because there isn’t enough demand.
In places like NYC most people who live there don’t drive at all, they just get around using the subway, which is ideal because of the population density. There are so many people that there is enough demand for there to be a subway station within a few blocks of you no matter where you are.
In a rural area though this isn’t as practical. Cars still make the most sense in that case.
If self driving car becomes the norm that kind of dilemma won't even exist anyway.
The most obvious development direction of the car will lean towards building network between cars so that every car will communicate with cars around them and any movement done by cars from let's say a kilometer ahead will be known by the car way behind him.
That is to say if one of the car have accident and had to suddenly stop it'll signal cars behind it to also brake minimizing possibilities of chain crash
this means that all cars will maintain a safe distance whenever possible that let them brake in emergency
it also means that they'll know the most optimum speed that the car should cruise at without ever being reckless.
Most certainly the mixed period between full human controlled to full automatic cars are the worst, but the problem will never be gone completely. Sure the car may know what every other car is doing at the moment, but what about that kid that run out in front? yeah it wont know about that, so unless you can 100% isolate roads with only autonomous cars such issues will exist. Yes those kind of accidents will from a mathematical standpoint be way lower than with human drivers so in that sense it does make sense to get them on the road ASAP, BUT the issue arise WHEN those accidents do happen, who is then responsible for paying for damages and such? Is it the passengers of the car (afterall at that point there would be only passengers, no drivers in cars), or is it the car manufacturer, or maybe the programmers, or some other people
@@GummieI You forget about the victim. The victim can also be deemed responsible, or in your case it's parent. But such thing are something relatively trivial that can be solved with law. That is to say it's way easier to make a law to solve easily foreseeable consequences rather than solving 99% of traffic problem.
Remember at one point in history not too long ago, pedestrian have the highest priority of road usage. Then it took less than a generation for car to occupy 80% of road, push pedestrian aside and create entirely new set of values and law.
People adapt faster than they themselves are aware off.. maybe because it relatively slow from human stand point but absurdly fast from historical stand point.
Rather some trivial hypothetical problem like random kid dying because the stars aligned, I'm more concerned about the real practical problem which is your privacy.
Having smart cars is one thing, having a car that track all your destination and definitely be used to bombard you with ads is not nice, ESPECIALLY if you pay the full price of it.
or other pressing concern like their serviceability where right now they can easily cockblock you from doing simple maintenance by citing BS reason.
Ethical conundrums can be programmed out.
In the situation where the car would have to choose between swerving into the SUV or the Motocycle, the situation forgets one simple tactic of avoidance that could be easily programmed into a self-driving car.
If (Vehicle ahead has dangerously insecure load -> if load is open to the air, if load is secured by ropes, etc.)
Then:
- Alert other self-driving vehicles via internet where available
- Keep distance (>100m)
- Change lanes if possible
If (Vehicle ahead loses load)
Then:
- Send alert message to other self driving cars behind
- Break immediately (The distance cushion of 100 meters allows this)
- Alert authorities of fallen load at GPS location
- Proceed around fallen load once area is safe
- Send route to other self driving vehicles to make them aware of fallen load
Even then, in a totally self driving environment, the vehicle with an insecure load wouldn't even exist, because it would most likely be packed by other robots.
Every single issue humans have on the road are caused by humans, the best way to remove the issues, is to remove human involvement.
But there is going to be a long stretch of time when self driving cars are going to share the road with mostly not self driving cars, for the very simple reason that self driving cars are going to be more expensive than normal models, and the used car market is huge and keeps things in market longer than you think.
Like, until a few years ago, maybe 3-5, a stick shift in every case would cost less than an automatic of the same model. I know this, because I drive a stick, and I'm going to stick with it as long as I can get one because it gives you more control and, for used cars which I buy because I don't have that much money, it is cheaper.
easily programmed? LOL
you are forgetting that the car has to use a camera and there are towing trucks, trucks with images printed on their sides, and lots of other vehicles that make this almost impossible to recognize through code (especially at a distance of 100 meters)
Self-driving cars wouldn't rely only on cameras; they are also be equiped with extra sensors like RADAR and LIDAR; even if the trucks come with printed images, that won't matter for such sensors.
+Luke Actually, self-learning AIs would take care of that. You don't have to program every last detail, because the machine can analyze similar situations and past data not only from itself, but also every other self-driving car out there. They can think, research, learn, calculate, and react in a matter of a split-second. Which is far better than I, you, or anyone else could do in those situations. And no, it isn't sci-fi.
+J T The biggest problem is still "sensors" and actually recognising the world and it's objects as it is. Sure, radar, lidar, ultrasonic distance sensor, cameras are all nice an dandy but aren't sufficient at all for every day traffic situations. The AI has enough trouble to actually decide if something is a moving vehicle or not. Think about unusual shapes like a timber transport or other unusual vehicles. Also Knee is right, most environment detection would be done via camera and pattern recognition / image processing.
There are even more extreme cases. I saw a dashcam video where an airplane crashed right next to a highway and a lot debris suddenly sweeped across the road. The driver has partly seen and heard the crash and slowed down before some of the debris catched his car. Most autonomous cars wouldn't have detected this or the response would be some random last minute actions. An AI has the advantange that it can react much quicker than a human, but it's "view" and sensory system is way more limited. Sure you could equip 20 highspeed-highresolution cameras around the vehicle and have like 100 CPUs doing all sorts of image analyzing as fast as possible. However that car would be so damn expensive that even the military wouldn't buy it ^^.
Also a big problem is: Who's responsible if something goes wrong? The car manufacturer?
Also a general problem: The more we automate the bigger is the attack surface for hacking and manipulation. If you have seen iRobot you know what i mean ^^. Sure it's "just" a movie, but a lot sci-fi has come true already. Cross communication between cars is a great idea to prevent crashes, but on the other hand the possibilities for manipulations are much greater.
ps: I am a programmer and know how the hardware and software works. I've created self learning / evolving AI from scratch.
another moral dilemma is. we live in a blame culture if an autonomous car causes a death who is at fault?
The company that designed it.
Feynstein 100 then software companies are now responsible for hardware degradation? no company will accept that responsibility so no one will produce autonomous cars, driver error deaths stay up.
Agreed, which will be the death of the autonomous car before it even starts.
No driver will accept personal responsibility for a choice made by his/her car when it was out of their control...
No automotive company will accept personal responsibility for injuries caused by their autonomous software...
smegskull that is a big IF, is the dead person a pedestrian ? another car passenger ? if its pedestrian, then it needs to be seen what caused it...well same goes for hitting another car. If the error is in the program then the company who made it is in charge, same goes if ur car has faulty throttle that stucks open, the car company gets the bill for it(or the manufacturer of the throttle)
The driver, There will always be someone behind the wheel ready to take control, If you fail to do that then the crash is your fault.
I'll be 18 next year, will get my driver's license, my generation will be last which will observe both combustion engines, electric engines, self driving and autonomous driving.
There'll be at least one more after you. Transition always takes a long time with big stuff.
@@DanielHatchman Things are accelerating so, so fast currently.
In 10 years it wouldn't surprise me if the majority of passengers cars were eletric. I'd be surprised if the majority are autonomous given there isn't the environmental driving force to overcome the cost like there is with electrification.
In a hundred years, people will think that our current generation is crazy because driving manually will be considered so insecure.
No, they won't.
We don't think that people at the start of 20th century were crazy because flying manually is more insecure than if you let the autopilot fly. People at the start of 20th century weren't crazy: the planes were only invented then, and they didn't have the technologies(electronics, software) to automate flying.
