The Trolley Problem - Mercedes has their own answer regarding self driving cars

[Copuis]

I dont know why they are looking at the car having the smarts, and not having the driver having overall say (thus is the one liable)

I mean we do that for pilots now right, they could have a computer taxi, take off, climb, level out, fly, land, taxi, all without actively controlling the plane beyond inputting the information, but they dont, they monitor it the whole time

I am all for "self driving" cars that mean the drivers are doing the boring part of driving, (keeping in the lane, maintaining speed, parking, etc) it take a load of, and removes for some drivers, something that is so hard they have to have their mouth open to carry it out

but I think that the driver should still be monitoring, and thus in charge of the car, ergo liable, as good as computers are, they are not going to sense that there is a big white truck trailer, or that there is a fire engine screaming up a side road (it might sense it after it is a little late to take a better action) facts are, as good as the tech is, it cant see everything, and the sensors have a limited range (they can see through car windows, and read the traffic beyond the car in front)
I think thats the way we should be looking at it, not some way to offload the responsibility of driving (and the actions of the car)

[VerticalDiscussions]

Anyone else here wants to kill the kid now xD?

 

I would shred him with a lawnmower for this heresy!

 

Well, at least computers dont have the desire to kill as that retarded kid does. Shoot him in the head now before he becomes a cereal killer...

 

 

[zMeul]

22 minutes ago, VerticalDiscussions said:

Anyone else here wants to kill the kid now xD?

 

I would shred him with a lawnmower for this heresy!

 

Well, at least computers dont have the desire to kill as that retarded kid does. Shoot him in the head now before he becomes a cereal killer...

[suicidalfranco]

i'd rather let a computer decide than a human, enough of this bullshit already. Let the superior being take the decisions 

http://www.theherald.com.au/story/4226873/trucks-car-in-pacific-highway-crash-photos/

[VerticalDiscussions]

2 hours ago, zMeul said:

-snip-

Can it be both :D? Well honestly, i dont remenber being a kid and ramming all of my human playmobile toys with a train XD. Nor putting animals in the middle of my portable track.

[tlink]

On 13-10-2016 at 9:58 PM, wrathoftheturkey said:

From a "mathematical and logical viewpoint," there is no answer if you start with the axiom that all lives are infinitely precious, since you can't compare two infinite values

actually yes you can, there are different sizes of infinity..  the aleph numbers are different sets of infinity.
 

[MoonSpot]

4 hours ago, VerticalDiscussions said:

... before he becomes a cereal killer...

Alternatively General Mills, Kellogg and Post could just hire better security.  It's not hard to defend yourself from this...

[techswede]

What i'm getting out of this is that i'm buying a Mercedes.

By the time self driving cars are a thing. (May that day never come) there will be so many safeguards in place that in the case where someone steps out in front of the self driving car in such a way that it cannot detect them and compensate accordingly in time.. They are at fault. 

The people in this thread that think that it is a problem for a self driving cars primary objective to be to protect the occupants are being ridiculous.  

 

If i know a car will choose to wrap itself around a tree to save bambi then i'm not fucking getting in one. 

[tlink]

34 minutes ago, techswede said:

What i'm getting out of this is that i'm buying a Mercedes.

By the time self driving cars are a thing. (May that day never come) there will be so many safeguards in place that in the case where someone steps out in front of the self driving car in such a way that it cannot detect them and compensate accordingly in time.. They are at fault. 

The people in this thread that think that it is a problem for a self driving cars primary objective to be to protect the occupants are being ridiculous.  

 

If i know a car will choose to wrap itself around a tree to save bambi then i'm not fucking getting in one. 

By that time it will detect the difference between bambi, a stroller, and a human. if the human is at fault make them pay for the damage they caused. the cars will be extremely safe by than anyways so there is very few risk for the driver to run into a tree or a pole.

[Bensemus]

On 10/14/2016 at 0:05 PM, zMeul said:

and who's blaming the car?!

ones to held responsible are the people building and programming them

AI is only as good as the people who programs it, and since humans aren't gods and can't think of each possible scenario with variables added to them, the AI can't efficiently assess the situation

People who were hurt. We always want to blame something and there is little satisfaction in blaming an inanimate object.

 

I've already pointed out that our first stab at making cars autonomous have been successful. Every single autonomous fleet has been safer than humans. They will only get safer as the software and hardware improves and as their percent of all cars increases as they are predictable and are replacing the potentially unpredictable humans who are easy to distract.

 

You don't have to tell the AI how to handle every situation. You give it rules and it applies those rules, exactly how humans solve problems. We don't know how to handle every situation. We apply our experience and knowledge to come up with the best solution. However we can only draw on our own knowledge and experiences. Fleets of autonomous cars will be able to draw on their collective knowledge and experience. Each car will be just as good as all the other cars and behave the same as they all share the same hivemind. Humans don't have this which means we potentially worse at handling situations, especially new drivers who have little to no experience to handle situations.

