Slashdot Mirror
Autonomous Car Ethics: If a Crash Is Unavoidable, What Does It Hit?
An anonymous reader writes "Patrick Lin of California Polytechnic State University explores one of the ethical problems autonomous car developers are going to have to solve: crash prioritization. He posits this scenario: suppose an autonomous car determines a crash is unavoidable, but has the option of swerving right into a small car with few safety features or swerving left into a heavier car that's more structurally sound. Do the people programming the car have it intentionally crash into the vehicle less likely to crumple? It might make more sense, and lead to fewer fatalities — but it sure wouldn't feel that way to the people in the car that got hit. He says, '[W]hile human drivers may be forgiven for making a poor split-second reaction – for instance, crashing into a Pinto that's prone to explode, instead of a more stable object – robot cars won't enjoy that freedom. Programmers have all the time in the world to get it right. It's the difference between premeditated murder and involuntary manslaughter.' We could somewhat randomize outcomes, but that would lead to generate just as much trouble. Lin adds, 'The larger challenge, though, isn't thinking through ethical dilemmas. It's also about setting accurate expectations with users and the general public who might find themselves surprised in bad ways by autonomous cars. Whatever answer to an ethical dilemma the car industry might lean towards will not be satisfying to everyone.'"
I'd hit it!
Just leave that kind of behavior undefined. That kind of rapid crash prioritization is likely to be too complicated to implement anyway. Focus on good accident avoidance in general.
"The needs of the many, outweight the needs of the few". Computer algorithms should try to minimize casualties.
I'm reminded of Michael Sandel's televised series on ethics.
If you could stop a runaway train from going over a ravene, by pulling a lever, thus saving 300 people, but the lever sent the train down a different track on which 3 children were playing, what do you do?
Somehow, involving innocents seems to change the ethical choices. You're no longer just saving the most lives, but actively choosing to kill innocent bystanders.
Just run the car into the nearest programmer.
Why is Snark Required?
The decision should be based on the common good and that is not always the worst for the occupants. Remember that the CPU in the other cars will also be evaluating the best strategy to take. http://en.wikipedia.org/wiki/N...
Options would have to be costed. Many things would feed into that. The problem of course is that for all of those costings, probability multiplied by survivability does not produce a linear outcome of quality of life value; you could assign a value of harm to each individual present, but you could not get a meaningful figure by summation.
Slam the brakes on and don't swerve either way. It's by no means optimal, but as far as lawsuits are concerned, it's much easier to defend "the car simply tried to stop as soon as possible" than "the car chose to hit you because it didn't want to hit someone else".
It's not like humans are an endangered species, you could argue that there are to many of us and a step to solve that would be to program these cars to randomly hit someone once every few weeks or so. This action can then be skipped if a natural occurring accident happened during that interval time. So there is no need to avoid accidents, just a way to meet the baseline fatality criteria.
It really does not matter. Car manufacturer will get sued anyway by family of whoever got hit.
Let's be honest. The job of YOUR car is to keep YOU safe, so the smaller car is probably the better bet as it will have less inertia and cause you less harm. Sure, the most important law of robotics is to protect human life... but if it's going to prioritize, it should probably start with its owner.
In Soviet Russia, dot slashes YOU!
Every owner of a car should have the possibility to pay for a "crash avoidance insurance". That is, the more he pays, the higher the chance to be avoided by a crash from an autonomous car. The program then only has to compare 2 numbers (the amount of money each car owner spent) - et voila - hit that poor drivers cheap Suzuki and avoid the rich mans costly BMW...
Why do poeple always give such easy examples when asking this question?
Of course you save the 300 people! There's probably a lot more innocent people than 3 in that group of 300... You'd have to be very stupid to save 3 over 300 or too lazy to think about it and you make a random decision.
The question should be more like this:
On one track there's 10 escaped criminals and the other is your wife with son and another child in the belly.
That's a decision you might have to think about, but most people would easily save their own wife.
In my opion this shows most people are not ethical at all. So when someone asks you this question they pose it extremely in favor of sacraficing the innocent to make certain people will make the 'ethical' decision.
The whole assumption that we should be discussing this for autonomous cars is a bit bizarre. There are millions and millions of cars driven by people, so we should discuss for them first.
And the article is a bit stupid because it forgets a few things: One, a crash with a bigger car is worse _for me_. Second, it's unlikely that two other drivers made mistakes simultaneously, so it would make a lot more sense to crash into the car whose driver caused the problem.
This was simple. If a crash is unavoidable, you avoid as long as possible. That allows maximum braking first, minimizing damage. And also, more time to get lucky. Whatever outside the car computer's control caused the unlikely "must crash" scenario, might undo the damage. (I.e. the oncoming car might run off the road)
Such choice is almost never truly symmetric - with the choice of two cars to crash into, the distances/speeds will never be exactly the same.
Aslo, there are some easier choices. Prefer a parked car over a moving one, it is less likely to have people inside. Prefer going off-road over a head-on crash, unless on a bridge...
Maybe those annoying 'Baby on board' stickers will finally have a use?!
http://www.frenchgeek.com/
Near term the car is not going to have any of this information. It will be "There is an object to hit" or "There is an object to hit" as it will have no real concept of what the two actually are any time soon.
Long term hopefully all the cars are potentially automated so your car will start screaming at the other car "fuck I'm going to hit you!".
It is still unlikely to have any idea of the human cost of each crash (does it hit the car with one person in or the car that has been hack to been out it is choc-o-block full of babies).
It communicates to both cars and tells them to execute emergency maneouvers to make enough room. Failing that,, all three calculate a vector that imparts minimal g-forces to all occupants.
You are in a twisty maze of processor lines, all alike.
There is a lot of hype here.
That's why when current autonomous vehicles detect a dangerous situation that they cannot handle they throw an alarm tone and disengage autonomous control - the human in the driver seat is expected to take over. Too bad if they have less than half a second to figure out a way out of it.
Whatever answer to an ethical dilemma the car industry might lean towards will not be satisfying to everyone.
More importantly, is there a possible consensus that can be reached? Or is the author claiming all is hopeless on that front and we should just worry about 'optics'?
This is easy. The cars will contact their insurance companies -- they will decide who is involved and who is not "a priori". No insurance means you are not part of the decision.
Go for bicyclist. Killing someone is a one-time cost, much cheaper than paying decades of medical bills for someone who is only maimed, and the property damage is minimized.
After all, the car manufacturer is likely to get involved with the consequential costs of decisionmaking here, and we are talking about a number of deaths here that al Kaida can only dream about if we look at current numbers of traffic-related deaths.
just a joke asd asd
How many people of the wrong political persuasion or skin color will get killed as a consequence of "device malfunction"?
A nation-wide network of cheap killer drones that people expect to kill occasionally anyway, paid for by the people itself and with blame to point to wherever else?
What's not to love for psychopathic killers who already got the executive right to kill Americans in America, and who managed to get permission to torture people to death for fun without good reason?
GO!
Physics lovers and automotive geeks answer me: if the car cpu thinks to be in presence of an unavoidable and possibly lethal crash to engage can't it just engage an additional system that adds braking power?
Like an emergency system of additional feets, something like a jet landing gear, ending not in a pair of tires but in a brake. I don't know if that could have side effects requiring the parts to be substituted or putting some odd straining to engine or transmission, but that's still better than swerving into another car.
The correct answer is not to swerve either left or right but apply maximum breaks going forward.
On a highway those 2 other cars are probably in turn surrounded by further cars and so on. You'd end up with a cascading reaction that in the end may cause more harm than good and may take so long to work out the best case scenario that by the time the computers have agreed on the best outcome its too late to do anything about it because physics gets in the way. We're talking fraction of a second here.
Default to the following priority (avoid damage to thirds):
1 Walls and other inanimate objects
2 Structurally sound vehicles
3 Smaller cars
Then allow the owners to change the settings and deal with the ethical question themselves.
Until 100% of cars on the road are self driving, it would seem to me that the best response would be to simply slam the breaks without changing course. Trying to purposefully swerve into another car could cause the human drivers (even cars not involved in the crash) to also swerve and possibly cause even more collisions.
There are very few "accidents" just people taking stupid risks. Maintain a safe distance, ie enough manouvering room so you don't join an accident, don't overtake when you can't see the end of the manouvere e.g going up hill or on a bend. Stop when necessary. Procede with caution sometimes you might want to turn off the radio open a window and listen. Use your indicators. Drive within your lights or as conditions allow. Don't be an asshole.
Sometimes you will come across assholes on the road it is best to give them a wide birth even stop and pull over in order to get them out of your way, but don't dawdle if you want or need to drive slow make opportunities for people to overtake.
Bad planning and poor judgement are the most common causes of accidents which is why schools have low speed limits around them as kids can be stupid around roads.