Same here, we didn't have technologies to automate driving, we needed a few decades for electronics and software craftsmanship to advance enough, so we can make it very cheap to automate driving, so people can buy it.
@@KulaGGin 1. You take this way too serious.
2. You compare apples to pears. Because flying manually isn't perceived to be so crazy if you are also driving cars manually which is still the case in our current times. Also, pilots are still trained to fly, they definitely can if they want to and there are also parts of flying which are still done manually even today.
My hope is that they'll think we're crazy for using cars at all, self driving or otherwise
@@martinmartin6300 "also pilots are trained to fly" so are people that drive cars what do you think driver's ed is for?
@@DavidGalvanwiz Sure, drivers trained to drive, too. Please read the comment that I have referred to. His argument was that pilots nowadays are flying on autopilot, only (which isn't true btw), and we don't consider former pilots to be crazy. The point that they are still trained to flight manually is just to debunk his point of view.
Veritasium
An element of marketing disguised as content.
Also, I agree with self driving cars. If one of them hits me, the manufacturer has MUCH deeper pockets, and they'll be responsible.
Cha-ching!!!
Eks calybur You aren't very observant are you? This wasn't disguised at all, plus it was content regardless of if it was sponsored. He made a good point. Just don't buy a BMW and it won't matter that it was sponsored
The man needs to eat something, doesn't he? So whats the harm in spreading some well supported and well reasoned facts (That actually fall right in line with his kind of reasoning and his real opinion) while taking some money for it. You watch this content for free, so take what you get or leave it. If its free, you are the product. (Or you are stealing it).
People stop using incorrect terms, it is a sponsorship, not paid promotions, advertising, or paid content.
that's like mixing up rape, homicide and manslaughter.
perhaps a less pejorative simile would be in order here.
The solution to the moral dilemma is simple in my personal opinion. The autonomous car should act selfishly and preserve the lives of its passengers at any cost. When you are theoreticizing it is easy to remove yourself from the situation and preach a grandiose view of morality, but when you are actually in the moment, and the threat to your life is very real, you act impulsively to save yourself. Autonomous cars will drastically reduce the amount of road accidents in any event, and the selfish behaviour might cancel itself out, for example if two cars are hurtling towards each other and both decide to swerve out of each other's way. Where a human driver is involved, accidents often unfold so fast that they do not have the luxury of being able to assess the moral implications. I would feel much happier in a car that was not always judging the value of my life.
Orrrrrr, it can just drive slower, more precisely, and at a safer distance from lead cars. That pretty much solves every one of these stupid moral dilemmas without killing anybody. It never has to weight one life over another, which is good because these aren't machines capable of qualifying life to begin with.
z beeblebrox totally agree
Also, in the future when the majority of cars are self driving cars/ all of them, they will be able to communicate with each other through wireless communication, they will know where they are, what they are doing (swerving or turning or what not), where they are going (to an extent of how it effects your car), and it will lessen the chances of two cars ever in fact colliding in the first place.
You're assuming a closed system where the only thing that exists in your path is the other cars. Have you ever had a deer or a child jump in front of your car? or a box that was in a pickup truck driving right in front of you fall out? It's not practical to leave such a huge space cushion all around your car. In a lot of the parts of the world such transportation space doesn't even exist from the high population density.
ash boogeyman The extremely early level of tech they already have in these cars has sensors for things like this, on all sides of the car, they can even detect moving things in the woods around the vehicle. Let alone 10-15 years in the future.
I've seen too many stories of autonomous cars screwing up to think it could be safer than human attention.
Hey Derek, just curious, what's your source for "airplanes flying in full autonomous mode are actually safer - studies show - than when pilots take control" at 2:55?
google
That's also because planes have more space to move around as compared to cars. In a car, you have to stay on road, and cannot drive on the walkway. Plus there are lot of cars ahead, behind and sideways, which isn't in case of airplanes so it's easy for autopilot to keep the plane safe
@@atharva1210 It wasn't about the safety records for planes vs cars. It was about safety records for planes on manual,and fully autopilot mode.
That's because pilots only take control during the most dangerous times during a flight which are takeoff and landing. Autopilot does very simple things like keeping the plane at a certain altitude at a certain airspeed at a certain heading or adjusting one of those values if the pilot wants it to. If anything bad happens during a flight the pilots immediately take control from autopilot so there is almost no opportunity for a plane to crash while autopilot is on. The only way it can really happen is if the pilots are unable to fly the aircraft, for example the 1999 South Dakota Leerjet crash happened because of a decompression which left everyone onboard unconscious and the plane ran out of fuel and crashed.
@@drabberfrog Actually autopilot is more likely to be used in bad weather than in clear weather. People have died because of pilots turning the autopilot off in clear weather to do it themselves and then screwing up. Autopilot is basically always used for landings during low-visibility weather.
should people still be required to go to driving school and get a drivers license if they will purchase a self driving car?
No this is Patrick depends if you will need to take the wheel or not I'm sure within 10 years you'll just hop in tell it where it go and it will take you no problem there may not even be a wheel in the car so I would assume at that point, no, a license would not be necessary.
Perhaps a lower license level that is easier to obtain, and restricts them to self driving cars.
Licence requirement might become something like airplane safety instructions - what to do in an emergency. How to brake, how to steer off the road safely and stop.
yes
yep, cuz if the driver needs to make a tough choice, they should be able to controll it
To all the fear-mongering folks preaching doomsday... Wow the ignorance in the comments section is almost unbareable. It's like everyone thinks "I, robot" is a goddamn documentary.
For you guys so adamently against self-driving cars. Think about this alternative scenario. Let's debate whether cars should be allowed cruise control, automatic transmissions, and sensors for auto braking. What about the moral dilema of letting current cars control our speed autonomously in crusise control? Your letting an AI control your gas and brakes, right? Or which gears to shift to with an auto transmission. What if it makes a mistake? What about the moral dilemma of letting our cars control distance and stopping when the car in front slams on the breaks and triggers the front sensors? "Cars should not automatically choose when to shift and when to brake, that's my right as a driver! And takes away my freedom! It's unamerican!" most of the youtube commenters might say...
What Im trying to say is that autonomous driving is just one more tool the driver "chooses" to use, with an expanded function to the convenience that cruise control, automatic transmission, and sensors for braking provide. Its a tool, it's up to the driver to use it correctly. That is all.
How about the moral dilemma of a third party taking control of your car? That is the concern, not increased safety and efficiency. I'd be all for it if there were no possible way for a third party to control any aspect of it. But when a device is automated and connected to other automated devices on a network, any party can gain access. Safety and efficiency goes out the window if an enemy nation gains access. These aren't planes on autopilot, we're talking about millions of cars on roads, all networked. I am rightfully concerned.
ShapelessHail 1984 was also a work of fiction... one which painted a clear warning regarding one potential future of humanity. As such, I, Robot remains in the realm of fiction if, & only if, steps are taken in the implementation of the system to prevent such things from ever occurring. To simply assume they will not is folly at best.
William Rumley this is already possible. We live in a world where, given a bit of time and some skills, a dedicated person can use the computers already in your NOT-self driving car to unlock it, set cruise control, disable brakes, etc.
Fear of cyberattacks is a good thing, but don't pretend like having you mostly in control of your vehicle makes you invulnerable to them. You can already be affected. Sleep well
Art Randall "You can already be affected." Yes. And you can also be hit on the head by a rock from space. Both events are technically possible; neither is terribly likely.
Big difference in how much you're letting the car do here. Braking control systems either don't affect your steering at all, or they activate only in emergencies.
A self-driving car could respond to a tricky moral dilemma by *detonating*, and they would still be safer than human drivers.
