Should Self-Driving Cars be Programmed to Kill You for the Greater Good?

The vehicle with autonomous driving capability will have a licensed driver ultimately in control...anybody who can't stop for pedestrians in a crosswalk is going too fast....that's not to say that somebody couldn't run in front of your vehicle before you could stop, but they would be jaywalking or crossing against the light and the responsibility for the accident would be the law breaker....but with crash avoidance system and adherence to the speed limits, the car would respond quicker than you could and steer around the obstacle if it's at all possible....

http://www.dailymail.co.uk/video/news/video-1045721/Fords-crash-proof-car-pedestrian-avoidance-system.html

http://www.iihs.org/iihs/ratings/crash-avoidance-features

https://en.wikipedia.org/wiki/Collision_avoidance_system

A licensed driver will only be present in this transition phaze: later when the technology can be certified (standards are not available yet) they will drive without driver.

...at which point personal ownership becomes an obsolete concept.

One would just dial up an autotaxi using an appropriate app instead.

Arrrooooogah! Fewer vehicles on the road overall means an upsurge in £$, recycling the existing ones...

The decline in the requirement to have personal ownership of horses was a tragic event for millions of people. They found other things to own.

Yes, there was a huge opportunity missed in the mushrooms production market when New York City switched from horses and carriages to the "horse-less" carriage. Not to mention all the jobs lost from the International Union of Horse Hockey Removers.

It save a few tons of coal that would have used transporting enormous tonnage of horse manure from New York City to Carlsbad Caverns (among others). The thing is, changing over from horses did nothing to change the concentration of horse puckey in Washington, D.C.

I just love politicians, don't you? Everyone should own one! Why not? You are paying by the throat for them already...LOL

The self-driving cars should be educated in the martial arts. There should be the judo version, the karate version, and the gung fu version. One flips the other vehicle, one gives the other vehicle (pedestrian) a hard kick, and one simply slides past off to the side (but safely).

There could be a blend of utilitarianism and deontology whereby the computer assesses the social status of the pedestrians. And, in the event of a decision to eliminate the cars (owner?), it should be dependent on remaining payment due and done quickly, without pain.

Not a bad answer but it is not so unlikely.

These incidents happen every day.

an autonomous vehicle will always avoid pedestrians or any object, but will not run off the road. It will be far more successful, on average, than people, at avoiding killing people.

autonomous vehicle will always avoid pedestrians or any object, but will not run off the road.
While autonomous vehicles may be responsible for killing fewer humans, I don't think we can yet say they will ALWAYS avoid pedestrians or any object but will not run off the road.

The most important notice here is that the bus will never be speeding in a zone where there it is legal that pedestrians are crossing a road.

But the question is still valid: what should such a car do in case a crash in inevitable: a child runs over to greet it's mom is the best example.

First task is that the vihicle comes to a standstill in a safe way without creating risks for other persons than those involved in the incident.

Secondary the car will have a memory to show what happend, so the attending person can't claim the bus was speeding on the wrong side of the road.

But most important: the bus will have noticed the persons at the pedestrian crossing and will slow down to let them pass. Not assuming they will wait till he passed.

An automatic driving system can track all persons on the sidewalk and when there is a sign that one kid is missing he can assume that this kid is running in between cars parked on the side of the road towards the road. Better human drivers also do this.

to be short: the self driving vehicle will slow down after which the human driver behind the bus will overtake and hit the pedestrians.

Just stand on the road and all traffic comes to a standstill

Self driving cars are the nightmare for those who don't want to walk the line.

Every time we get another autonomous vehicle story we hint at the real crux of the problem that remains unsolved and ignored in every discussion I bring this up

When an autonomous vehicle gets into an accident (it will happen some day) that causes significant damage to people and/or property who will assume liability if no obvious violator exists? Will it be the owner of the vehicle? Will it be the programmers of the vehicle? Will it be the occupants of the vehicle for not overriding the autopilot? Will any human operated vehicle assume responsibility? Will the autonomous vehicle be automatically absolved of any liability because the machine never makes a mistake?

It depends: if the vehicle is liable for the damage the insurance of the vehicle will cover damage to 3th parties. Just as it is now. (This is already confirmed by the insurance federation in Belgium, no legal changes are necessary)

It is then for the insurance company to find out who's error it was and to claim the money they paid to them.

Google already explained that in case it is an error of the cars software they will pay for it. (and they will get a back up insurance, which will make sure the software is ok before insuring a series production)

So in theory nothing changed, unless fatalities occur. After which the investigation will find out who's error it exactly was.

Big chance that the car will even not leave the parking spot if the insurance fee is due.

who is at fault if a building burns down? The owner, the occupant, the arsonist, the jet that crashed into it, the weather, the builder, the insurance company? People sue, courts assign liability. States and financial institutions require owners to create safe operating processes and insurance. The sky is not falling Fred, it is just progress of an identical sort that has been taking place for centuries.