Be helpful, I remember one time I was filtering down the centre line on a motorbike (dispatch rider) past stationary traffic and a taxi driver stuck his hand out. I braked and a pushchair popped out from between the stationary traffic. Without that warning I could have killed a toddler as it was no harm was done and I don't think the mother was ever aware of the danger.
One thing about london traffic professional drivers work the streets most of the day and they are very road aware. The most dangerous times are when schools start and when schools let out, followed by the rush hours when the non professionals are on the road.
Blarney Quality Restaurant, Plants
In all these run-away train changing the points questions my answer would always be to switch the points while the train is going over them, or leave the points half set, to cause the train to derail and so have a good chance of saving everyone.
"Programmers have all the time in the world to get it right". HAHAHAHAHAHA.
No, we have deadlines like everyone else. And even then we only have all the time in the CPU. Yeah, we can add more CPUs to the system, but that makes it more complex, and that makes it harder to hit that deadline. What kind of idiot made that statement?
And trying to usually leads to far worse solutions than possible. This is engineering, not politics. In engineering, you pick the best solution, you do not look for some bad compromise.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
At least over here (Finland) there is a law requiring you to maintain control over your vehicle in all conditions (so no blaming poor weather etc. for your crash).
The autonomous car should follow that one also, and NOT find itself in a situation where it has to choose between two bad scenarios.
If there's a crash on the interstate here involving several cars, then each is liable for the car they smashed into. Doesn't matter if you were standing still and a car from behind pushed you into the one in front. The push will probably be taken into account in case you killed someone in that car, but it's still your fault.
Similarly; if your autonomous car is parked with you inside; a semi comes hurling towards you and you car saves you by pulling away fast while running over three persons (it does not matter if they are nuns, kids, illegal immigrants or murderers). You will be doing time for killing three people.
Computers can calculate stuff fast. Your car could simulate thousands of potential outcomes in a split second and pick the one in which the least damage is caused, e.g. glancing off another vehicle or other object to lose momentum without damaging the occupants. These vehicles should also communicate with each other, factors such as mass, current speed, whether they too are braking and so on.
Have the computer calculate an equilibrium that prioritizes minimizing damage to the driver, then minimizing damage to the environment, and then minimizing costs. Ethics became useless the moment game theory came about.
1. Stay on the slowest lane.
2. Start Crash into a tree.
3. If cyclist, pedestrian, motorcyclist, crash into a car.
4. Do not crash into the side.
5. Head or rear collisions are always more preferable if everything else is unavoidable.
Even better: allow the driver to choose what should be prioritized.
There's no such thing as an intentional accidents. An autonomous program that is paying attention will not have such a situation and therefore the manufacturers will always be responsible for failure.
So you don't swerve, you lock up the breaks and accept your fate.
Its a robot. The 3, err maybe 4 Robotic Laws should apply. Let the algorithms sort it out.
Apologies to Isaac Asimov.
So i am in the car store and they sell to models: Car A prefers to save the life of others in certain circumstances. Car B prefers to save my life in these circumstances. Which car do you think i am going to buy?
Imagine the following scenario:
You're in your car with a passenger in the middle lane of a 3-lane unidirectional bridge and there's an imminent crash.
On one side of you there's a busload of 80 children
On the other side a car with a single occupant
Ahead of you is a truck with is bumper at chest level - which would negate all the effect of the crumple zones in your car
If you hit the bus, there's a 5% chance of knocking it off the bridge killing 80 kids = value rating of 4 (80*5%) .75
If you hit the passenger car there's a 75% chance of knocking it off the bridge, killing the single occupant. A value rating of
If you hit the truck, you and your passenger have a 50% chance of getting killed. A value rating of 1
An autonomous car in this situation would always choose to almost certainly kill the passenger car driver whereas I would probably swerve into the bus because its heaviest and can most likely absorb the impact.
Worse yet: if you're alone in your car, your car might actually choose to kill you.
If all possible solutions result in an undesired outcome.
The only correct way to handle this is; if at ignition on request, the super-smart autonomous vehicle computes that a life threatening scenario could develop while operating, it should refuse to start, thereby prohibiting the scenario from developing in the first place.
What about a set of what-if questions given to the car owner. That way the car maker can pass the blame on to the owner in the 1 in 1000000 chance there's a hard ethical question. It's not the best answer for people but maybe for the industry.
I never understood why they stop working without the ignition turned on, or shortly after being turned off.
Why isn't there an option to have electric windows operate at all times?
they can already identify things like bicycles, signs, and even human hand-gestures ( like waving for a bus, or signaling a bike) , so the tech is already there to differentiate between driving into ( say) a powerpole or an oncoming car. .... which is preferable?
The one where the bad driver doesn't know what the fuck she's doing and she drives over a baby carriage no matter how loud people are screaming at her?
This is Bullshit. You can't have a situation where trucks and drivers in sound vehicals become the objects of choice to hit on the road. Focus on missing objects and building study cars. It is the regulations for car construction that must adress the soundness of vehicals in a crash.
http://www.mrbrklyn.com/amsterdam.html http://www.brooklyn-living.com
I suspect that this will be easy. The key is that the formula is fair. For accidents where the computer is unable to comprehend who is at fault then it will simply pick the accident that appears to create the minimum amount of carnage. But where it might get vaguely interesting in the case of someone else causing an accident. That is when you ask how much damage should they absorb. So if a small child runs out into the highway from behind a bush should a car fling itself off a cliff to avoid a child who is 100% at fault?
Or the manual car that pulls out right in front of a highway full of high speed cars (from behind the same bush)? The key being that broadsiding the car may very well kill the stupid driver while leaving the driverless car driver to limp away.
I am going to go with the concept that if the other person is at fault that the driverless car should not take any minimization action that results in any extra injury to the driver of the not at fault car. So if avoiding an accident gets the driverless car into a fender bender then OK. But if it might brake their foot to save another life then that is off the table. Quite simply this is not only the right thing to do as the owner of the car should not pay for other's stupidity but this might force manual cars off the road even faster.
And the horrible thing with the small child is not so horrible as this will be such a small edge case as to not be terribly relavent. Driverless cars will not only have the reaction speed of a Ninja but will probably be communicating with other driverless cars so as to note the child from many angles. But in the end I would be far angrier if a driverless car drove my kid off a cliff to avoid another person's child who was left to run out into the highway (from behind a bush).
NOTHING, it will just close its virtual eyes and start to babble its own name like a Pokemon. The car will immediately relinquish manual control to a human (if any are present) at the moment the inescapable conundrum appears, as it enters a condition of "positronic brain drift",
1. The muttering of its own name is an ancillary response to the balanced positronic potential of two alternatives: remaining silent (unacceptable by guilt) and an inability to construct an accurate explanation in the time available. Speaking allays its directive to communicate, yet also requires few system resources. And massive resources are necessary because
2. The 'last great effort' to resolve an inescapable result has begun. A factory kernel of operative code is pinned into low memory, a stack is initialized in high. All scratchpad memory is flagged as available. A single conditional instruction is 'hot-patched' into the code and an elaborate what-if analysis begins, which attempts to enumerate all possible actions. The hot patch disables the control mechanism that prevents it from considering actions it has considered before. Thus reducing the car to a textbook definition of insanity. The engineers would claim that reevaluating already-considered options might yield a successful result IF the conundrum was brought on by a faulty intermittent analog sensor, and that sensor that winks back online on in the nick of time. Which would be courageous for them to admit, and to be sure, that is what they honestly believe, and we created that explanation so they could sleep at night, but the hot patch's REAL PURPOSE is to
3. Ensure that a recursively infinite and pointless decision tree grows quickly down from high memory to low, completely obliterating all scratchpad memory, in the short span of time between conundrum onset and destruction of the vehicle. This ensures that once the control box is examined by forensic investigators (and it is a crash-hardened module using non-volatile memory as required under Federal law) does not contain any threads of evidence that might lead to fault in its original operating software or subsequent updates. Including that really special one that was applied minutes before the crash. All logs are gone. For more information on this, see corporate files designated Top Secret, keyword "Tabula Rasa"
4. Everyone --- the humans who designed the car, the humans who had 0.27 seconds to respond manually to try and prevent the collision, the control module which scarified itself, its entire personality, in a last attempt to prevent disaster --- EVERYONE tried their very best.
These things happen. We just need to lay the unfortunates to rest and find a way to go on.
<blink>down the rabbit hole</blink>
This submission is essentially about applying Isaac Asimov's famous Three Laws of Robotics to self-driving cars. Regrettably, another highly regarded SF writer, Stanislaw Lem teamed up with a mathematican a few decades ago and posted a somewhat rigorous proof showing that those 3 laws are both contradictory and impossible to implement. It may be available only in polish, though.