Exactly the point and I love how it evaded the question! There shouldn't be a question for such because the machine's choice is only limited by its ability in a situation.
So would he go to jail if the car did something wrong and accidentally killed someone?
Robert Gratto that's going to be very interesting. Surely the car manufacturer should be held responsible, but we all know that's not gonna happen... until a landmark case occurs, and THAT will change the industry forever.
Robert Gratto I'd assume so up until a certain point in time when people aren't considered drivers but only passengers because they're expected to take control of the vehicle at any time. The company would probably also be held responsible for flawed programming. That's just my guess though
I'd say the driver would be held responsible as the car company would put giant *keep eyes on road* stickers everywhere as to not get sued
By law i think you would but interesting enough only one accident with autonomous car has ever been registered
Yes indeeeed.
The interesting part about these questions and scenarios is I have yet to see anyone take into account that, unlike people who can't watch every sector within their field of view autonomous vehicles come with several specialized cameras designed to watch 360 degrees around a car at distances farther than a person can see and pay attention all at the same time thus being capable of seeing things that could potentially signal an accident happening (in other words predicting a possible accident), all this while being programmed to drive as a person is supposed to, a proper speeds, with proper distance from the vehicle in front, with nearly up to date maps and GPS and potentially programmed with how stop lights are programmed thus avoiding running red lights or even catching one.
Basically an autonomous vehicle can (or is supposed to be able to) see more than we can, all at once, without distractions, combining this data with the preprogrammed traffic rules, maps, traffic light settings and updated traffic warnings all analyzed is seconds and thus these scenarios can be avoided almost entirely.
Granted like any other new tech there will be bugs and thus these scenarios will happen but it's expected to have extremely low rates of happening and if it does then at this point the vehicle may not be able to make a choice, moral or otherwise because I think the scenario would have to be extreme enough that even a computer won't be able to react in time.
And if by chance a chance was to be made I think the autonomous vehicle should not risk others lives for the sake of saving the passenger. The vehicle should just take the hit and hope all the data gathered right before the accident will allow it to react in a way that can save the passenger and potentially prevent any hard or serious harm.
That's what I think is the way it should work even with the level of testing we have today.
>yay in 2 years I can finally drive a car
>2019: nope, you cant drive a vehicle silly human
in soviet russia.....
Nope don't believe the hype buddy hope you got a nice whip
I am a time traveler from the future. Self-driving cars won’t exist for a while yet.
@@tasmanmillen sad
@Maximal's Personal Profile you don't have to. If your city has it you could always use public transportation, or a bike. It would economically friendly for you and environmentally friendly
Hi Derek,
Until recently I was ... reticent about autonomous cars. Yes they have better reaction speeds/etc, but my track record on the road was perfect - better than perfect! I've not only avoided causing an accident in the ~15 years I've been driving, I've also avoided the accidents of other drivers.
But something hapoened a couple of years ago that convinced me to rethink this.
I was in 2 accidents in a 6-week span. Both were the other driver's fault, and both occurred while my car was stationary at an intersection.
What I realised is that autonomous cars aren't better because I'm a bad driver, they're better because everyone else is a bad driver.
I know that sounds stuck-up but it's not just how I think, it's how everyone thinks. Everyone I know thinks they're a good driver, and some of them Definitely are not. Chances are I'm not as good as I think either.
People don't think they need autonomous cars because THEY are a good driver, but it's not about them, it's about all the other idiots on the road who are putting them in danger.
Case closed. Hahahah. Yeah it’s a great argument for people, who think they might be better than the robot.
It usually takes the smartest guy in the room to point at the most obvious thing...
FunBotan You mean like "the numbers aren't comparable because the situations are different"?
No, I mean the winning strategy is so obvious there shouldn't even be a debate
So eugenics is a go? It is just so obvious there is no need for debate. Imagine a perfect master race of humans free of disease and with high IQ's.
How did this become a discussion about eugenism
easy for someone sitting alone in a basement to say. :)
Here is some professional advice (class a driver's license). People are concerned about who you would sue if this situation actually happened? The trucking company because the load on the trailer wasn't secured correctly by the truck driver! If the following distance between the car and truck was further apart, a collusion wouldn't be a problem if the conditions were good. Point being don't tailgate a semi. I want a self driving car and excited for the future! [:
I could go on and on about this topic haha
James Atomic can't sue the truck driver because there won't be one. It will be driverless. You would probably have to sue the CEO of the company that builds the robots that pack trucks.
Each drivers responsible for their own vehicles breaking distance and path, not the person they're following. I dont get why truckers bitch if they cant see someone behind them, its not like they're going to be in reverse on the interstate. Maybe they dont want someone bending their rear footsteps. Can a truck even outbreak a small vehicle? They usually have a lot more weight, but they also have a lot more tire contact.
First last, depends on each individual truck. Look up "amazing volvo truck braking' on TH-cam. It's amazing how fast the truck stops in the video.
first last, let me tell you what. I just did some work on a truck that was going 60 on the interstate and got rear ended. It tore the hood almost completely off the smaller car. They don't like tail gaiters because things happen that can kill someone and they don't want to be part of that.
As with most statistics, the interpretation of the data is key. Even if 94% of collisions are caused by driver error, you have to consider if the average driver's likelyhood of causing a crash is proportional.
If a very small percentage of drivers are "bad", and causing those accidents, the self driving car may beat the average in aggregate, while still putting the average driver at greater risk than they would be driving themselves.
I don't know if this is the case or not, but most of the statistics I've heard are pretty simplistic.
That's an interesting point. But the thing is, self driving cars are immensely better drivers than your average human.
Fascinating! A haptic adaptive interface...and we're a hundred years ahead of time!
Damn! That's a nice elevator.
2013 - 300,000 subs , 2017 almost 4 mil subs ... U guys have come a long way
I love this concept. I am going to write about it and see I can't improve on the idea.
You still cannot skip the tough questions. Choices have to be made, and it is not the program but the programmer, who has to make them.
Spot on.
This goes for medicine and tech developments in general I believe.
Where I live in Europe for example we have incredibly strict rules for the pharmaceutical industry. So strict in fact that the cost of development and the delay in getting new medicine out on the marker are killing a lot of people that are in need.
I get why its this way. Its a huge tragedy when unforeseen effects pop up after market introduction. It can sink a company. Any company.
But we need a larger perspective in our political discourse. We have overemphasized safety to the point of negligence.
You don't see a problem with having to worry about the negative health effects of drugs meant to improve your health?
+09erre I don't see and issue in most cases. Bear in mind that there are degrees of safety here. Obviously we don't want to be reckless. But if there is a badly needed drug there needs to be ways to cut corners to get it out faster in some cases. Patients being fully informed of course.
And it should be already understood that *any* medication can have complications. And any responsible doctor will tell you that.
The public might think differently though, and that is an issue I will agree.
the problem with 'cutting corners' for anything medical and bringing out a drug or procedure before it has been fully assessed, it is goes against one of the most basic principles of medicine: "first, do no harm".
09erre all drugs have negative health effects. You're an idiot if you don't worry. There's a fine line between discovering the negatives and beating a dead horse for another 10 years. I'm part of a drug trial for a sleeping pill, gentler than any of its contemporaries. Human testing started in 2006. It might hit the market in 2020. 14 years of human testing for an ambien alternative with fewer side effects before it becomes publicly available. People die every year from illnesses with safe treatments that have been in testing for over 10 years. You don't see a problem with that?
Serah Wint If you truly cared about getting the right drugs to people faster then you would be creating a movement, getting signatures to reform policy, or trying to get elected to make that change.