Autonomous vehicles will eventually carry only automatons that, incase of an accident, will run away from the vehicle, and at the first opportunity, report the vehicle stolen. This will require "no fault" and uninsured autonomous vehicle insurance.

When the proliferation of this kind of technology eliminates human control and human responsibilities, it eliminates human morality and dignity, too. So why worry about safety? When we are conditioned to become like farm animals, ethical considerations become irrelevant, and our deaths are no more consequential than the deaths of pork-, poultry-, or beef-bearing livestock. We (other than the elites who make the social-engineering decisions) will be reduced to our primitive, pre-conscious, pre-linguistic condition, and there will be no reason to worry about the deaths of the masses. Some elites are already on board with this kind of thinking. Democracy was a fun experiment, until the baser impulses of humanity took over, like a fascination with safety

"When the proliferation of this kind of technology eliminates human control and human responsibilities, it eliminates human morality and dignity, too. So why worry about safety? When we are conditioned to become like farm animals, ethical considerations become irrelevant, and our deaths are no more consequential than the deaths of pork-, poultry-, or beef-bearing livestock."

So, if this is the future you envision, will you choose to be Eloi or Morlock? We *DO* have the choice, each of us; let the machines do our thinking for us and fall under their control, or learn about the machines, learn to THINK and become the one who controls the machine, rather than the cattle controlled by it.

(For those who choose Morlock: Living underground and eating Eloi are both optional.)

Soylent Green is People!!!!!

The way I see it autonomous vehicles will have multiple video and other such monitoring systems on the working on all sides at once plus automatic recording of all actions and inputs making it fairly easy to determine what went wrong in the event of a accident.

Right now we human primary use two eyes that only see a limited range and distance that are backed up by a well known to be less than accurate memory system.

To me that means that if a person runs out in front of an automated vehicle they have been recorded accurately and indirectly from multiple sensory systems all at once.

On top of that more than likely at some point the primary sensory and systems of all near by automated vehicles will have some degree of linking which would allow any active automated vehicle the ability to get extra views and sensory data from any other nearby vehicle as well which would in a way give even more views of an accident to determine what exactly had happened and who should be at fault.

Right now I see this as being very similar to the now commonly used dash cams people use. How many YouTube videos are there now of insurance scammers that deliberately jumped out in front of a vehicle or such in order to claim the driver hit them and was at fault to make an easy buck that got foiled or worse because the vehicle or vehicles involved had active dash cams that recorded the whole incident.

Sure someone ran out in front of a automated bus and got ran over. The thing is the bus has recordings of them and where they were from multiple views and multiple sensory systems and every automated vehicle nearby does as well which presently very few of us humans have that level of observations and data to fall back on as proof as to who is really at fault.

Conky, patron saint of autonomous vehicles:

Similar to the guidance we provide in Defensive Driving Courses... your 'safe clearance distance' increases with speed and thus so should the sensor range on the vehicle.

Similarly, as self-driving vehicles develop so will the surrounding infrastructure and social awareness. Meaning, stop lights will emit a signal informing surrounding cars of its current and pending status and pedestrians will be more likely to look both ways knowing there is a likelihood that a self-driving vehicle with the 'driver' PMCS'ing the inside of their eye lids is on the road.

Finally, it will take the first instance of a vehicular manslaughter case due to an autonomous vehicle killing a pedestrian; finding the driver guilty, for 'most' people to never trust a computer (with a finite set of 'rules of engagement') to safely guide a vehicle through a dynamic and unpredictable environment.

I have a better scenario though (better in that it is more likely to occur): Rush hour, you're in a center lane, cars to the left, right, rear, and front. You're vehicle is in autonomous-mode and average speed of all traffic is 45 MPH. Since it is in autonomous-mode it is maintaining a 'proper' safe following distance based on the 3-Second-Rule.

45 MPH = 66 FPS = 198 Feet Assured Clearance Distance (That's roughly 13 car-lengths)

First, what happens when you leave 13 car-lengths in rush-hour traffic... or 'any' kind of traffic for that matter? Other cars fill the gap. This occurs so your car slows, the person behind you get's mad... can't pass because of cars on their side as well, so they start tail-gating you. A pick-up truck lane changes in front of you by roughly 2 car lengths and a plastic bag flies out of their cargo area towards your vehicle. The sensors just see an object... they can't determine the object's density.

Summary - traveling 45 MPH, cars all around, car behind you is tail-gating, and object 1' x 1' x unknown depth and density is now 10' from your windshield and closing fast... we know it's a plastic bag so we, as a skilled driver, make no changes. The autonomous sensor can't discern density... what does it do?