[Bensemus]

10 hours ago, Copuis said:

I mean we do that for pilots now right, they could have a computer taxi, take off, climb, level out, fly, land, taxi, all without actively controlling the plane beyond inputting the information, but they dont, they monitor it the whole time

Autopilot can't handle failures of the plane's systems though. We still need pilots to handle those extreme situations.

 

And before zMeul says I'm contradicting myself cars and planes operate in vastly different places with little parallel between the two.

[Copuis]

1 hour ago, Bensemus said:

Autopilot can't handle failures of the plane's systems though. We still need pilots to handle those extreme situations.

 

And before zMeul says I'm contradicting myself cars and planes operate in vastly different places with little parallel between the two.

 

neither can cars systems???, 

also, planes system have at the end of the day, a much much easier time of it, (because of them operating in vastly different places) as for the most part, they only have environmentally concerns for the most part (the distances between planes is mostly kept so large that it isn't a factor for automated systems)

cars have the above problems, plus, many many collision factors (other cars, debris on road, animals, people, etc)
you also have the sensor problem, a faulty sensor in a plane is often found to be faulty before there is an issue, where in a car, (and due to people acting like they can do everything but monitor the road, and car) a faulty sensor is going to result more likely in death (just ask that guys who's telsla couldn't see the big trailer) 
look at the reaction that happened post that
they all blamed the car

not the person

like I said
"I dont know why they are looking at the car having the smarts, and not having the driver having overall say (thus is the one liable)"
 

[BuckGup]

A self driving car should be smart enough that a situation like this doesn't happen, and if it does happen it will be the pedestrians fault for not following simple signs

[Copuis]

you're saying that a self driving car should be able to read all the road, yes

however there are limits to sensors, the mark one eyeball, and a good driver*, combined with some lane minding, and radar cruise makes for a better and safer journey (*good driver being alert to what is still happening around not glued to a phone or reading a book) than any amount of completely self driven cars

if you want to be able to read and use your phone, take a bus/train/taxi/uber, 
I mean really, you can drop an extra $30~$40k over the price of a normal car for self drive, or, hire a part time driver for three+ years 

I've not yet been to north america (where one of the big drivers of this tech is) but I can only imagine that there is a lack of good public transport options (or towns/city's set up for non car centred users)
there is a good interest here in Australia, for a number of reasons (imo, there is a really low bar to getting your licence, and drivers are pretty poorly mannered and lack awareness (I'm not only someone to have driven more miles than most a year, but involved in motorsports and driver training (car control) also a biker (motor)) europe has a mix of insane drivers, but all are vastly better for a number of reasons and the manners are there that dont seem to be on my local roads

self driving cars seem to be a way out for those drivers that either lack the skill, or "drive"/want/enjoyment of driving, it's a means of personal transport nothing more, and thats not right

if a plane/boat/train under control of an automated system crashes, the pilot/captain/driver is still liable, that should still apply to cars, why should they be treated any differently from other automated systems???????   

[hph6203]

The person at fault is the person that caused the incident. Theoretically speaking if one ship were on autopilot and the other wasn't, and the non-autopilot ship incorrectly cut in front of the ship on autopilot then the non-autopilot ship is at fault.

 

 

There are laws that govern our roads, and if those laws are followed then this situation shouldn't arise. Stop at stop signs, yield for pedestrians at crosswalks, stop at red lights, cross on green. If the car is properly programmed to follow these rules, then 99.999% of the time the pedestrian is at fault, and if there is no other safe alternative then the person who makes the error suffers the consequences.

 

I don't want my car or any car to run an algorithm to weigh the life of individuals, I want it to follow the rules and I want pedestrians to do the same.

[Copuis]

If the ship on autopilot is at fault, the captain is liable, not the autopilot 
and that is part of the difference seemingly applied to cars at the moment,
it the person leaves it to the tech, if the tech still to blame if the persons inaction (be that in the monitoring of the car, or stopping the car, or even understanding the limits of the tech) caused the incident??

look at the blame when the car died in the tesla, sure the system failed to see the trailer due to the colour of the sky and the colour of the trailer, and they were all looking at blame, was it the fault of the car, was it the fault of the trailer maker for a non standard colouring, was it a fault of the programmers
the conversation didn't seem to go, should the driver have been paying better attention, nope, the blame went elsewhere first, and thats not right (imo) tech shouldn't be used to remove responsibly, just to remove risks, thus shouldn't be blamed the dialogue isn't going in the right way