Not wasting your time on a TH-cam comment.
It's facinating how peopole are unwilling to have a machine make decisions for them even when proved it is safer.
Martin H eh, freedom, free will, etc. Turning all of society over to machines would prove safer, driving, flying, government, business decisions, etc. Wall Street Corp just put an AI in charge of HR, makes all the hiring/firing decisions. Humans aren't necessary for very much, anymore. Classic slippery slope fallacy, but how far is too far?
In a different video, you talk about thinking and Drew's inability to hold too many pieces of information. As a middle school teacher, I've had students ask if they can us the calculator, and I always say no. Because we know that young children need to learn calculations so well that it becomes automatic. Every year, to prove my point, I would challenge a student to come up with a multiple digit multiplication problem. Another student and I would race; he/she would use a calculator and I would do it in my head. I almost always won. But I also practice. I know; I'm a geek. My point here is that if we allow drivers, especially young ones but really any age, to release their responsibility to the vehicle, then those people will lose the ability to analyze road conditions and make smart choices. I predict that we'll see this first with lane departure warnings. People will get used to hearing a beep, until it becomes part of the background that they don't hear it anymore. So yes, for a while you'll see fewer accidents, but I think in the long run it will backfire. Someone needs to put cell phone jammers in cars so that as long as the tires are moving, the cell phone won't work. I think that will do a lot more to reduce accidents than an autonomous car.
Hey Veritasium, you are one of the only youtubers were I give an thumb up, before watching the video. Heven't seen a bad video from you. I really appreciate what you are doing here. Never stop doing that!
Literally everything in this video is speculation based on nothing.
ChaosmanOne umm no
0:55 - How much should you trust an autonomous car?
Seeing as computers _literally_ can't perceive reality (relying SOLELY on pre-programmed instructions to navigate), this is like letting a blind person drive who is being given second hand details about their surroundings. Sure, computers can do this really, really, quickly, but this is basically what you are doing. So, no, I don't trust them, and I don't think you should either.
1:01 - This car is now driving itself.
No, not really. It's on a highway with no obstacles, no lights, no stops, no turns, no decisions, in perfect conditions (and I mean perfect), going in a straight line. As soon as that "self driving car" has to do any kind of real driving (a weird line on the road, rain, stop lights, wrong way signs, potholes...literally anything), it will fail.
1:35 - We can imagine a lot of situations where an autonomous vehicle has to make a tough choice, but the real moral dilemma, is accidents are happening right now!
False dichotomy, otherwise known as a false choice or false binary. Veritasium claims to be intelligent and should KNOW that this is reprehensible! It is impossible to "program" choices for every possible circumstance a car might encounter, so the REAL question is, "What is the moral responsibility of programmers with regards to collision avoidance programming?" Eventually, far in the future, IF this technology works out, there may be fewer collisions in traffic, but for now, this is a complete unknown.
2:00 - So, even this level of technology we've demonstrated today, autonomous driving on a highway (literally the only thing this technology can currently do - _kind of_), could save a lot of lives.
So, after pointing out some scary statistics (lies, damn lies and statistics), Veritasium blurts out something purely, 100%, utterly, speculative about a technology that literally doesn't exist yet. That is probably the most dishonest thing I've ever heard him say. Gross. It's like saying, "This new teleportation technology (after demonstration of teleportation of a photon across the city of Calgary in 2016), will eliminate the need for cars altogether, and save a lot of lives" *smug grin* Idiotic.
2:19 So, if we're not driving, we'd better hope that the tech gets to a level where the car's can drive for us.
And there's the fact. This tech CANNOT, currently, drive a car. Period. It just plain can't. Honestly I wish I could say that this was not so, I really do, but I can't and won't lie to myself just to make myself feel better...it's the same reason I'm not religious.
2:30 - 3:02 False equivalence. I'm seriously starting to doubt Veritasium's intelligence. Sad.
3:12 - The longer we wait to get autonomous vehicles on the road, the more people will die.
Silly. We're not waiting. The technology LITERALLY doesn't exist! "The longer we wait to get hyperships in space, the fewer extra-solar planets we can explore!" Total, absolute, nonsense.
I guess Veritasium just spewed what BMW wanted him to, because he sure as HELL didn't think about anything he said in this video!
ChaosmanOne 30,000 people die every year in car crashes. That is not speculation. 94% of those are human error. That is not speculation. Trolly problem like scenarios are incredibly rare in the real world. That is not speculation. If the cars make errors at rate less than that of humans, fewer people will die in crashes. That is not speculation. Where exactly is the speculation you're talking about?
how do you know you can perceive reality though?
ChaosmanOne your argument falls flat right out the gate. Self driving cars aremnot preprogrammed. Google machine learning and deep learning. These machines are learn from the environment, they are not preprogrammed.
I'm fifty-two. I'm hoping that by the time I am no longer safe to drive a car, my car can drive itself.
There will come a time when humans will not be allowed to drive cars - it'll be morally irresponsible. I'd give it.... 15-20 years.
One thing about self driving cars which I think is important to consider (whether you care or not is up to you) is where the motivation for their development is coming from. What car companies tell the public vs what they tell their investors as to why it's worthwhile for them to be researching and developing self driving vehicles are two separate things. The message to the public is basically what you hear in this video - the technology has real potential to save lives (and, not mentioned, to lower emissions of transportation by decreasing traffic). Those benefits are real, and while it's certainly a benefit for businesses pursuing the technology, it's not the main reason. Obviously the main reason is profit, whatever, that's the motivation behind any business venture. The problem is that most of the profit comes from data collected on the passengers of the autonomous vehicle - data on where people are going and when, most broadly, and this information can then be sold to whomever - retailers, advertisers, marketers, product planners, financial analysts, government agencies, anyone willing to pay for it, and who can then use that information to build profiles about you. It's nothing new, it's the same thing Facebook, Google and basically every other internet business does to internet users, just with information about your physical habits rather than digital. Whether you find this collection of personal data to be an unacceptable breach of privacy or not is up to you, I don't really care. I just think it's worth thinking about.
Here in the north, when it is snowing, edges of smaller roads become sometimes very hidden, and one must be careful not to drop off the edge or to drive on thick snow. I think it would be even harder for automatics to find where in the middle of reflective snow the road goes. On the other hand, in good conditions in areas with relatively stable weather, self-driving cars would help a lot. And in big warehouses they have successfully used robotic wagons to move stuff around for long time already. The problem is, you would have to remove all the people from roads to make it safe for self-driving cars, and get rid of the problem about who is responsible of what and who has to pay for who when something happens.
You hit the nail on the head. I'm all for self-driving cars soon.
Not likely.
+ChaosmanOne, it's already happening. Uber's trying to replace all of their people with self driving cars in the following years. They're not programmed correctly right now, causing it to be delayed, but it's only a matter of time until self driving cars replace manual ones. That can also be applied to jobs in general with the dawn of automation.
The problem with adoption isn't going to be the tech, that will fully mature quite soon. The delays will be caused by errant public perception (and any dumb regulations that result from it) and the fact that cars are major, expensive purchases. The average car in America is 10 years old, so even if we rolled out self driving cars TODAY it wouldn't reach a majority of on the roads cars until the late 2020's
Wordsnwood (Art Mulder) eh, I just hope they're mainstream in 2 years (I'm 14)
Chandler Short
Grow some balls kid. Learn to drive, it's fun.
I just now noticed that the atomic number of Veritasium is 42.0 (420).
I find that Amusing.
pretty sure its 42 as in the meaning of life.