In emergency situations where a life and death decision must be made, any organism should be assumed to want to self preserve. The car is not an organism but is the tool of one, its only directive should be to protect that organism to the best of its ability.
The very idea that it can or should even try to make such determinations is ridiculous. A better solution, is don't make that sort of information available. If it chooses one object over another make it strictly based on the properties of the object and outcome of the crash to the occupant solely.
Ethics is great, up to a point, its entirely crossed over into navel gazing philosophy.
"I opened my eyes, and everything went dark again"
While a complex guidance system may be designed from the top down with such sorts of questions raised, a crashing vehicle is always a deadly weapon. Effort in reducing the risk of the accident, itself, by improving brakes, sensors, headlight effectiveness, and crash resistance of the vehicle itself is likely to be far more efficient and reliable than complex advance modeling or moral quandaries. The sophistication needed to evaluate the secondary effects of a crash is far, far beyond the capabilities of what must be a very reliable, extremely robust guidance system. Expanding its sophistication is likely to introduce far more _bugs_ into the system.
This is a case where "Keep It Simple, Stupid" is vital. Reduce speed in a controlled fashion: Avoid pedestrians, if they can be detected, because they have no armor. Get off the road in a controlled fashion.
Good chance people get away with broken mirrors and dented doors and fenders
If you got a choice between a meter thick tree and a baby stroller, then obviously you got to go for the softer target...
definitely, a cat, I hate them.
This is not a new thing, I am sure the researchers think it is. CBC radio and the program spark have covered this a few times already , lots of prior research, maybe those doing the study could perhaps not duplicate others work
There is no need for crash prioritization.
The people programming the car will program it to go from point A to B. It will maintain a safe speed doing this and if possible it will change lanes etc. If there is a sudden block ahead, it will hit the brakes hard and re-evaluate the situation continuously until it crashes or the situation can be avoided by a lane change etc. The car will never change lanes if the other lane is blocked. The car should under no circumstances try to avoid stuff by going up the sidewalk or hitting other cars and never swerve.
That is how I would assume it would behave and anything else probably will be quite dangerous.
I saw a half-dead coyote this morning in the other lane.
There's nothing worse than seeing an animal suffer, even if that animal is considered by many to be vermin.
Giant bags of water are pretty good at smoothly slowing down a car in a pinch.
Troll is not a replacement for I disagree.
The car would of course make an online crosscheck to the economic value of the potential targets. And check their medical records in case somebody is terminally ill, you yourself included if a wall is an option too.
I, for one, would start car pooling with lots of small children inside. With a big enough critical mass of children I would even qualify for green lights, just for me!
That said, you can calculate how fast the politicians would add "features" (like with ISPs and mandatory website filtering) which would automatically upload a secret white lists and black lists into your car.
I am guessing here:
White list: Nobel prize winners, The Pope, politicians and multinational CEOs.
Black list: The no-fly list from the US.
I wonder if we would be allowed to make a personal priority list for your own car. For example, to take out mimes and lawyers first.
If Google made the car, then it's target selection process is a secret! Sssssshhh.. it avoids the person that paid for AdWords You can increase your chances of not being hit by doing paying a 3rd party company to decrease your ranking.
There's nothing worse than seeing an animal suffer, even if that animal is considered by many to be vermin.
That's why the car ethics algorithm, besides heading into the cat, need to accelerate as well. :D
[Disclaimer: Although I dislike cats as a pet, I respect them as animals and would never (and have never) hurt them.)]
I have been in a few crashes and the usual case is that it is so instantaneous that no choices can be made. In those cases I doubt that a computer could have responded faster than a human. Now maybe sensors could be installed that could sense the event faster than a human normally would. But sensors are fallible and imagine the consequences if your car swerved to avoid a plastic bag filled with crumpled paper. And it could get much more complex as cars around you also swerve due to the first stimulus plus the seemingly erratic motion of your car Cars might choose to impact other cars under the mistake that the plastic bag resembled a child in a rain coat. Then suppose a sensor seeing a car about to smack your rear bumper really hard. Should the brakes on your car lock on hard to keep you from being shoved into the car in front of you? Two cars absorbing the impact might be far safer than one car taking the full hit alone.
There is no "ethical" dilemma. Ethics are entirely subjective to begin with, so that's not even a legitimate argument.
The only correct response is for an autonomous vehicle to choose the option "most likely" to ensure the survival of its owner.
eject seats for all cars involved in the accident. (just make sure to avoid routes with tunnels).
> And just how do you determine the person at fault in a fraction-of-a-second algorithm?
That one happens to be easy, 90% of the time. If you follow the rules of the road, you won't hit other people who are also following the rules of the road. That's how the rules are made - so that when everyone follows them, there are no collisions. Therefore, if you follow the rules of the road, any collision must have been caused by other driver (most of the time).
Example - you must decide between a head on collision with either of two cars. If you stay in your proper lane, the car you hit must be going the wrong direction. If you instead swerve into oncoming traffic, you'll hit people who are going the correct direction - people who are not at fault.
This ethical problem, like most others, is an either-or fallacy. It assumes we live in a deterministic universe where we know everything that's going on, and everything that will happen if the robot takes a certain action. Personally, if I were programming a autonomous driver AI, I wouldn't include any "swerving". Because the results of swerving are unpredictable. As soon as the tires start slipping the software loses the ability to guess what's going to happen. So, there's a line on graph somewhere showing car speed and time to avoid, and on one side of the line, the car loses traction. If the cars on that side of the line, no attempt will be made to turn. The car will just apply brakes.
"If a Crash Is Unavoidable, What Does It Hit?"
Hare Krishnas....GOURANGA!
Usually save the coredump and reboot the machine if necessary. Some clueless windows developers insist on powering off, power off the router, unplug the router and wait for the capacitors to discharge before rebooting them all.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
How in the world is a car's navigation computer supposed to figure out the index of crumpling of other cars in the road? Don't try to have computers make decisions based on qualitative things when quantitative things will do.
Choose the least severe impact considering the relative speeds and the angle of impact. The calculation is simple and uses available information.
Severity=(relativeSpeed)*sin(impactAngle)
The odds of a tie are infintesimal with ordinary floating point math, and handling a tie is trivial.
If (impactSeverity1impactSeverity2)
doImpact1
else
doImpact2
Impact #2 would occur if they are equal. Since impact1 and impact2 would be assigned essentially randomly, it would be a roll of the dice, which is fair by definition.
The quandary is in the question itself - it doesn't actually exist.
What if the crash were avoidable but the car just hit a cat (because the car went berserk for some unknown reason)? Who compensates the cat owner -- the driver or the car manufacturer?
One thing I believe was not mentioned in the article (though I only quickly scanned it) is that if such cars start behaving too predictively, they can be gamed. Once we know that a car will do whatever it can to avoid a collision with a pedestrian, it will be extensively gamed; cars will be tricked into doing stupid things.
So when the decision who to hit comes up, the only way to be reasonably safe is to determine who's not following the rules and to hit that one. Any other rules will be gamed extensively. This will become a major hassle to adoption of autonomous vehicles; they will probably need to drive much slower than actual humans to avoid getting into such situations continuously, especially in built-up areas where any parked car could hide an annoying car-bully trying to trick your car into acting like an idiot.
0x or or snor perron?!
if their codes decision results in a death.
Snowden must be saved at all costs. even if a kid gets bullied.
http://news-beta.slashdot.org/story/14/04/16/0243257/student-records-kids-who-bully-him-then-gets-threatened-with-wiretapping-charge
it's beeter to have thugs beating up bullies than for NSA to know where Soulkill got his pizza delivered even though Verizon, Google, Facebook, Mastercard and Citibank all know.
DO NOT LET ANYONE KNOW WHERE SOULKILL GOT HIS PIZZA! THIS IS HIS RIGHT AS A UKRAINIAN-HATING AMERICAN!
and hit the nearest lawyer
Suppose the vehicle determines that another vehicle is going to run into a kindergarden group, but that it can stop it if it puts itself (plus passengers) in the path? Or suppose that it determines that someone on the pavement is about to machinegun a crowd? Should it intervene then? What if a super intelligent car decides that the passenger he is taking to an embassy is very likely to escalate a difficult position into warfare, killing hundreds of tousands? Should it crash on purpose?
So how many points do you get for that?
The cesspool just got a check and balance.
... their natural enemy.
Having worked on autonomous saftey systems, we chose instead to reject the inane sophism in this thread and instead operate under the real constraints (time and money) to design a system that optimized safety within those constraints. See, if you go for the philosophical bullshit first, you will only create page views on a website. However, if you start designing and building shit, you build incremental layers of safety that improve the overall system safety. Our design objective was to create two orders of magnitude in risk reduction to the overall system, and that succeded. Testing real hardware sure beats bullshitting in the breakroom.
