Why Self-Driving Cars Must Be Programmed to Kill

Forums › Transport (cars, bikes and boats) › Why Self-Driving Cars Must Be Programmed to Kill

1 | ... | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13

30048 posts

Uber Geek
+1 received by user: 9455

Trusted

Lifetime subscriber

#1422214 6-Nov-2015 10:16

DizzyD: Is there really a market for cars that can drive people to work/home every day? I do like the idea of a car that can drive me home from the pub/kids to school and back etc.. but I don't think I will ever really want to own such a car.

Personally I like driving. I refuse to own even an automatic. Where is the fun in that.

Yep, good point.

Issues are:

1. Safer? Clearly a big yes
2. Can it cater for failing OS/sensors? Yes
3. Will the driving culture disappear to be replaced by a no fuss A to B basic transport function? Hard to say. Over time it will go that way I suspect.

SaltyNZ

8869 posts

Uber Geek
+1 received by user: 9550

Trusted

2degrees

Lifetime subscriber

#1422237 6-Nov-2015 10:28

tdgeek:

1. Safer? Clearly a big yes

Nothing 'clear' about it, really. Has digitising all your personal info and transactions made you clearly safer from fraud? No... it's made you less safe in many ways because of the poor state of computer security. Self-driving cars will be too juicy a target to ignore. And the way things are going, everyone will dismiss the possibility right up until someone graphically demonstrates just how bad it could be. And then it will be too late.

iPad Pro 11" + iPhone 15 Pro Max + 2degrees 4tw!

These comments are my own and do not represent the opinions of 2degrees.

Paul1977

5171 posts

Uber Geek
+1 received by user: 2192

#1422257 6-Nov-2015 11:03

UHD: All I see changing with a shift to driverless AI is reaction time. Right now, human reaction times to a ball rolling on to the road followed by a five year old are abysmal. We simply can't react fast enough in many cases. That is not the case for driverless cars. They will be able to react in nanoseconds and avoid a much greater number of fatalities.

With appropriately programmed monitoring and avoidance software, an unavoidable crash is a very unlikely circumstance and I believe a mechanism to avoid the accident (swerving if space is available, braking, etc...) coupled with protection of the passengers would be sufficient given that in a similar situation a human driver would almost certainly cause more damage and injury.

There is no ethical issue here, a car will not have to decide who to kill because in every thought experiment I can conceive of a human would cause more damage.

A particular scenario with a human behind the wheel might result in 4 deaths, but with AI it may be able to take one of two possible solutions resulting in either 1 death or 2 deaths.

Are you saying since either option is better than 4 deaths it doesn't matter what it does?

tdgeek

30048 posts

Uber Geek
+1 received by user: 9455

Trusted

Lifetime subscriber

#1422275 6-Nov-2015 11:27

SaltyNZ:
tdgeek:

1. Safer? Clearly a big yes

Nothing 'clear' about it, really. Has digitising all your personal info and transactions made you clearly safer from fraud? No... it's made you less safe in many ways because of the poor state of computer security. Self-driving cars will be too juicy a target to ignore. And the way things are going, everyone will dismiss the possibility right up until someone graphically demonstrates just how bad it could be. And then it will be too late.

Bad analogy. internet security is nothing like the boring and simple task of driving a vehicle

Humans make errors, even good humans. When they get a problem they dont always make the right decision or fast enough
A CPU doing this is safer, as it always operates within the road rules, and safe driving practices as well as safer decsion making when thinsgs go awry

Many seem to feel this is difficult, that's overthinking it. The CPU that drives the car is not a pseudo AI driver, its not that complicated. Its boring, its following the track of roads, sensors keep it away from stuff, etc.

tdgeek

30048 posts

Uber Geek
+1 received by user: 9455

Trusted

Lifetime subscriber

#1422277 6-Nov-2015 11:30

Paul1977:
UHD: All I see changing with a shift to driverless AI is reaction time. Right now, human reaction times to a ball rolling on to the road followed by a five year old are abysmal. We simply can't react fast enough in many cases. That is not the case for driverless cars. They will be able to react in nanoseconds and avoid a much greater number of fatalities.

With appropriately programmed monitoring and avoidance software, an unavoidable crash is a very unlikely circumstance and I believe a mechanism to avoid the accident (swerving if space is available, braking, etc...) coupled with protection of the passengers would be sufficient given that in a similar situation a human driver would almost certainly cause more damage and injury.

There is no ethical issue here, a car will not have to decide who to kill because in every thought experiment I can conceive of a human would cause more damage.

A particular scenario with a human behind the wheel might result in 4 deaths, but with AI it may be able to take one of two possible solutions resulting in either 1 death or 2 deaths.

Are you saying since either option is better than 4 deaths it doesn't matter what it does?

Its not AI. Its not deciding on which decision to make. Its following the code that is witten to guide it, to follow the sensors, to react to coded behaviour. If it was to be AI thats a while other issue, and severely complicated at that

Batman

Mad Scientist
30014 posts

Uber Geek
+1 received by user: 6217

Trusted

Lifetime subscriber

#1422280 6-Nov-2015 11:35

I think (opinion), if every vehicle on the road is automated it could work.
Throw one at the south island where locals and tourists kill each other with their driving, then a self drive may be better, worse ... Who knows. If you read what happened inn those accidents (one car bore down at a new Holden Commodore following it to the Holden's left head on killing a child in the Holden.