That was a pretty cheap conclusion. I thought he'd talk about tradeoff between risk aversion and speed. In the future it's entirely possible for "jailbroken" cars - or different safety preferences to compromise everyone else in the road. Being autonomous also requires uniformity
It's the exact same as our current problem of drug users and distracted drivers at the wheel. People will probably hack their cars, and they'll probably be punished if/when those cars cause accidents.
I can guarantee if uniformity becomes mandatory, there will be a booming market for car guys like myself who want optional manual control. Autonomous vehicles should be like smart cruise control, and nothing more. Life is about a balance between "safety" and freedom. And because I trust myself more than I trust everyone else to keep me safe, I choose freedom. And so should you.
@@danielhebard1865 So why even live in a society? Go out and live in the woods if that's how you feel about others.
jamrenzee I’m not misanthropic. I’m just not a collectivist. Humans aren’t ants; we don’t ever sacrifice self for the “greater good” of people we don’t even know. Small communities experience altruism, but this falls apart when you apply it large scale. The fact is the freedom to drive your own car should not be given up lightly; this “but it’s for the greater good” is exactly how a people become disarmed, have their free speech taken away, and ultimately, lose their freedom entirely.
@@danielhebard1865 I trust myself more than you to keep me safe. You should be delineated to a self-driving car along with the rest of us.
i love how this video has lots of comments about autonomous cars being ready and a MUST, but then the recent video with the Waymo car all the comments are about it "being an ad" maybe you should do a video about that kind of human behaviour Derek lol
The idea of autonomous cars used to scare me. Then one day I was driving to work and glanced over to see the driver next to me changing his pants while driving! A few days after that I saw another driver reading a freaking newspaper while driving!
Those two instances crystallized in my mind the reality that most people don't like driving and would rather not do it. Sure, they love the freedom of movement it offers, but they don't like actually driving. They would rather be changing their pants, or reading a newspaper, or texting their bff, but instead they have to drive right now so they "compromise" and do those things WHILE they drive.
A computer by contrast may make slightly worse decisions in a pinch (though I'm not even convinced of this anymore), but it doesn't have anything better to do than drive. The computer driving your autonomous car will never be distracted by its pants, the newspaper, or its bff. It will pay attention to the road and only the road, and so even if its decisions are otherwise not quite as good as yours would be (which, again, I'm no longer convinced is the case) it will still have fewer accidents.
Self-driving cars are already beginning to exceed the capabilities of even attentive drivers. Then you consider the possibility of them talking to each other, never getting lost. You mention distracted driving. but self-driving cars also never get sleepy, they never get road rage, never drive recklessly, never drive drunk, never crowd people in a lane or speed or even blink. They also have access to information humans do not and don't have blind spots.
In addition to safety these cars give benefits to those not able to drive, and commercial befits like cheaper cars and shipping prices.
Good points. There is one particular concern for me though, and that is the accuracy of the information being fed to the car. Take maps for example: My GPS once told me to turn off a bridge into the Chesapeake Bay. Obviously I didn't, but how would an autonomous car respond in that situation? Presumably, it's other sensors would tell it that it was not safe to proceed in that way, but how confident are you in taking that chance?
Swimming sounds fun.
Self-driving cars have decent collision avoidance. It certainly would drive you into a lake. these cars only rely on GPS for directions anyway, GPS alone is not acreate enough to keep a car on the road.
really the biggest issue of these cars to over come is their ability to visually recognize objects. Image recognition is easy for us. But for a computer telling the difference between a toddler, some birds, or a newspaper in the road is a very difficult thing for a computer to do. But these are issues that are being worked on right now. Self-driving cars are on the road already, with humman drivers at the moment. It has become a common site to see self driving Uber cabs in my home city. Everytime these cars, make a mistake, is and opportunity to work the kinks out of them. Results are already showing, that they are driving better than human drivers.
You ask whether it's worth the risk, but the thing is when you drive your car normally your already taking a risk. Self-driving cars will never be perfect, there will always be accidents. but in the near future, taking a ride in your self-driving car, will become far less risky than driving the car your self. it's just a matter of time.
"It certainly would drive you in the lake. "
Oh okay... wait WOULD?!
Fun!
Also, accurate. A create isn't a word, but does make more sense.
And yeah, I think self driving cars a great. The only reasonable arguments I've heard against them are for rural areas, but I don't know what the data on that is really. If the road fades out and has plenty of pot holes, I'm not sure if a computer now could decipher that there's anything useful out there.
Being a very old guy, I understand intellectually that letting the car drive itself would be better, but I don't think I could ever trust it.
And that's what's interesting. Humans don't like being safe, they like feeling safe.
HugoH What's the difference when you only know what your senses tell you?
+R9000
That's exactly the purpose of Science. We attempt to determine what is better based on observation, mathematics and statistics. Its the same reason we put very little weight on a single person's perspective as evidence when we have much more accurate methods available to us.
I think the difference is not "feeling safe" over "being safe" but "Control of whether I live or die is in my hands" over "whether I live or die is in a computers hand"
+HugoH So true, from a scientific point of view Nuclear energy is as safe as other forms of energy production. Still many people are afraid of them.
The comparison with planes is a mistake: The initial and continuous training to be a pilot is far beyond the training or a driving licence, so it's not a "full autonomous plane" in the sense that there is no pilot.
There are some experiment to have 1 pilot only in some configuration (because the co pilot would be on the ground) but so far there are 2 to 3 pilots in each flight, ready to take over at any moment there is.a problem:
The plane is not autonomous, it's just able to face expected situations without human input, but autonomy would be deadly in any edge case. Would you take a plane if it crashes in the rare but not absent possibility that there is any internal malfunction or external issue like birds, weather, etc?
exactly. There IS NO such thing as an autonomous airline plane as of today. Yes, automation makes it safer, but it's still not yet at this point that in case of some error it knows what to do / how to improvise.
Except that pilots can only train so much. you could run millions of simulation for every scenario that has ever happend or can be imagned, take months to analyse the situation and have the system do the right thing. This is on top of effectively perfect concentration at all times, millisecond response times and constant, detailed communication with any nearby planes.
This may not exist today but it could and it should. And when it does it would absolutely be the right thing to do to implement and use this technology. The reality is that air travel is already so safe that there's not great urgency and groups of people are dumb so they they'll resist it.
I also think that the pilot's salary is negligible compared to the rest of the operating costs. Maybe once we start having small, electric planes for short distances this will become a bigger factor.
And yes, I would absolutely trust a self flying airplane.
@@Hasturoth machines train for known possibilities, humans train to adapt, improvise, overcome.
They're complimentary, my point from the beginning is that if you remove 1 of the 2, you weaken the whole. And it's the same for cars: Adding "intelligence" to a car could make it safer, but replacing the driver is suboptimal (meaning people are gonna die).
There is currently 1.11 deaths/100 million miles in USA. Tesla claims that under their autopilot it's 1 per 320 millions. But the autopilot disengages when it's too complicated, so they're basically saying that they have 1/320 fatality rate in good conditions.
But the important part here is "the autopilot disengages when it can't handle it" So you still need a driver, in full capacity, at any time.
I rest my case.
@@Hasturoth You're absolutely right about the theory of all this. I agree that in a perfect world, with a perfect sytem, the self-flying airplane would be the best - in an environment with other self-flying airplanes.
But I think the world and especially a technical sytem like an aircraft filled with humans, is too complex for us to handle completely automatically.
It's the same reason why cars aren't already driving completely autonomously. Yes, they could do it on a straight highway with other computer-cars in good weather with perfect lane lines.
But the world is complex, unpredictable.