(i'm not sure if it is a culture thing, I mean I know the US love to blame everything and everyone before common sense is applied, but still)

also, the rules differ so much, and dont cover much of this at all, 

also, as for the pedestrians being at fault 99.999999% of the time if the car is programmed properly, thats not true at all, and discounts alot of other factors 
ie, travelling within the rules, a car might be travelling normally, and a childs ball, small, and not an issue for the car, bounces from between two parked cars, the road speed is 60kph,  the car may not slow, it will likely go "object small, moving, no risk dont apply brakes as the car following is close"
where a human, monitoring would go, "childs balls from between car, child might follow" and hopfully having reacted in time, might stop the death of a child

now if the worse happens, child runs out too late for the car to stop in time, and steering left and right are not options, the child dies, whos at fault, the car?, the driver?, the parents (the last thing to blame under the legal system) i would say the driver would be more to blame than the car, there isn't a way the car could have known, but the driver could have

 

[Bensemus]

On 10/15/2016 at 5:57 PM, Copuis said:

neither can cars systems???, 

also, planes system have at the end of the day, a much much easier time of it, (because of them operating in vastly different places) as for the most part, they only have environmentally concerns for the most part (the distances between planes is mostly kept so large that it isn't a factor for automated systems)

A plane has to handle any failures and keep flying. All the car has to do it pull over. Which one seems easier?

 

Quote

you also have the sensor problem, a faulty sensor in a plane is often found to be faulty before there is an issue, where in a car, (and due to people acting like they can do everything but monitor the road, and car) a faulty sensor is going to result more likely in death (just ask that guys who's telsla couldn't see the big trailer) 
look at the reaction that happened post that
they all blamed the car

Where are you getting that info? 20% of plane crashes are mechanical which would include sensors. (Pilots account for 50%). A faulty sensor on the plane can be quite bad as you have to keep flying until you get to an airport. A car can simply pull over.

 

The Tesla accident doesn't show anything. The car's sensors did fail but guess what? The driver was supposed to be paying attention too as it's clearly stated the cars aren't ready to work without human oversight. Despite that they have a better safety record than humans. 

 

On 10/15/2016 at 5:57 PM, Copuis said:

like I said
"I dont know why they are looking at the car having the smarts, and not having the driver having overall say (thus is the one liable)"

You have to separate current autonomous cars and the ideal final version. Right now the driver is at fault as they are overseeing the car. In the ideal final version there won't be pedals or a wheel so there won't be a human driver, just the car. In that scenario the human can't be at fault as they have no input. The car is the one at fault and as the car is an inanimate object the manufacturer bears the responsibility.

[Bensemus]

16 hours ago, Copuis said:

also, as for the pedestrians being at fault 99.999999% of the time if the car is programmed properly, thats not true at all, and discounts alot of other factors 
ie, travelling within the rules, a car might be travelling normally, and a childs ball, small, and not an issue for the car, bounces from between two parked cars, the road speed is 60kph, the car may not slow, it will likely go "object small, moving, no risk dont apply brakes as the car following is close"
where a human, monitoring would go, "childs balls from between car, child might follow" and hopfully having reacted in time, might stop the death of a child

So then program the car to act like the human.Computers can be programed to recognize objects. You could classify and group any small round objects and note that a kid may follow the object. However 60Km/h is quite fast to be going near any places kids are likely to play so I don't see that scenario coming up. All parks and elementary schools have a 30km/h speed limit zone starting at least a block away from them in Canada. residential areas have a 40km/h limit. If the speed limit is that high then most likely only older kids who know not to run onto the street are playing there.

 

Also the child would still be at fault. The laws don't change and say kids are allowed to run into streets. It's terrible when it does happen.

[AlTech]

On ‎10‎/‎14‎/‎2016 at 3:43 AM, ThinkWithPortals said:

I would have thought the best option would be the one that would result in the fewest casualties, so it should steer into the one person. It's the best option, from a mathematical and logical viewpoint.

How about the option of no casualties and fucking use your brakes when you see a damn person? Or does it turn you on to see your car hit somebody else?

[Copuis]

3 hours ago, Bensemus said:

A plane has to handle any failures and keep flying. All the car has to do it pull over. Which one seems easier?

 

Where are you getting that info? 20% of plane crashes are mechanical which would include sensors. (Pilots account for 50%). A faulty sensor on the plane can be quite bad as you have to keep flying until you get to an airport. A car can simply pull over.

 

The Tesla accident doesn't show anything. The car's sensors did fail but guess what? The driver was supposed to be paying attention too as it's clearly stated the cars aren't ready to work without human oversight. Despite that they have a better safety record than humans. 