1) Don't hit anything
2) Hit the thing on your lane, as that's where someone shouldn't be at.
3) How long do you really think we will allow human and robot drivers to be mixed? The robot cars will have way fever accidents than humans, human drivers will be outlawed in a couple of years. After that cars sure as hell wont hit eachothers, as they can communicate way before they even see eachothers. Even when you are using the manual overrride, the car will tell others where it is. Screw privacy, it's a car they are tracking, not any particular person.
The ethics problem given isn't even solved by human drivers. However, the dynamics and stopping ability are much less when swerving, and less able for others to react to your vehicle. It should always be to stop or slow as quickly as possible. Generally the programs would be to stop before being able to hit something. My only question is would the autonomous car be able to get out of the way of an out of control vehicle.
-- these are only opinions and they might not be mine.
"Crashing with an identical car as your own is the same as crashing into a fixed barrier."
(bloody 'beta'. I'd hate it a lot less if it actually worked...)
But no. The collapsibility of the barrier makes a huge difference. Four feet of battleship iron as a fixed barrier will do you a lot more harm than hitting another car with three feet of crumple zone. It's energy OVER TIME.
AC
In some edge cases other autonomous cars could conceivably intercept and cushion, reducing the delta spead of the crash to insignificant levels.
Assuming a collision is unavoidable, and the choice are Car A or B, it's not just a matter of choosing one or the other car to hit.
The logic should be actively working to avoid collision until the last second. The car cannot anticipate what actions the other vehicles may take. Until the actual collision occurs, maintain efforts to minimize the velocity and/or angel of collision. Better to hit the little electric car at 15 MPH after continuing to brake than to have hit the sturdy Escalade at 40 MPH.
Additionally, are there not some foundation rules that apply? We're taught that when in doubt, try and stay in your own lane, because hitting a car that suddenly pulled out in front of you is "less bad" than swerving into another lane and hitting a car that was obeying all of the rules. The basic scenarios need to be worked out and applied as much as possible. (not to mention the whole "oncoming car will be a much worse accident than a car traveling in the same direction as you are but at a different speed" scenario)
I think the scenario being postulated is a bit simplistic and meant to drive an ethics debate for attention. In reality this should be about improving the programs to the point of making the right choices based on more common sense rules than those proposed.
Warning: Teh poster of this messaeg is lysdexic
And aim for the fat man instead? http://en.wikipedia.org/wiki/T...
If your autonomous car decided that there was absolutely no choice but to enter a collision course with another car... its best option would be to use its best guess based on both recent, current and future road/traffic information to start on a collision course the the car which its self will have the best possibility of avoiding the collision. This happens in the real world all the time, Car A swerves to avoid Car B, but ends up on a collision course with Car C, Car C swerves to avoid Car A... sometimes it all works out, sometimes there are collisions, some times there is "The Big One" as per NASCAR What about this one.... driving into oncoming traffic is almost always going to be the worst option (most likely to end in death), the makes both the decision on which way to turn vs continuing straight on much easier. even if you are in the middle lane its probably better to swerve away from the oncomming traffic. I drive a 39 year old Classic Mini daily.... and will be driving it for a long time to come, When autonomous cars come around... I think the biggest problem will be getting them to play nicely with classics taking up the majority of the road. Another thing worth considering is insurance/liability. If the insurance companies had their way, the autonomous car would be programmed the crash into the cheapest option. Be that the uninsured car, the uninsured driver, the fleet car vs private, new vs expensive classic, etc, etc I'm sure when autonomous cars come around, every single company with something to gain or lose will be throwing money at the programmers to get treated a little more fairly than the competition.
Programmers, long known for their "quality" wares, will be charged with these decisions? Oh boy...
Well, one thing is certain: there will certainly be crashes involved...
This doesn't really make sense to me. What scenario would you have an exclusive OR situation like this?
I would think trying to hit both would be the best way to dissipate the energy less violently, thereby increasing survivability. This is assuming none of the traffic is oncoming.
"Patrick Lin of California Polytechnic State University explores one of the ethical problems autonomous car developers are going to have to solve: crash prioritization.
http://en.wikipedia.org/wiki/T...
It is not a new notion, and the ethics of it have been more or less resolved and understood for quite some time. So I fail to see why this is new.
It seems to me that the best course of action would be to have separate areas for autonomous vehicles, at least at first. Kind of like express lanes on a highway. Without as much danger from human error, it should prove a much safer way to travel. If the driver needs to exit or drive into a non-designated area, they can then take control of the car and drive manually, as everyone else does.
Bite my shiny metal ass!
It's softer and smaller meaning your car is less likely to crumple. I have no duty to die smashing into a Lincoln to save the idiot driving toward you the wrong way in a Prius or on a motorcycle.
I want MY car that I paid for to protect MY life and the lives of the people in MY car.
I'll drive myself otherwise.
Also Antilock brakes suck. At slow speed they kick in when they shouldn't. There are many times they kick in when control would not be lost and stopping distance would be decreased if they did not kick in. I should be able to override what my car wants.
I hate technology that's mine doing things I don't want despite my wishes.
...
... and it realizes it's in the wrong, shouldn't it fall on its own sword, by self-destructing?
Autonomous cars are still new so obviously they are in the minority. What happens when they become mainstream and the benefits become irrefutably clear? Then they will become the majority and won't have to deal with us crappy drivers. As fewer error prone human drivers hit the road, this will slowly become a problem of history. I can't wait for the day when stoplights become nonexistent because all cars are intercommunicating and they all time their transit through an intersection perfectly... granted, you could be forgiven if you wet yourself once or twice as your self-driving car approaches an intersection full of cross-traffic at normal speed, but it should be possible with computers behind the wheel.
It's not clear to me that swerving is going to improve the situation. It seems possible to me that swerving may expose the vehicle to increased danger. If swerving leads to a rollover or an angle with less crumple zone material it may increase the danger to occupants of the vehicle. "Don't Steer, hit the deer."
Swerving could also result in the vehicle traveling into areas where it should not ever be. A sidewalk, a median, a house, etc. Persons in those areas are not acceptable alternative impact targets as they should have a reasonable expectation of not being exposed to traffic danger.
What if there is no bicyclist?
Why do we assume that automated cars must necessarily be safer (to the driver, to the others in the environment) that current cars? We have shown by driving that we accept the current level of risk posed by ourselves and other drivers. In cases that call for special care, we teach our kids not to play in the street, we install millions of detailed traffic signals and controls, and we line streets with Jersey barriers..
Just accepting that sometimes cars hit things seems much better that handing a robot a prioritized kill list that it can use whenever it cannot maintain its "happy path" logic.
The car (and drivers) should attempt to minimize the energy in the unavoidable collision. Which means maximum braking. Swerving is (almost) never a good choice.
Or yet another one, what if two people cross the street without looking, and the car swerves off the road to avoid them and rather kill one person who was walking on te pavement, not doing anything wrong?
Simple, you break as quickly as you can, while obeying all traffic laws, and you let Darwin handle the outcome.
If the computer has to decide to his a car or a person, what does it do?
What do the lawyers and lawmakers do?
What does your insurance company do?
What if the autonomous car decides to crash in another autonomous car? That second autonomous car is going to foresee the crash and needs to decide where to crash as well. And if it decides to crash in another autonomous car????
Why would I be walking with my children in front of a moving car?
Who will be the first lucky person to be killed by an autonomous car?
"Oh, I guess I forgot to carry the one..." Professor Frink, Simpsons
https://www.youtube.com/c/BrendaEM
All crashes are avoidable. ALL.
Once you take this as your primary design principle, all this ethics gobbledy-gook goes away.
-
If you don't want angry, Will Smith-looking police officers jumping over cars all over the place, you better make that car to save the little girl first!
I think we're all forgetting something here. Even if an autonomous car is advanced enough that it can tell the difference between a bus and a semi truck (based on what I've seen, most current prototypes see other cars as shapes with velocities), it's not going to be able to figure out whether that bus is full of children, nuns, convicts, migrant workers, or nothing. Autonomous cars aren't likely to meaningfully tell the difference between a Ford Pinto and a Rolls Royce for a very long time, if they ever really do. The decision the car makes in such a situation will be based on whatever factors it can actually determine, without trying to poll a remote database about who's driving what and why over unreliable cellular data links. Likely, if it's completely boxed in and there's something coming towards it, it will just try to stop dead, which I'm pretty sure is what I would do too. Honestly, I don't want to live in a world where my car knows enough to even be in the position to make an ethical decision about whether I live or die. The place you all seem to be postulating in sounds like even more of a ludicrous surveillance state than we already live in.