I don't see a self-flying aircraft in the next 100 years, considering the extremely conservative introduction of new technologies into aircrafts and other systems.
Think about it. Why do we still have human ATC personal? Why do they talk on analoge radio?!
Because it's complex. Really complex.
When do you go around? When there's a vehicle on the runway? What if you'Re really low on fuel and only have another approach left (maybe)? What if the vehicle is driven by a human being who could decide to leave the runway just in time for you to land on it? What if you actually speak to the guy in the vehicle? What if he is suicidal and keeps on driving there? What if it's accelerating and you only have fuel for this one try? What if your plane is full of 500 people, what if it's empty or just a little cargo? Do you risk your life, do you risk 500 souls?
i support autonomous driving, i just have one concern with it's mass implementation-and it's human related. Those sensors will need regular maintenance, especially optical cameras that can be affected by dirt and debris (i know most of the sensors are LiDAR based these days, but they still require cameras too). Look how many people drive their cars despite warning lights, or clear problems with the vehicle that needs repair, and refuse to take them in. Either they don't have the time, or the cost is too high-and once the car's sensor system becomes compromised it's no longer as safe (and depending which systems go down, it could make it extremely dangerous). And realistically, there is no way to enforce ensuring people get their proper regular maintenance on those systems (the only 'surefire' way would be to have the system detect it's own flaws and lock out the driving system from starting once it gets bad, so you can't drive if it knows sensors are down-but locking people out from using their own purchased equipment would lead to a lot of legal issues).
THERE ARE NO MORAL QUESTIONS FOR DRIVERLESS CARS. I keep seeing this very annoying meme pop up all over, that driverless cars will need to necessarily require us to tackle moral 'trolley car' problems. This is absolutely false. The people programming driverless cars do not include any moral behavior or choices in their algorithms. The false choice presented in this video for instance, "will the car swerve left into a station wagon or right into a motorcycle" is NOT an actual dilemma. The car does not swerve left or right into obstacles, EVER. The car does its best to avoid all obstacles while braking to minimize the collision. The car and its occupants will be the ones to bear the brunt of the accident in every situation. Only a human and their ego would even come up with such an immoral idea that they should swerve left or right and take the life of a station wagon driver or motorcycle rider in order to minimize their own risk, but I guess it makes for a good click bait TED Talk. When you get into these cars YOU assume all risk, it isn't mitigated out to everyone around you just because you are about to get in an accident. If the car cannot stop itself from collision it does not swerve into another different collision. Veritasiums point is different obviously but the meme concerning moral dilemmas of car AIs is nonsense.
fullyawakened I like this argument, but it doesn't consider cases in which swerving would cause less harm to others than simple braking. For instance, what if a person ran onto the road and the lane adjacent was occupied by another vehicle? To swerve would likely cause property damage, whereas to only brake would cause death to the pedestrian.
but, you are not considering the actions of the person running into traffic. The AI should take the least hazardous action to what it has control of. that would mean only braking. A person stepping into traffic assumes their own separate consequence of action, separate from the vehicle/driver. If the their action led to their death, they have suffered the consequence of their choice.
Everyone also seems to forget that if all cars are autonomous, then the vehicle that detects the danger will communicate with all neighboring vehicles, causing all vehicles to respond optimally to the situation. In this particular case, the car will swerve to the right and mere milliseconds later the biker will gently steer to the right, creating enough space within ample time for all vehicles to safely resume their course.
This system can even by tied into other devices with location/GPS tracking, eg. a mobile phone or a sub-dermal device. This way it can even detect that a pedestrian is getting dangerously close to fast moving vehicles, causing all vehicles to recalculate their paths so that no cars get dangerously close to the pedestrian.
but everything should be design to protect people from their own stupidity. like why does balconies have guardrails ? if someone slip and fall from the 15th floor its his fault for not being cautious. autonomous cars should the same if someone fall in front of the car and it just breaks, the guy that got hit can claim that he wouldn't get hurt if the car turned to the side like a human pilot would do instead of just breaking, so who is to blame ?
Oni Raptor But the AI also has the choice of changing lanes and avoiding the pedestrian. According to Isaac Asimov's 1st law of robotics, "A robot may not injure a human being or, through inaction, allow a human being to come to harm." It seems crazy for a human to choose to hit a pedestrian on the road just because it's not the driver's fault that the pedestrian was there, so why would that be acceptable for a robot? And would it matter if the pedestrian ran across the road on purpose or if they were pushed or blind or disabled or a child? In any case, it's pretty sick to run them down when you have the ability not to.
I've been getting really tired of people doing trolley problems instead of realizing that automation in driving is still objectively a good idea. The fallacy of salience is super annoying sometimes. A semi-automated car getting in an accident is more newsworthy than the thousands of non-automated accidents happening at a given moment.
I feel like we just enjoy feeling like we have control over our situation, even if that very control is a myth of the mind.
A really interesting observation. The trolley problem can be a bit of a red herring, but it's important to still remember that other things are at play: what if a sudden availability of cheap driverless taxis leads to people abandon walking and cycling? The increase in congestion, pollution and physical inactivity might not be as shocking as road deaths, but it could still cost years of human life
Remember the blackberry days when your device froze and required a battery pull? What happens when the cars computer freezes?
your car gradually slows down????
it will give you control of the car
you restart it. Just like present-day gasoline car engines.
What happens when the autopilot in planes freezes? What happens when the computers that control the stock market, moving trillions of dollars, freeze? What happens when the algorithm which controls your building elevator crashes? What happens when the computer that controls the power grid that goes the electricity supply to hospitals crash?
Computers have faults and fail sometimes and, spoiler warning, some people will die. The point made by Derek in this video (and many others) is that human drivers are *already* making catastrophic errors, all the time. If you want to compare it to your Blackberry freezing up, think about how many people literally fall asleep at the wheel, leading to thousands upon thousands of deaths.
Instant death
I would imagine the transition period is the most challenging one with automated and people driven cars. After there are mostly automated cars they can communicate between themselves and coordinate so that there is less risk of accidents.
so with the ethical question it showed, 1:34
What would an actual human choose
the fact that it had to ask what to choose makes it clear its not easy for humans to solve it, so why are we making machines do it for us? they said, what should a machine choose but the real question is what should a human choose? and what would they choose
and that is the reason why it's a dilemma, there isn't a absollute correct answer, some people, have problems with that, because we are letting that the machine take that decision for us, no matter if it take the worst of the best desicion
Andres Perez
We do that alraedy
So, absolutely nothing is changed. There isn't a correct answer, so why is it a problem that a computer is just as incapable as a person of coming up with a correct answer in a hypothetical situation that a properly behaving machine should never but itself into in the first place?
The sad thing is that a self-driving car will do everything it can to avoid ever having to make an impossible choice. It can always calculate the result should something truly unexpected happens and make sure it always has a good "worst case scenario." The moment the car realised if it needed to make a sudden stop it would be screwed, it would start making changes to prevent that future.
Austin Pinheiro Its simple, you swerve into the suv instead at putting the motorcycle driver's life in immediate danger. A fender bender for a car is nothing, while a fender bender for a motorcycle could mean serious injury. The problem with this is will a computer be able to choose between the lesser of two evils? If i was in that situation, I would swerve into the suv to cause as little collateral injury as possible.
I completely concur with all that is said in this video. Concerns over how a computer might juggle the choices between, as the thinking goes, it will prioritise when endangering 3rd parties, is unrealistic - i dont suppose any algorithm will be based on 'whom should i kill first?'. It also overlooks the fact that such decisions are already being made, by humans.