 

You have to separate current autonomous cars and the ideal final version. Right now the driver is at fault as they are overseeing the car. In the ideal final version there won't be pedals or a wheel so there won't be a human driver, just the car. In that scenario the human can't be at fault as they have no input. The car is the one at fault and as the car is an inanimate object the manufacturer bears the responsibility.

where are you getting your information regarding that???? biggest cause of "crashes" has been pilots committing to landings that should have been go arounds 
and the difference between a faulty sensor in a car vs a plane is the training and awareness (discounting the two MH flights, and german wings, that had been the soul reason for lose of life in fact, and those landings were beyond what the computer systems (due to aircraft or airports system) ability (and pilots too)
 
a pilot is more likely to spot a faulty sensor (something that the computer cant always do, and when there has been a reliance on the computers over humans, well, ask air france how that went for them), and can divert no issue, (remember a faulty sensor that might mean the automation isn't fully functional, but the plane is full serviceable, just like a car)
 
vs a faulty sensor in a car, where the trend is to purely free up people to do whatever, meaning the computer can act on faulty sensors and well, thats not going to go well

as for the telsa accident, it in fact shows a lot
first off, the sensor wasn't faulty, it was working 100%, there was a limitation with the sensor, something that will NOT be able to be over come, it can be mitigated only
(and a BIG part of my point is that all these self driving cars tout no attention, but it is, and SHOULD be required)

the aim shouldn't be to have a pod that your able to use like a car, among cars on the roads filled with vehicles of differing tech and ages, there are things in place for that already, and I mentioned them already, 
if you're wanting that, there needs to be a another layer on the road, a pod lane or something


oh, and your understanding of the law of computers and traffic is flawed
you could program the car to see any small round thing, or every fluffy toy, or small pet dog, and have it come to a screaming stop in case a child runs out, and cause the car behind to plow into your car, but many round things aren't always balls, somethings it is a bit of trash, a lid from a jar, or a pet brown rabbit vs a wild brown rabbit, it is just too complex a database to have the stop dont stop things (and the larger the data, the more factors that can go wrong)

and if a child runs out on the street, following a ball, and you fail to stop, it is strictly speaking manslaughter, you actions (or in actions) lead without intent to the death of that child, there isn't some "well the child shouldn't have been there, they werent following the law" get out of jail free card (not that it would be likely that you'd be jailed, depends on the country) you might be able to say the parents are liable, as they should have been supervising, but you should have stopped when you saw the ball

(now add to that if you'd been in charge, but not in control of the car, (ie it was self driving) how fucked up mentally that would leave you??)

[Copuis]

On 15/10/2016 at 9:34 PM, Grinners said:

I don't even understand what you are saying or how you came to that conclusion from my post. 

 

If you refer to the part of my post that you cut out it might help you understand the point. 

 

" It doesn't have to be perfect, it just has to be better than a human, which it already is. "

but it isn't, sure in some aspects, I mean the car is less likely to get distracted, but thats about it

can the car maintain better spacing, and speed than most if not all humans, yes
can it read the road ahead and get into different lane because of the flashing lights a kay away?, or the crash that just happened, nope, not yet, or for a number of years yet

can the car brake harder and faster than MOST humans, yes but thats most, so it is better than a human, yeah, depending on where the bar is set, it also can only reacted to the surface changes ONCE they've happen,
can they avoid a crash better than humans, yes, no, and not enough to say it is over all safer

an alert driver of modest training is better than a tesla at driving outside the pristine (tesla autodrive fails to work correctly where I'm from as the road markings are not well maintained, and in poor/wet weather the system is unable to lane keep, something I have no issue with)

so, it is better than a human, yes, but only in the fact it isn't distracted, and only if things are just right
any other time, human wins, hands down, everytime 

 

[Grinners]

20 minutes ago, Copuis said:

but it isn't, sure in some aspects, I mean the car is less likely to get distracted, but thats about it

can the car maintain better spacing, and speed than most if not all humans, yes
can it read the road ahead and get into different lane because of the flashing lights a kay away?, or the crash that just happened, nope, not yet, or for a number of years yet

can the car brake harder and faster than MOST humans, yes but thats most, so it is better than a human, yeah, depending on where the bar is set, it also can only reacted to the surface changes ONCE they've happen,
can they avoid a crash better than humans, yes, no, and not enough to say it is over all safer

an alert driver of modest training is better than a tesla at driving outside the pristine (tesla autodrive fails to work correctly where I'm from as the road markings are not well maintained, and in poor/wet weather the system is unable to lane keep, something I have no issue with)

so, it is better than a human, yes, but only in the fact it isn't distracted, and only if things are just right
any other time, human wins, hands down, everytime 

 

I'm talking about fatalities per 100 million miles driven. 

[Copuis]

6 minutes ago, Grinners said:

I'm talking about fatalities per 100 million miles driven. 

flatly, not enough data outside of controlled environments to make that statement, and you'd be wrong under most metrics

add to that, from country to country there would be so many other factors at play it wouldn't be a fair thing (ie, in the states, sure, maybe, but what are the other factors, were they drunk etc etc etc, but then in a place like sweden, nope, humans, would win, hands down, then here, in cities, maybe, but then, urban, and rural, nope, for many many reasons and factors)