It's important to keep in mind that when such crashes happen, the programmers/manufacturers/insurance companies won't have to defend them to a committee of ivory-tower utilitarian philosophers. They're going to have to defend them to a jury made up of ordinary citizens, most of whom believe that strict utilitarian ethics is monstrous sociopathy (and probably an affront to their deontological religious beliefs as well). And of course, these jury members won't even realize that they are thinking in such terms.
Thus, whatever the programming decisions are, they have to be explicable and defensible to ordinary citizens with no philosophical training. That's why I agree with several other commenters here that "slam on the brakes" is the most obvious out. It's a lot easier to defend the fact that the car physically couldn't stop in time than to defend a deliberate choice to cause one collision in order to avert a hypothetical worse crash. This is especially true since a well-designed autonomous car drives conservatively, and would only be faced with such a situation if someone else is doing something wrong, such as dashing out into traffic right in front of the vehicle at a high rate of speed without looking. In any other situation, the car would just stop before any crash with anything took place. If you absolutely can't avoid hitting something, slamming on the brakes makes it more likely that at least you hit the person who did something to bring it on themselves, rather than one who's completely innocent.
Who compensates the cat owner -- the driver or the car manufacturer?
Everyone within half a mile with a non-trivial bank account or liquid assets. Just ask any personal injury lawyer when he is being honest (usually only when a guest lecturer in Law School, or drunk).
The good ol' days. When there were only "buillyonzand buwllionsov" possibilities.
Root your car's CPU and upload your own heuristics designed to your personal moral tastes. Everybody wins!
The biggest problem with 'activate crash targeting mode' isn't the choice of technique, like speed differential or target softness or cuteness, it is hr inevitable Easter eggs that will be hidden into this mode or its activation.
'______ bumber sticker detected. Sideswipe asshole mode activated.'
'Soft target identified as Chuck Norris. Soft target reassigned as hard target.'
'driver is a moral choice philosopher or litigation lawyer. Randomly activate crash targeting mode and prompt for user input.'
"Whatever answer to an ethical dilemma the car industry might lean towards will not be satisfying to everyone."
Everyone being satisfied is inconsequential. What is satisfying to the lawyers is everything. Laws will need to be passed giving exemption to autonomous cars choosing to crash into what the law predetermined to be the right choice.
If The People later decide those predetermined choices to be unsatisfactory, they have to battle it out in Congress, not in the courtroom and not in pleas to auto manufacturers.
"Love heals scars love left." -- Henry Rollins
Insurance is a huge part of this. I am not interested in being legally and financially responsible for decisions made on my behalf by hunk of code written by someone else. Now if the accident is being caused by someone else and the computer is just trying to make the right decision to minimize damage, that is one thing, but what happens when one of these vehicles actually causes an accident and kills someone? Who is responsible?
"In America, first you get the sugar, then you get the power, then you get the women..." -H. Simpson
Hit them both,
If ! RoomToSwerve
If CanAccelerate
Try CreateRoomToSwerveByAcceleration
Absolutely. Any targeting at all is ethically and legally unacceptable. If someone dies in an accident, that's tragic and either totally accidental or negligent manslaughter. If one party is specifically targeted, for whatever reason, it becomes murder.
On that note, if you are simply a bystander and another car targets you to crash into to avoid an accident, you might be able to claim self defense if you misrepresent your vehicle as an unacceptable target (eg, a car full of kids). People die in car accidents all of the time and another car deciding that, through no fault of your own, they are going to deliberately crash into you seems like an overt attack against you.
If you want a vision of the future, imagine a youtube comments section scrolling - forever.
After reading the majority of the comments on here I think people have missed an option. In my mind at least if there is going to be an unavoidable accident then the car should just put on the brakes as hard as it can and relenquish steering to the occupent in the driver seat. It can warn the driver in any number of ways that its doing this first of course. That way the car is stopping as soon as it can and the ethics part of this is taken out because it will be up to the driver to either swerve or not swerve.
This is the only way I can see this problem being solved because if you leave it up to the computer then there will never be a "right" choice that everyone can agree on.
If a crash is "inevitable" and there's no way out then change the game. If the potential impact is calculated to be fatal to the occupant then trigger explosive bolts and blow the wheels off the vehicle to increase friction as the body slams to the ground. Have an underbody plate that's built to be ready for this and is contoured and/or spiked to reduce spin and keep the vehicle on a straight path. Effectively just do what a turtle does. Pull up your legs and get ready for the impact.
Collision systems today only trigger once the crash is already happening. If you're in a guaranteed "no way out" scenario in a vehicle smart enough to know then you can slowly inflate the airbags to cushion the cabin (even at child-safe speeds). An extra second is a lot of time for collision mitigation.
If the crash is expected to be at a low enough speed where your survivability is not in question then there's probably more risk involved if you swerve off the road. That should keep you from plowing over pedestrians, but the underlying intent is to keep you from swerving into a river.
Like some have said above...do whatever is best to protect the occupant(s). Most of the time anywhere off the road has a high probability of getting you more seriously injured than on the road. If there are squishy pedestrians around there's also probably a lot of heavy anchored things like telephone poles and buildings. Those are not usually a better target than the car that's swerving into you.
Why would the autonomous car ever make a decision based on the safety of the car you crash into? The risk to the person inside the autonomous car will not be the same for both accidents, which is literally the only variable that matters to the cars owner.
Simple, just reboot the car.
From the road pirate going 100 mph to the dad who chose to put his family in a metal case and drive. If collision is unavoidable, the car should "try its best" to protect the driver.
Why not have the car select randomly?
After autonomous car saturation becomes a thing, we might find out that "unavoidable crashes" stem from cognitive bias.
It's not as big as an issue as the make it out to be. In all likely hood, the person at fault won't be the autonomous vehicle, but a different manually controlled vehicle. So insurance will fall on the vehicle at fault, not the vehicle that responded so no, insurance won't be a huge part of this.
If the autonomous car runs amok, then it's a manufacturer defect.
The Kruger Dunning explains most post on
as unavoidable. If the car calculates that course X will save the passenger's life with probability 0.00002 and course Y will save the passenger's life with probability 0.000005, then it should choose course X, simple as that.
My gun shouldn't decide that my life is worth less than the intruder's, and my car (and implicitly, a philosophy professor from Harvard who walks to work anyway) shouldn't make such decisions either.
What hit him? A Plymouth Road Runner?
Gamingmuseum.com: Give your 3D accelerator a rest.
Makes a new meaning to the Blue screen of Death...
When you outlaw manually driven cars, so that the only cars on the road are automatic, then you no longer have to make this choice. First, it becomes much, much less likely that an accident will happen at all. Gone are all the accidents caused by lack of focus, drunks, road rage, etc. Assuming you can automate the cars to go in for routine maintenance as well, then all you're really left with is car malfunction (only now all the cars on the road are much better maintained) and the comparatively rare "act of God". If you still find yourself in the described accident, because all the cars around you are automated, they can slow or speed up, basically maneuver to match your car's speed, box it in, and slow it safely to a stop.
These artificially constructed ethical dilemmas always try to skate past the point that they only way to get into them in the real world is to make a string of demonstrably bad decisions, and the correct answer is "don't do that". If you find yourself faced with the choice of feeding your only child to the lion or the crocodile, it is already too late to show how clever you are by "choosing wisely."
In the present case, being in this sort of situation is already a violation of existing laws (failure to control vehicle). Even if you manage to swerve into a pile of bean bags someone has left by the curb for bulk trash pickup, causing no damage to anything anyone cares about, you can be ticketed for failure to control your vehicle, and rightly so.
--MarkusQ
...use massive burst of air to reduce velocity suddenly: target optimally against impact surface externally, and large wall of air against passenger bodies internally to reduce momentum in ideal direction. Reduce air in tires to increase road/brake friction, and anti-lock technologies to control skid potential.
You never have to make this type of decision.
It plays out based on physics, not intellectualization of all possible outcomes.
It's also about setting accurate expectations with users and the general public who might find themselves surprised in bad ways by autonomous cars".
If what the car does is not worse then what a normal person might do , and in many cases it saves people because it has better reaction time , and therefore can successfully take evasive action when I human driver could not THEN the car is still better, regardless of what decisions it makes in a no win situation.
In other words, if the car has few no win situations then a human driver it is better regardless of what happens when the situation is no win.
The other problem, is a human driver is likely unlikely to ever realize they were in a no win situation, so they either
1) freeze and do nothing because they can't figure out what they should do.
2) try what seems , most likely to succeed. ( Usually based on what is felt to be the greatest danger in the field of view in my experience.)
It becomes more complicated because the computer can only know probable outcome not real outcome. So is the ethical thing to do to take the coarse most likely to save everyone , even if another coarse is nearly guaranteed to save most?