My only concern about autonomous cars is that it will lead to me being excluded from having a right to drive at all, on the basis that you cant trust humans.
two answers to that:
a/ because humans need to be able to enjoy their lives, and driving cars is something enjoyed by many.
b/ by saying "risk so many lives" you are grossly overstating the risks posed by human drivers, to others.
When you remove deaths in single car accidents where the driver is also the victim, risk takers on motorbikes, and pedestrian deaths which are COMPLETELY preventable by pedestrians not walking in front of cars, then the % of all road deaths caused by an 'other' driver is quite small. Plus, the risk of being killed in a car accident is tiny in first place.
so, yes, there is a risk in having humans drive cars, but it is a tiny one compared to a range of life's risks, and is entirely justifiable as a legitimate form of life enjoyment. this really is only an issue of importance for safety nazis who care nothing about the broader benefit to society. if avoiding any risk at any price was really worthwhile, then it would be illegal to go on a frivolous drive on the weekend for sightseeing. and we would have 40km/h speed limits on motorways. instead, the vast majority of people are happy to take some risk, in return for pragmatism and life enjoyment.
lastly, exactly the same tecnological aids which will make driverless cars less risky, can be operational to some extent when cars are piloted, even further reducing the argument for not allowing people to drive cars.
If you want to drive, do it at a racecourse or something similar. I would much prefer autonomous traffic to humans behind the wheel.
Alexander Stollznow I agree with you because the algorithm isn’t based on who should i hit, but how do i avoid everyone.
I dont think you guys understand that humans dont want complete safety, if it would mean giving up any kind of enjoyment or freedom.
One reason why people will play sports and watch sports is because of the humans doing said activitys.
Humans want entertainment and gladly take some risks to have fun and get memories and what makes life worth living.
"If you want to drive, go to races" is a really naive way of thought. The same would be: "If you want to walk why not taking the safer option and walking on treadmills as being killed or injured of cars is extremely high, *especially* that we have massive distractions around us *all the time*
What I want to say is, Humans will take risks to enjoy their lives, to enjoy things. To have their own reasons to live. Because I for sure, dont want to live in a world where I cant do *anything* because its deemed "not safe", resulting in taken things away that make life, worth living and endure through to your death.
Why do you think people 100 years ago danced to Jazz, even tho it was forbidden, prosecuted and socially unnacceptable. They took risks to enjoy things they like. Thats what makes human Human!
@@tramachi7027 I couldn't agree more.
7 years ago. Now what. Where are we with this?
I think thats a stupid analogy. Any school goer already knows there is in fact an equation of motion. I bet a driverless car takes into account its speed, brake force, and gap distance. Its either possible, or that its already being put into the programming that the car should always keep a gap in front thats enough for it to brake if somehow the vehicle in front immediately stops moving as if it has an infinite brake force. And I bet the safety factor for that is not 1
As an airline pilot, I don't agree what you said about "Airplane flying in full autonomous mode are actually safer than when pilots take control". First of all, compare plane's autopilot with autonomous car is a stupid comparison. Pilots are trained professional, and the training includes how and when to use autopilot, limitations about autopilot. And pilots are there if autopilot are not doing what the pilots expected. "Airplane flying in full autonomous mode are actually safer" that's because it gives pilot more mental capacity to deal with whatever happens inside/outside of the airplane, that's what we call "situational awareness". Autopilot does NOT relief the pilot from flying the airplane, the pilots still holds the final authority and responsibility. Autopilot only takes the "physical part" of pilot inputs away by using some buttons and knobs. Pilots are still there to make sensible and flexible decisions where the autopilot is incapable of. It's not "switch the autopilot on, then we are off duty" sort of thing like many think about what autonomous cars should or could do. Believe it or not, many pilots out there can hand fly the airplane as good as, if not better than, what autopilot can do. And there're some scenarios that hand fly the plane are MANDATORY. Are you dare to be a passenger on pilot-less plane? No, I do not, because when it comes to judgement and decision making, real human beings are still the best we have got.
Study > anecdotal. He said, studys show.... Furthermore, you can exchange every pilot/plane in your comment with driver/car and it would tell the same story.
Also the times when pilots are most likely to hand fly are times like take off and landing, which are just more inherently dangerous than cruising, when autopilot is used more
Completely Correct! I am looking to be in the airlines one day, and have been on a simulator for over 1,000hrs! In a 737-800, 727-100-200-200F, 737-200, etc... Older aircraft are the best! They have more responsibility than the 777X, which uses touch screens and easy-to-use navigation. Back then, you had to input your coordinance (Lat, Long) in order to properly set your current location. Nowadays, the FMC automatically calculates where you are, you still have to put in KLAX --> KSAN. Something like that. It automatically inputs waypoints, VORs, DMEs, ILS Landing Info, etc... Back in the 50's when we had the best airliner ever made, the 727, you had to pull a chart out the day before the flight and plan your flight manually.
I remember way back in the early 90s reading what was then an already old sci-fi book dealing with exactly the same dilemma of self-driving smart cars. IIRC it was set in a smart-car retirement village. Mind you, those cars had achieved self-awareness so somewhat alters the outcome.
It´s about whether or not someone would buy a car that can “consciously” choose to sacrifice them if necessary…
@Sune Wallentin Goettler Those are purely academic situations brought up by anti-self-driving pundits, as there is (to my knowledge) literally not a single self-driving algorithm in the world right now which deals with trolley problems. The answer to 99% of those problems is brake to minimize all damage, or don't go so fast that a moral dilemma will happen. For some reason people just love fantasizing about moral dilemmas but the chance of it happening in real life are almost nil.
it only takes one coding error, one bug, one hacker, for 10 million self driving cars to crash
dude. every technology, every innovation, every invention has a plot hole.
the one who has designed even the best security system will know all the ways to break into it. If you are thinking in this way that a bug can disrupt all the cars, they will have a failsafe. They always have those.
@@jaitirthpage1041 chernobyl had failsafes too lol. my point is rather than a single car jsut breaking, tragety but managable, an entire nations cars could fail catastrophically.
@@Kotih by the way the workers in Chernobyl reactor 4 were told to disable fail-safes. They couldn't turn them back on.
Thank goodness! I am a rather awful driver and I'd love to have a self-driving car that just eases my anxiety a little.
So good to hear someone actually ADMIT they are not a good driver! Thank you!
Everyone pretends like "oh it could never be *me* in an accident because I'm a great driver!" when in reality we are all human, we are all prone to distraction, letting our minds wander, being more tired while driving than we will admit to ourselves, etc.
So thanks for being honest with yourself and others! Stay safe out there!
@@rrrrrfffff I will tell you more. I consider myself a very good drive, but I would be so happy if autopilots take control. It is not just it is gonna be better than I am (surely, it will), but also everyone else around will be much better, so the scenarios where you just can't do anything to save yourself or react will be nonsense. It is gonna be hundreds, maybe thousands times safer than it is right now. Sadly, the majority of people don't understand simple facts, are terrible when statistics show up, and in general, are morons =(
Excellent point! Despite the risks, autonomous cars will save more lives than they lose. We need to put the risks into perspective.
Well the reason is that the ai is incredibly hard to develop. Autonomous vehicles would be perfect if they replaced all vehicles on the road at the same time, but as you mentioned, driver error and reckless driving happens a lot, making it incredibly hard to predict what to do. In common situations such as someone changing lanes without a turn signal and cutting people off, it's really hard for an ai to safely manuever the situation.
The only reason I'd like to have a driverless car is to get my drunk ass home.