The answer is so obvious, and right out of "The Expendables" movie:

(1) 20 mike mike bay opens on the rear of your autonomous car, tailgater either backs off within 0.25 seconds (he won't make it)

(2)20 mike mike cannons (twin or quad options at dealership) open up on tailgater

(3)your auto-drive puts the brakes on hard, but tailgater is not feeling well at all by now, and cannot apply brakes. Fortunately, his car has the new Lyn-Dor Industries auto-stop driving mode that takes over within 0.1 sec, and applied full brakes as well. The tail-gating car still hits your rear bumper, but it has the new auto-repel feature that shoves the tail-gater back where he came from, perhaps even into yesterday.

(4) your auto-drive now realizes that it was indeed a plastic bag, speeds up to tail-gate the pickup-truck ahead of you, before anyone moves in to intervene in lane.

(5) once tracking distance is established the front-facing 20mm cannon bays open and open fire on the pickup while applying full braking once again to avoid hitting the flaming debris of the guy who cut you off in traffic.

Hey this is starting to sound more like L.A. and an episode of "Fast and Furious".

As long as I am able to crawl back into the electronics bay and start pulling cards, I guess it would be ok. Daisy, Daisy, give me..._{your annnswerrrrr......}

Isaac Asimov would have loved this!

... should it attempt to minimize the total harm in any given situation ...

.
What else?

.
Why would we want anything else?We could of course include the self driving cars in the internet of things and let someone else drive it. Someone else can make the decision if we are going to die or not.

.
Currently we allow drivers with minimal driving skills on the road so chances are we will have autonomous vehicles too.Will there be less death? No. Minimize total harm would always lead to more than usual harm until the problems with the so called algorithms are sorted out.

.
We can not sort human brains and we will make errors for the autonomous cars.
I'd say look what happens. But I would not use a vehicle given it would possibly opt to kill me instead of the neighbor which I try to run over!

That argument is so illogical I can hardly think of were to start. I'll start with the obvious, which you actually highlighted! Any approved autonomous vehicle will be superior in safety to most nincompoops on the road. Death rates will decline. People will still die, even some passengers, but it will be smaller

So these new cars will have death of the non-driver passengers by lethal injection to the cervical spinal cord? Count me out!

I would sooner opt for a drone-type car with a driver in India (supposedly paying attention to the road, and not the India edition of Playboy (or Playgirl as the case may be). Or even if the drone driver has to take a potty break in the middle of a congested work commute? IDK - at least the drone driver could alert the passenger/driver that he needs to wake up and fly, or wake up and drive. At least they could offer a service where a real driver "took the wheel" for you while you text, smoke, take a nap, or a pee bottle break.

F4F wildcat pilots in WWII on long missions (or after several missions in the same day) rigged up a poor man's auto-pilot: They made loops of inner tube and fastened them to the stick by stretching them equally and stringing them over protruding knobs, etc.

The result was more or less straight and level flight (supposing they knew what they were doing with the trim adjustments). The flight leader (a friend of mine from a multiplayer game, Aces High II), would yell at them over the radio, "Hey Mulkowski, wake up and fly!!!" when they started drifting out of formation. These fellows were way brave, and their fear of sharks was highly surpassing their fear of the Japanese A6m3 pilots they faced in combat. Never forget what they did for us all, the sacrifice, nor the great skills they acquired in navigating open ocean, and most of the time making it back to the little grey dot in the field of blue that was the carrier deck! Salute to all naval aviators everywhere!

In spite of the movies, we still do not have the ability to (1) use more than a limited capacity % of our brains, nor do we (2) have the ability to creat an AI that can think for itself, have existence (in the philosophical sense), be senient, or produce completely safe results.

In a safety-conscious country as ours, we may have in fact wake up and drive, wake up and realise that mass-transit in megalopolis is the best answer, and not simply coddle ourselves out of existence by not being able to face real hazards any more.

Life is hazardous, and none of us are making it out of here alive.

The problem people keep on forgetting is that neither the AI or human driver will have an accurate picture of all of the relevant information at the time that a decision must be made. A human driver will at least be risking their own life and commitments with their decision. The AI literally cannot care any less and has nothing to risk.

And, during those brief seconds when a danger is identified, most cognitive brain function stops and reflexes take over.

http://www.stanforddaily.com/2015/02/18/stanford-researchers-build-autonomous-cars-that-imitate-real-drivers/

Also, tunnel vision sets in and the worse part of the threat becomes the only focus and mitigating only that portion of the entire circumstance become priority... preservation of self.

It is generally rare that the natural instinct of a person is to take one for the team rather than preserve self.

Take PSD's for example... Personal Security Details... the natural reflex of a person upon hearing a gun shot or explosion is to take cover... PSD's on the other hand have to "Get Big" to shield the person they are charged with protecting. These actions go against our natural wiring.