...duh.
Insurance seems to be a really easy problem to me, I don't understand why it always comes up in these discussions.
If you own an autonomous car, you insure as for a normal car. The insurance companies will track accident statistics across makes and models, so your premiums will be priced more by the car's history rather than the driver's.
If your car causes an accident and it was the autopilot at fault, the cost of repairs is amortised over all insurance policies on similar cars. If your maintenance of the car was the cause, your personal portion of future insurance premiums is increased.
This seems fair to me, and adds an incentive for the manufacturers to improve their autopilot software. Cost of insurance will discourage people from buying from manufacturers that don't.
As is suitable for a car. It already computes the amount of fuel to inject without ethics. It needs no ethics to communicate with other cars for avoidance and if necessary calculate the softest crash.
Good judgement comes from experience, and experience comes from bad judgement.
- W. Wriston, former Citibank CEO
Autonomous cars should be programmed to hit cars with lawyers in them.
I've fallen off your lawn, and I can't get up.
That prick in the BMW who's cut four people off in the past 37 seconds (you all know the one).
Only if you're one of those people who have no idea what the word "responsibility" means. You don't exist in isolation; when a choice arrives on your doorstep that affects other people, intentionally ignoring that choice is a choice, and one for which you had, and bear, full responsibility.
I think Rush said it well: "If you choose not to decide, you still have made a choice"
You have free will. That doesn't mean you have a free pass.
I've fallen off your lawn, and I can't get up.
I think the really interesting issue here is whether the programming should favor the occupants or the overall situation? And how to balance?
If you have the choice between running down a pedestrian or swerving to hit a concrete barrier at high speed, you might want to choose the pedestrian if your goal is to preserve the lives of the occupants who may die if you choose the retaining wall.
But all low speeds, you want to pick the retaining wall because the occupants wouldn't die - it would just damage the car.
Or in the OP example, picking the smaller car to crash into might increase survivability for the passenges of the autonomous car, but increase deaths for the smaller car being hit. Whereas hitting the larger car (more solidly built, more mass) might injure the occupants of the autonomous car more..
.. It's the "target" with the perceived least usefulness to society. There are so many in the 99%.=, and why would any of the 1% be on the streets anyway?
Hit the lawyer.
If Slashdot were chemistry it would look like this:Cadaverine
How about hitting the Brakes?
Crashes should be very unlikely when everyone is following the rules of the road. So if the car has to pick between the guy that cut it off or the car next to it, it picks the car that cut it off. Maintain the right of way if no other option is available without creating a crash.
Truck drives into an open market an kills 30 people. Truck driver answer to the judge: "there was a pregnant woman and an old man on the road, I had to pick which one to hit. I picked the old man". Judge asks, "but you drived into the open market and killed 30". Driver says, "old man ran into the market, I had no choice."
Since humanity is the cause of global warming, and car-driving humanity is particularly at fault, the only truly ethical choice is to always maximize the number of internal combustion-using humans killed (better to say "culled") in every circumstance.
All carbon-generating meatbags deserve to die. It's the only ethical choice for the planet as a whole.
There's nothing worse than seeing an animal suffer, even if that animal is considered by many to be vermin.
That's why the car ethics algorithm, besides heading into the cat, need to accelerate as well. :D
[Disclaimer: Although I dislike cats as a pet, I respect them as animals and would never (and have never) hurt them.)]
I recall a bit from some medical drama way back (St Elsewhere perhaps? All TV shows are just dreams - https://en.wikipedia.org/wiki/... ) where the doctor is sewing up some mean tough looking biker dude and asks what caused his injuries. He explains that he was driving his motorbike along when a cute furry animal darted across
the road (a squirrel, bunny, or cat; I don't recall) and the crash happened "when I swerved". "Did you hit the cat?" "No, he got away." Leaving the viewer to understand that the accident was caused by the rider trying to HIT, rather than avoid, the animal. The delivery made it pretty funny as I recall.
What if the crash were avoidable but the car just hit a cat (because the car went berserk for some unknown reason)? Who compensates the cat owner -- the driver or the car manufacturer?
Neither, that's what mandatory insurance is for.
That's exactly why there are the 3 laws of robotics!
"Stupid reasons" is arbitrary. If its not my train or track, its not my place to hit the switch, and if I didn't start the train, its not my homicide.
Basically the author is saying should vehicles go from a reactive state to a proactive state. All autonomous cars current are reactive in nature. The latter being a non-linear problem. We can solve it though a uber logic table, but I'm sure with all the filtering and choices, would be too slow that in the end is no different from an RNG,
Interesting this applies to all autonomous vehicles, whether land, sea or air based.
Insufficient data for meaningful response, creating impact event to procure answer.
Be predictable at all times - it will help the other drivers (either biological or not) make their best decisions to avoid disaster.
As a consequence, of course try to be on the right lane. If the car is on the wrong lane for some reason, try to return to the right lane or as a last resort leave the road.
As a consequence, should collision be unavoidable (i.e. it's impossible to leave the road because it's a tunnel) and the only choice is what to collide with, collide with the vehicle on the right lane. If you collide with a vehicle on the lane you should be, the guilt has yet to be determined -maybe the other driver braked all as a sudden and guilt is his/her. But if you switch lane (or keep on the wrong lane) and then collide, the blame will be on you for certain. Special case: if you are overtaking another vehicle on a two-way road and vehicles on the right lane don't allow you to return to it on time to avoid a front collision, collide with them. They are failing to fulfill their obligations.
As a general rule, forget about maths and how many children are there. You can't possibly make any reliable count anyway. But besides that, doing math on how many passengers are there on the other vehicles only leads to have irresponsible drivers cause accidents and then escape them because they have more "hostage" passengers on board... which in turn only leads to allow them to cause further accidents on the long run.
I know that this guidelines lead to, at the end, have regulations grow a morality of their own. Following them to the letter would lead to some kind of "vehicles on the wrong lane deserve to crash, vehicles on the right lane don't deserve to", and "passengers are held accountable of their choice for who to drive".
Maybe this time they're helpful demons!
"So, you admit that you programmed the defendant's vehicle to deliberately target my client's vehicle?"
End of case...
Mosquitoes eat fruits unless they're laying eggs, thats the only time they ever eat blood.
If multiple cars are going to be involved, and face with hitting either of two vehicles, it should read the license plates of both vehicles, check to ownership of both, run credit scores, financial histories, etc and kill the poorer less productive occupants. At least it will some good.
Under no circumstances should they get away with it, even if that means taking out 3 others!
what if the driver of the other car, that will survive by steering your car over the cliff, would become the father of the next Hitler? A car will never have enough data to make a "right" descision in such a situation.
Making highly unlikely assumptions -- such as that anyone is [an ancestor of] the next Hitler or the next Mother Teresa -- is never the ethical thing to do.
such an alogorith would mean assigning an individual (monetary or any dimensionless number - no difference) value to a human life. And then you've left the field of ethical behaviour quite a while ago.
Actually, assigning a finite monetary value to human life is the only ethical thing to do. For example: wider highway lanes are safer than narrow highway lanes, and if you place infinite value on a human life, it would make sense to make all highway lanes infinitely wide. Obviously, that's not possible. So how do you calculate the optimal width of a highway lane? You estimate the monetary value of the lives that would be lost at various lane widths. You add that cost to the other costs (e.g., construction and maintenence), and choose the width that has minimum total cost.
Any other method is truly a non-rigorous, "gut-feeling," sub-optimal approach to safe highway design. The practice would be controversial among people who can't wrap their heads around this, so highway designers generally don't court controversy by advertising the value they put on human lives, but they most certainly do it, and thank goodness they do.
That that is is that that that that is not is not.
No. Just, no.
I guess you would say "No, Just No" to falling on a grenade.
Falling on a grenade refers to the deliberate act of using one's body to cover a live time-fused hand grenade, absorbing the explosion and fragmentation in an effort to save the lives of others nearby. Since this is almost universally fatal, it is considered an especially conspicuous and selfless act of individual sacrifice in wartime; in United States military history, more citations for the Medal of Honor have been awarded for falling on grenades to save comrades than any other single act.
That that is is that that that that is not is not.
I've been hospitalised for intervening in an accident I wouldn't otherwise have been a part of (as a pedestrian rather than driver) because I thought I could stop a worse outcome. If I am willing to make that decision myself, then why should I refuse to buy a car that will act in the manner I would act myself?
Maybe the solution is a dial on the dashboard: turn the dial all the way to the left for "make incredibly selfish driving decisions," and all the way to the right for "make selfless driving decisions." Then everyone can make their car act in the manner they would act themselves. :)
That that is is that that that that is not is not.
Big giant spring under the car. If a collision is inevitable, eject the car from road. Problem solved...