Hoss Cartwright call a taxi
Unpaid parking ticket? Government shuts down your car until you pay. Warrant for your arrest? Car locks you in and takes you to the Police station.
yeah.. in 2019 revolt motorcycles launched their first electric motorcycle in india.. it has a 4G sim built in. so it is connected to the internet almost all the time.. the matter is.. you have to pay the price of the bike in installments.. and If you fail to pay an installment, they can remotely shut down the motorcycle untill you pay the money... okay that's a good thing.. no money.. no bike..so what my point is... in the future.. this can be done in the other way.. just remember the car hack scenes in the movie *The Fate of the Furious* .. and also from the movie *Fast & Furious 6* when Dom demands cars without electronics..becoz their modern BMWs were ruined by somthing that looks like a hockey puck...🧐🧐
Probably someone already stated this, but I'll repeat it anyway. ;-)
There is a statement, that the real moral dilemma is not if the programming for tough decisions is right, but how soon we allow self driving cars on the streets. This is only true, if ONE human life is considered "finitely precious".
If a human life is considered "INfinitely precious", one human life is equally precious to 1+n human lives. Therefore the "really real" question is: How precious is ONE human life?
I feel like the real solution is changes in city design, reducing dependence on motor vehicles to do daily activities
Veritasium had a good segment on probability, my observation should have been on his mind when he made the comment that planes on autopilot are safer. This is a poor example. Pilots put planes on autopilot when the probability of things going wrong is lowest. Planes crash on take off and landing. Rarely do they crash in between when pilots put them on autopilot.
What percentage of crashes during takeoff and landing are due to pilot error?
Phoenix autopilot is almost never used on landings as it causes more crashes, Humans are the safer option for landings and takeoffs.
@Jonnym610 Sources for your claim? AFAIK there were only two serious accidents due to autoland. And autoland is used under very specific conditions and must be controllable at any time by pilot what further restricts its use and reaction time.
The main hindrance for implementation of full autopilots is the insurance industry, not technology.
Many years ago, I was driving along the Nimitz Freeway, I-880, near San Leandro CA. Suddenly, a truck wheel appeared right in front of me, rolling towards me. It was impossible for me to react, much less figure out how to react. The end result was me hitting the wheel square on, vaulting me up in the air and landing, bursting all four of my tires as I skidded, luckily, to a gore point. There were no other vehicles to the sides of me, but other cars did come up behind me several seconds later.
Had this been an autonomous an autonomous vehicle, (a) it would have been able to identify the oncoming object sooner (it was dark and rainy, but LiDAR can see through rain and zero ambient light), and (b) it would have been able to swerve properly without causing a worse rollover.
But even in the case that it couldn't, the moral decision for the AV is to protect its passengers, not the nearby motorcyclist or SUV. If I paid for the use of an AV, that AV should protect me. So, that truck that dropped something off its back will have its plates photographed and sent to the cloud. When all's said and done, that truck's liability insurance will pay for everyone injured, including either the SUV or the motorcyclist or both plus others.
Hey its cruising along in Vegas!
when we switch to self driving cars, the man will know our every move.
I want the world to just plunge into full automation. So much stuff would be better and safer.
ScreamsoftheDead We have a long way to go until that's remotely practical though.
I don't think the "real" dilemma is getting self driving cars on the roads faster or quicker. The real dilemma is who do we blame when accidents happen?
Without self-driving cars, liability is extremely simple. The driver is in control of the vehicle and therefore takes the blame for making a mistake that causes an accident. So what happens when electronics and computers make a mistake? Who takes the blame? Importantly: what are the different types of mistakes these computers can make? Sensor errors, misinterpretation, miscalculation, glitch/crashing, probability/statistical errors, what else?
Also, you only mentioned accident statistics for human controlled driving. I'm not sure why you didn't mention accident/fatality rates of self driving cars. That seems to be a huge piece of the puzzle left out that we need to weigh and consider. The psychology is simple: We like to BE in control. We'd much rather be in control of our fate while taking higher chances of death/injury/etc., than not be in control and effectively "roll the dice" of being "randomly" injured or killed even if the chances are much lower.
No Google, I don't want to use my real name. I don't agree and no amount of statistics will ever convince me otherwise. Autonomous cars are a bad idea! I'm warning you all now: there will be more accidents and incidents than ever before! California was actually smart this time around!
You kind of answer your own question. If the cause of the crash is because of the computer / car itself it is no fault. But you should be excepted to take all responsibility since you are behind that wheel and should be ready to take control whenever needed.
drumman22 But if the machine is the one that gets the person into that situation. How can you blame the person if he didn't get himself into that situation. Whereas if a man has full control over the car and is so stupid to get into that situation then it is his fault.
drumman22 So how can you tell that the computer made the error and not the driver himself? The car most likely won't register or "know" the error (because it would have prevented it if it knew that this was an error). Also if the human always has responsibility it can (and will) lead to lazy built cars and/or errors not being found because the error itself will never be associated with the computer.
The number of self-driving cars is still very low.. and yet, you hear about them being involved in accidents.. therefore, the accident rate must be incredibly high, compared to other cars right now.. just because there are so few of them around right now.
It's such a good point. We could save thousands of lives with this technology but instead we focus on a one-in-a-trillion trolley scenario.
Hi Veritasium, greetings from Germany. I´m just finishing my Examen on "automated driving from an ethical perspective" and wished I found your video much earlier. Thx for it:) It seems hard to overestimate emotional influences on this topic. But as you mentioned the elevator is not the best comparism. I think you gave the strongest counter argument by refering to the trouble of responsibility. That´s a puzzle with no potential of a just solution - not for the programmers, drivers or gouvernment from my point of view. It´s also problematic to forge an argument by the numberizing people. Even when you decide to decree with political power, where should be the line of acceptance - even if every fatality is one too much without question. I would like to here from you. Very much enjoy your stuff:)
The real moral dilemma is why we let stupid people drive cars and cause accidents.
Because if we didn't there'd be no one on the roads...
sm901ftw He meant getting self driving cars :/
Because we built so many cities in such a way that there's no choice.
Because making stupidity illegal would be seen as bigoted.....
Exelius that's ridiculous. If I want to drive my car I will do it.
"Hey wanna drive our self driving BMW?"
Reason is because I like driving and want to drive myself. I like the roar of the engine and how it's an extension of myself. If the car drives me I might as well take the bus. For people that don't care for driving, it's a great option for them.
I totally agree. I enjoy driving. it is a loss for people who like to drive.
Hey. Make one I can afford and I'm right with ya. I have lived my whole life dreaming of having a chauffeur that could drive for me, and I've fallen asleep at the wheel resulting in roll over. *I'm all for this tech!*
freedom before safety
driving should always be an option even if it means we kill each other in accidents
GUUUUUURL I want a self-driving Tesla.
A big thing is that we actually match autonomous cars against a whole generation which has experience since they made their licenses where there wasn't even an alternative like autonomous car. A beginner, however, would probably think completely different when it comes to safety in driving. I remember very well what a hazzle it was to figure everything out and to get used to it. It is kind of a lengthy process until someone can drive somewhat safely. You also see it in the figures: Most accidents are caused by inexperienced drivers (besides the overconfident or drunken ones).
Great graphics!
you can't rly compare the straight pat of autonomous flight with a self driving car
A little.
In the moral dilemma section at 1:30 it only gave 2 choices swerve left into a car or right into a motorbike... There is actually a third option which to not swerve at all, which is actually the choice the car would actually make.
Now who wants to buy a car that controls your very life! It only costs 5x more to maintain and inspect!
The answer is there are a large number of technical issues to work out including: how to make them work with drivered cars, currently prohibitive costs, and biggest of all- how to keep them from being hacked