How this relates to autonomous cars is totality of circumstance and what portion of that circumstance the car has influence over.

IMO, no autonomous vehicle should be programmed for evasive maneuvers... there are just too many variables that it can not possibly account for, road surface, weather conditions, reactions of things (pedestrians, other vehicles) outside of its scope of influence.

Rather, it should sound an alarm alerting the driver and begin a controlled stop while maintaining current direction of travel. Any action outside of that needs to be initiated by the driver.

There is mention of "Matured technology" and "approved systems". Who or what would make this determination?

How about we leave who the car protects up to the driver?

Let the driver set his or her preference for self preservation by selecting a dial that goes from 0 (maximize probability of the most survivors) to 100 (maximize the driver's survivability).

That's what we currently do as vehicle operators....even though it may not be a decision we consciously make upon getting into the car.

Could the dial read: "Highway to Heaven"---------------"Mad Max"?

I was thinking "ME"----------------------"THEM"

Several comments have been along the lines of how much better autonomous vehicles will be than humans at avoid loss of life (passengers or others). The commentors mention things like more sensors and such. While reading through this thread there are a few things that came to mind:

1) When deciding on purchasing a vehicle, most people will sacrifice safety for a lower price to some degree. I suspect over time autonomous cars will have the bare minimum number of sensors it needs to operate by the safety parameters it (the car) is programmed to follow.

If society's laws/regulations force maximum safety of all involved in an accident then manufacturers will provide a super-set of sensors (proximity, cameras, thermal, audible, etc.) making the vehicles too costly for average citizens.

2) It's all well and good talking about how computers can make quicker decisions than humans, and overall that may result in a huge reduction in traffic fatalities and injuries. But no matter how beneficial it is for the general public, someone will end up being dead that their relatives think should not be dead, if not for that computer. The lawyers will get involved and someone will be blamed (owner, persons in the car, manufacturer). The scenarios that have been put forth sound as if it's reasonable that the computer in the vehicle would be able to recognize impending danger and avoid it, thus not having to make a choice between death of a passenger and death of a pedestrian (or others).

However, it's not hard to imagine scenarios that where the bandwidth of the autonomous vehicle (total response of the vehicle, not just the computer) will be too slow to avoid running over a person...and it doesn't even have to be that person's fault (such as running out into traffic). Granted those situations may be quite small compared to the current number of accidents happening on a yearly basis, but it won't take many to attract the attention of the shark lawyers.

Then there is the actual case of faulty installed equipment, can anyone say "airbags?"

Maybe better would be driver warning systems, that if totally ignored mean the driver is incapacitated, and the vehicle goes into "Lo-Jack" mode, and shuts down after auto-steering to a safe shoulder, or exit, or parking lot.

I will just keep driving my 2001 Chevy Silverado, until her legs give out. Then I will see what else is out there consistent with my attitude/age, ability to learn or put up with, that BHO hasn't already had crushed.

When you go through all the remarks mentioned here you will understand that manufacturers will install protection systems in the cars to prevent non approved software to be installed.

It's not just the cars we have to worry about killings us now....but the auto manufacturing robots, too.

worker-killed-volkswagen-robot-accident

I love the last line of the article, "VW said the robot had not suffered a technical defect".

My first read of that line I mistakenly interpreted it as VW saying the robot had not suffered from killing the worker.

You cannot hope to install the three laws of robotics until sentience can resides in a robot.

Stupid human.

I''m not sure what's more frightening, having sentient robots, or relying on the 3 laws to protect us from them.

Watch the new TV series "HUMANS", talk about creepy.

I've seen commercials for it. Looks interesting. But I rarely watch TV so I doubt I will ever see it.

The "three laws" are mentioned, and are the main theme. All I know is, if your house maid starts staring at the moon, watch out!

Depends which moon she is staring at, you might need to watch out for the other women in the house.

To err is human, to fault is electric.

Would that be the new moon, or the full moon, please do clarify!?

I wish I had a robot that looked like me, and was completely respectful and obedient to my wife, so I could be free to get back out to my workshop and finish building my doomsday survival machines.

I don't believe they will be here soon, but they will be here.

I don't know enough to predict how soon they will be here, but considering how rapidly things are advancing (Japanese robots, autonomous cars, etc) I wouldn't be shocked if the time comes rather soon (before I'm pushing up daisies).

One intriguing and disturbing idea is that AI sentience happens by accident. The AI person(?) chooses to stay hidden from humanity at first.

I can envision three rowdy robots, drunk on 10W-30, on a wild autonomous joyride, careening off what is left of the human race. *Beep-Hic....Yahoo!*

While reading Thermography and Vibration Analysis Reports ^{(Robot Porn)}

Oh, no you didant! It's my charge plug and I will cycle it as fast as I want!