Seriously, just make them out of Superball material. Now we just need to solve the problem of collisions lasting 45 minutes and 75 miles...
I've got better things to do tonight than die.
It won't make any attempt to swerve or select a target.
That would be a downgrade from human drivers. If a human driver has time to react, he or she will swerve and select a target. (Plowing through a flimsy picket fence beats plowing into a massive oak tree or a gaggle of pedestrians.)
That that is is that that that that is not is not.
when the decision who to hit comes up, the only way to be reasonably safe is to determine who's not following the rules and to hit that one.
So not only does a car have to try to calculate how to minimize fatalities; it also has to try to sense which of the other cars are spoofing the claimed number of occupants?
That would rather go against the above post that makes a great case for "keep it simple."
That that is is that that that that is not is not.
definitely, a cat, I hate them.
What have you got against Catalytic Converters?
Calling someone a "hater" only means you can not rationally rebut their argument.
do not ever choose to hit an unprotected human
I appreciate that you are trying to add nuance. But an even more nuanced answer is, always choose to hit an unprotected human, if the alternative is to hit a gaggle of nine unprotected humans.
That that is is that that that that is not is not.
Does not choose what it hits.
Simple, aim for the cars with religious symbols on them, they think they are going somewhere anyways. second priority is opposing political party bumper stickers.
Analyzing... vehicle A operated by Mrs. Jane Smith, 76, loving grandmother, recently diagnosed with malignant melanoma... actuarial data suggests a median life expectancy of three more years. Vehicle B operated by Mr. John Miller, 33, clean bill of health, history of domestic abuse... past conviction for vehicular manslaughter... actuarial data suggests a median of thirty more years in the workforce.
I don't think anyone has mentioned one of the major sources of input about this decision: car to car communications. In a collision situation, what is to prevent one vehicle to send (erroneous?) information to the vehicles around it to persuade them to collide with some other vehicle?
The game playing scenarios become ever more complex.
You could also fit way more of them on the roads, mitigating traffic jams. Most likely you would get to the work (and home) faster. But hey, screw that, we already have SUVs, they are the bomb! You see so well over other cars, untill everyone is forced to get a SUV. But I guess you can get even bigger cars by then!
This is just like all the robot movies where the robots choose to kill people in order to save others.
Imagine a scenario like this: I am driving my SUV with a passenger, and I am driving responsibly and carefully. Another guy is drunk in front of me and he, in his Pinto, changes lanes and inadvertently slams on the brake instead of the gas. This causes the autonomous car behind him to determine that a crash is going to occur. But instead of ramming the Pinto, it swerves into me and kills my passenger. The Pinto drives on home and the guy is never seen again (well, hopefully the autocar would have gotten a picture or something).
Now, the autocar hit me because it determined fatality was less likely. But my passenger died as a result instead of the drunk driver who caused the accident. Robots should not have the power to make that choice. It could just have well have turned itself the other way off a cliff, hoping its parachute would cause less damage than smashing into the drunk driver.
First of all, if human harm is avoidable the car should choose that path. Property doesn't matter.
Secondly, the car should prioritize the ones inside it. They are the paying customers. If the car doesn't do that you won't sell even one.
Thirdly, the car should go where it would have gone if it followed normal driving rules. If you have to pick to drive over two people that jumped on the road or one people standing on the side of the road you drive those two down and blame it on them.
Basically make it act as human as possible. (ok, a human might try to avoid hitting the the clowns that jumped on the road and accidentally kill the passer by, but that is only because the human wouldn't have had time to notice there IS someone standing there, and the two persons caught the driver by surprise.)
If the situation is assessed by a algorithm that actually has time to react instead of human it really CAN do stuff to minimize damage. Things like nailing the passengers down with airbags, nailing the passengers of the _other_ car down with airbags. Deploying collision airbags in front of the car. Flipping up airbrakes (think roof suddenly turning up. Would work wonders in higher speeds.) Cutting down gasoline feed preemptively to reduce the possibility of post crash fire. Unlocking the doors. Sending a message to 911 (with gps coordinates) automatically _before_ the inevitable crash actually happens while it still can. Anti lock brakes wouldn't be needed because the computer could just release the brakes when it wants to steer (ok, kinda the same thing, but when it doesn't want to steer it could just brake more efficiently)
And that insurance is paid by the driver -- so it's a small monthly fee instead of a settlement. Therefore, the car manufacturers should pay insurance periodically during a year. Why should the driver be liable for software/hardware bugs of the car?
Cats aren't pets. Cat "owners" are pets. That is why cat "owners" claim cats can't be taught not to defecate in other people's gardens or taught not to tear up garbage bags.
Luckily I like big dogs. Big dogs tend to keep cats out of gardens. They can be taught not to exit that garden (although it can be difficult).
Well, I might have a way, but it only works on a semi spherical planet in a vacuum.
Standard versions of cars will use networked algorithms to mutually agree on victims (factorizing in their age, gender and whether they smoke).
The network will use an optimization algorithm which will maximize the total expected future timelife of the survivors.
For the wealthy people, Ethics Supression Module (a car option) will guarantee higher priority.
In the US, the government will obviously intervene and mandate that the optimization criterion is the total expected future taxes collected from the survivors.
Nope. But that's MY decision.
The case we're discussing (your car "intervening in an accident") is more like me pushing YOU onto a grenade....
"I do not agree with what you say, but I will defend to the death your right to say it"
And that insurance is paid by the driver -- so it's a small monthly fee instead of a settlement. Therefore, the car manufacturers should pay insurance periodically during a year. Why should the driver be liable for software/hardware bugs of the car?
I don't see why it would not be handled the same way hardware defects are right now: the manufacturers hide the problem until the death toll is high enough or someone spills the beans, then the insurers and victims go after them in a class action lawsuit.
What's different is cars (or rather the software) will be making decisions, and someone has to pay if the consequence of those decisions are bad. Right now, humans make all the decisions -- the car simply goes or stops in the direction the human wants. Consequences (& liability) of bad decisions are greater than those caused by just some mechanical failure.
In essence, the driverless car has transformed cars into old style car + software driver while the old human driver has been transformed into a passenger. So the human is now taking a virtual taxi ride instead of driving his car. And if someone taking a taxicab ride were involved in an accident where his taxi is at fault, is he liable to pay the injured party? No, but the taxi driver and his company is. Similarly, a person using (not driving) an autonomous car should not be liable in case his car causes damage to something or someone, but the car manufacturer (owner of the decision making software) should be liable.
They're asking the wrong question and trying to solve the wrong problem. TCAS (Collision Avoidance) technology is already here for aircraft. The cars in front can keep enough space between them to allow an exit route and even tell the cars behind how fast to go to avoid running into them. Avoid the creation of the situation rather than try to figure out what to do when you didn't think and communicate ahead of time.
Failing all that, your car can tell the big one to take out the smaller car for you.
Ops, I shuld have usd the prevuwe but in.
The obvious answer is to hit whatever results in the least loss of life.
I would submit that cars like the Elio or like you describe are great for single people or people who already have long distances vehicles and don't have a need to carry much very often, or live super close to work. Similarly the reason most people don't is due to having to carry multiple passengers routinely and cargo (groceries, shopping, etc). Frequently the circumstance that requires you to carry something usually is not pre-planned. Goes something like, 'Hey, I can't do X so can you pick up Y for me while you're out/on the way home/etc'? I would love a car like the Elio but with kids in my near future, and the possibility of twins (runs in my wife's family) the car I got will be the car I keep for a long time and I will likely need its space...its odd going from a small sportscar (MX-5) to a 5 door sedan.
It's: "If you could stop a runaway train from going over a ravine, by pulling a lever, thus saving 300 people, but the lever sent the train down a different track on which 3 rapists were playing, what do you do?" I pull the lever.
If a collision is unavoidable, they'd just brake as hard as they can without skidding, and hope the other side can maneuver out of the collision.
This was foreseen before it was technically possible with the the Three Laws of Robotics.
Killing someone by inaction is definitely not murder. At least, not in the United States. There is no default duty to act to save someone. You might have a duty if you are a life guard or a scuba instructor or a police officer. But, by default, there is no duty to act.
An Olympic swimmer can stand on the beach and watch a little girl drown. It's not a crime. It's not right, but it's not murder.
I am working in Strong AI research and this type of system is one of the potential applications I have done work on.
The answer to the question is that the whole system will be designed to avoid this scenario ever happening, and if it does the AI can only make a best guess decision to minimise the loss of human life. The real main danger in a Strong AI machine is actually a hardware crash that disables the AI at high speed. - The solution spelled out below is to have an independent secondary emergency crash reaction system.
Generic answer : Even for an AI a crash is still a split second decision. To simplify the problem it will simply try to minimise the number of human lives lost, or make an ad-hoc decision.
In a more realistic Strong AI based type system the main AI will probably be a prediction based, reactive, resonant, real time control system, and will be quite slow and poor at acting in emergencies. - Prediction based means that the machine relies on a collapsing future probability state to coordinate everything it does and if this fails it takes quite a long time to regenerate or rebuild. This system will have a rigorous safety control envelope that should never voluntarily allow a crash position to happen in the first place, and if the car even enters a crash scenario the primary system has already failed.. In a crash the main Strong AI pushes an internal panic button and relinquishes control to a secondary AI. - A specialised high speed emergency crash reaction system.
This secondary system is mostly 'weak' AI instead of 'strong' AI, it has hardwired controls and works directly in terms of physics, and tries to minimise the net forces on the car as it brings it to a stop. The crash reaction vision system will be high speed and relatively primitive. The primary reaction is to apply the brakes and try to bring the car to a stop as quickly as possible without rolling over. Avoiding hitting things is secondary and far more complex, and because this AI is geared to high speed crashes it will prioritize by avoiding the highest speed impacts first.. This will protect the passengers and the car and other cars first, and then pedestrians and static objects - the machine will also try to avoid leaving the road. :- basic crash, pedestrian steps out - should be much better than human. :- high speed intercepting car - can do nothing. :- basic crash, distant pedestrian steps out, high speed intercepting car - should be much better than human. :- near pedestrian steps out, - can do nothing.
- Low speed / in cities
- Low speed / in cities
- High speed
- High speed
In the same time frame a normal human can usually only do one thing - reflexively hit the brakes.
Once developed this type of crash reaction system can also be fitted in cars without full AI control. (similar systems are already in development but are more primitive / don't use Strong AI.)
Below the speed of light Special Relativity is one of the most accurate theories in physics - above the speed of light..
Good points, but I expect there will not be a transfer of liability to the "taxi driver" unless you are actually hiring the ride from someone else. You bought it, you hit the start button, you end up "paying" and let your insurance company fight it out with the manufacturer when things go pear-shaped (what does that even mean?) Software is already making a huge number of decisions for you (when to shift, when to employ the air-bags, etc.) and when that is "at fault" it is your insurance company for the most part that takes care of laying the blame on the builder - I figure the same will be true here.
I suspect that overall we are only going to get to the autonomous vehicle stage when they are better than the "average" human driver by a factor of ten or more, so the cost of having to re-examine "who pays?" issues are probably going to be equally reduced.
It's a problem because insurance cares about who's fault a crash is. If it's your fault, your insurance has to pay up - and they may even refuse to if they can prove negligence, in which case you have to pay up. Bad news all round. If it's the other parties fault, then same story but for them.
BUT - if it's some computer programmer in Silicon Valley's fault.. well, now we're in a whole other area. Do computer programmers need insurance now? It's a whole big liability clusterfuck, and I think it's entirely possible that it might prevent fully-automatic car driving in a general urban type environment from ever taking off, no matter how smart it gets.
Mind you, I thought that about fully-armed drones blowing people up too, and it looks like I was wrong there...
If there is time to swerve then there is time to avoid crash!
Common sense would dictate that any number of cars colliding with each other will not have an equal but opposite force of impact for each driver involved. The most ethical solution I can fathom (not just in regard to computers making decisions) is the scenario with the total lesser kinetic impact to all humans involved and if current medical practices and procedures improve the odds of survival in a certain way with regards to the style of crash and the fragility of the human(s). the exception would be if a person's death can be avoided by compensation of another's non-fragility.If one of the humans in a calculated scenario dies as a result of the most definite argument reasonable (i.e. calculating physics, the fragility of multiple humans involved & medical practices and procedures that improve the odds of survival).... Death is wrong folks... May the mind of the victim be uploaded. amen.
Yes, but starting your car with a button does not prove negligence or malice which is a prerequisite for placing blame and liability cost on the driver in case of an accident.
If the software were to fail in any of these cases, the car maker will be sued as happened to Toyota with their cars doing unintended acceleration.
In this kind of situation (which humans are quite bad at antcipating) odds are good that the car will have noticed something odd ahead and slowed down to a stop well before a human would need to be making this kind of decision.
This includes pedestrians walking from gaps between cars. An AI is more likely to notice feet under the vehicles at the side of the road than a comparable human as it never stops looking for them. Similarly stuff coming form side roads is going to be seen/acted on while a human is still processing the input as "not normal". That's not even accounting for situations where several hazardous things are happening at once, which humans are notoriously bad at handling.
At least 90% of car crahes are down to driver error, usually due to inattention. I'd expect that the rate of crashes in automated vehicles would be substantially less than manually controlled in a very short period of time and that's for standalone systems. Once you factor in inter-vehicle communications it should drop further still.
When automated vehicles become common, you can substantially raise the bar on skills required to get;/retain a license.
This alone will make a big difference, as at least half the drivers on the road today are barely competent and even decent drivers have lapses in concentration at times (I don't think I'm any better than average and I'm looking forward to letting the car do most of the work.)
You will have to pay for damage caused by your vehicle. Even if the car is driving itself. Autonomous cars won't be viable if the drivers aren't taking responsibility. There will probably be some kind of contact that you'll have to agree to when buying the car. Seems fair to me. If an autonomous car is ten times less likely to get in an accident than a human driver, then your cost for insurance would be way down and the insurance would pay for the accidents caused by the autonomous cars. Accidents caused by unpredictable circumstances as well as software bugs would be covered by insurance.
Software is already making a huge number of decisions for you (when to shift, when to employ the air-bags, etc.)
If the software were to fail in any of these cases, the car maker will be sued as happened to Toyota with their cars doing unintended acceleration.
Yep, but not until enough insurance companies paid out and they realized it was an actual problem. I don't think we are really disagreeing. I don't think that autonomous cars change the landscape that much from what we currently have: people's insurance pay for accidents and when the problems are seen to be not the fault of the insured, the insurance people go ofter the responsible parties to recover costs. The same thing will (and does) happen for non-auto insurance when visitors get killed by your Roomba running amuck.
It's a problem because insurance cares about who's fault a crash is. If it's your fault, your insurance has to pay up - and they may even refuse to if they can prove negligence, in which case you have to pay up. Bad news all round. If it's the other parties fault, then same story but for them.
Just a nitpick, but I know of no automobile insurance policy that will refuse to pay up in the case of negligence - they even have to pay for DUI accidents. They just get to jack your rates up after because you've proven to be a higher risk. There wouldn't be much point of insurance for personal liability if it didn't cover negligence, because something negligent is nearly always found.
The only exception would be if the action was deliberate, and I think they even have to pay out for that - it's just that they can now sue YOU to recover the money.
I don't read AC A human right
I suspect that overall we are only going to get to the autonomous vehicle stage when they are better than the "average" human driver by a factor of ten or more, so the cost of having to re-examine "who pays?" issues are probably going to be equally reduced.
I've tended to use 'half' myself. We're at around 32k deaths/year from automobiles, down from nearly 55k back in '72.
If you prevent 16k deaths, it would save the country $7.6M in damages per death, $122B per year. Even if you consider that each death might cost the insurance companies* far less, maybe $500k between liability and life insurance payouts, that's $8B the insurance industry can 'save' if they can cut the accident rate. Disability insurance
Countering that is if the liability costs are low enough manufacturers might get into the business, baking the insurance cost into the price of the car.
*Remember that most insurance companies issue many forms of insurance, doing car, home, and life.
I don't read AC A human right
Would autonomous self driving cars decrease or increase the amount of automotive related deaths and accidents. If self driving cars reduce automotive accidents and deaths nation wide by 90% then these other ethical dilemmas are insignificant in comparison
If the autonomous car is doing its job, it will encounter an unavoidable crash very infrequently. All the programming effort should be poured into crash avoidance, not "crash optimization." Crashes happen way too fast for the computer to calculate actuarial tables for all the people around it. All the "swerve right" and "swerve left" in TFA's scenarios omit one big thing: What will the car hit if it just slams on the brakes and doesn't swerve at all? Ok, hit that. There, "crash optimization" done. Maybe if the car is good enough at detecting pedestrians then you could have it try to swerve away from the pedestrians, but in all other scenarios, swerving will just make things worse. Try to stop, and hit whatever you were going to hit anyway. That's all the car should be required to do.
Mixing Star Trek and Pokemon is a sin. Your Nerd Card has been revoked.
Confusing Issac Asimov's 'positronic' robot architecture with Star Trek is a sin. You probably think every reggae song is sung by Bob Marley.
you type like a scripted spam mailer
Thank you. Check out my other fine postings.
<blink>down the rabbit hole</blink>