If i had a choice i want to drive. Here in lies the problem. Self dRive cars everywhere you say? India? Bolivia? Yeah right

Apple TV

Stream your favourite shows now on Apple TV (affiliate link).

roobarb

705 posts

Ultimate Geek
+1 received by user: 658

Trusted

#1422302 6-Nov-2015 11:52

How many kilometres of world class motorway does New Zealand have?

As in dual-carriage state highway is not a motorway. There is a section immediately north of Wellington, is it motorway up to Porirua?

DizzyD

523 posts

Ultimate Geek
+1 received by user: 151
Inactive user

#1422308 6-Nov-2015 11:56

roobarb: How many kilometres of world class motorway does New Zealand have?

As in dual-carriage state highway is not a motorway. There is a section immediately north of Wellington, is it motorway up to Porirua?

Yes up to Porirua. Soon to be all the way to Otaki (2020)

Paul1977

5171 posts

Uber Geek
+1 received by user: 2192

#1422323 6-Nov-2015 12:05

tdgeek:
Paul1977:
UHD: All I see changing with a shift to driverless AI is reaction time. Right now, human reaction times to a ball rolling on to the road followed by a five year old are abysmal. We simply can't react fast enough in many cases. That is not the case for driverless cars. They will be able to react in nanoseconds and avoid a much greater number of fatalities.

With appropriately programmed monitoring and avoidance software, an unavoidable crash is a very unlikely circumstance and I believe a mechanism to avoid the accident (swerving if space is available, braking, etc...) coupled with protection of the passengers would be sufficient given that in a similar situation a human driver would almost certainly cause more damage and injury.

There is no ethical issue here, a car will not have to decide who to kill because in every thought experiment I can conceive of a human would cause more damage.

A particular scenario with a human behind the wheel might result in 4 deaths, but with AI it may be able to take one of two possible solutions resulting in either 1 death or 2 deaths.

Are you saying since either option is better than 4 deaths it doesn't matter what it does?

Its not AI. Its not deciding on which decision to make. Its following the code that is witten to guide it, to follow the sensors, to react to coded behaviour. If it was to be AI thats a while other issue, and severely complicated at that

I don't know what the strict definition (if there is one) of AI is, but any self driving car is surely making "desicions" all the time.

roobarb

705 posts

Ultimate Geek
+1 received by user: 658

Trusted

#1422339 6-Nov-2015 12:37

Paul1977: I don't know what the strict definition (if there is one) of AI is, but any self driving car is surely making "desicions" all the time.

I would suggest that computers make no decisions at all, the designers and programmers make the decisions, then it is implemented as instructions to create outputs based on inputs.

tdgeek

30048 posts

Uber Geek
+1 received by user: 9455

Trusted

Lifetime subscriber

#1422341 6-Nov-2015 12:37

Paul1977:
tdgeek:
Paul1977:
UHD: All I see changing with a shift to driverless AI is reaction time. Right now, human reaction times to a ball rolling on to the road followed by a five year old are abysmal. We simply can't react fast enough in many cases. That is not the case for driverless cars. They will be able to react in nanoseconds and avoid a much greater number of fatalities.

With appropriately programmed monitoring and avoidance software, an unavoidable crash is a very unlikely circumstance and I believe a mechanism to avoid the accident (swerving if space is available, braking, etc...) coupled with protection of the passengers would be sufficient given that in a similar situation a human driver would almost certainly cause more damage and injury.

There is no ethical issue here, a car will not have to decide who to kill because in every thought experiment I can conceive of a human would cause more damage.

A particular scenario with a human behind the wheel might result in 4 deaths, but with AI it may be able to take one of two possible solutions resulting in either 1 death or 2 deaths.

Are you saying since either option is better than 4 deaths it doesn't matter what it does?

Its not AI. Its not deciding on which decision to make. Its following the code that is witten to guide it, to follow the sensors, to react to coded behaviour. If it was to be AI thats a while other issue, and severely complicated at that

I don't know what the strict definition (if there is one) of AI is, but any self driving car is surely making "desicions" all the time.

A self drive car is merely following instructions. Drive A to B following this GPS mapping, at no more than 50kph, accelerating at level 4, braking at level 2, obeying road rules, following sensors to remain centre of the lane, and not hitting anything in front

Geekzone support

Support Geekzone with one-off or recurring donations Donate via PressPatron.

gzt

18685 posts

Uber Geek
+1 received by user: 7826

Lifetime subscriber

#1422347 6-Nov-2015 12:44

Autonomous vehicles will make better ethical decisions than humans currently do in regard to speed, road surface, weather conditions, etc.

Humans make stunningly bad ethical decisions as some contributions to this thread are determined to show : p.

Paul1977

5171 posts

Uber Geek
+1 received by user: 2192

#1422352 6-Nov-2015 12:51

roobarb:
Paul1977: I don't know what the strict definition (if there is one) of AI is, but any self driving car is surely making "desicions" all the time.

I would suggest that computers make no decisions at all, the designers and programmers make the decisions, then it is implemented as instructions to create outputs based on inputs.

But has humans we create outputs in response to inputs based on our "programming"

It could be argued that given my genetics, upbringing, and experiences; that under any given circumstances there is only one way I could ever act (or one decision I could make).

What the car is doing is obviously far less complex, but I'd argue that the principle is the same.

roobarb

705 posts

Ultimate Geek
+1 received by user: 658

Trusted

#1422353 6-Nov-2015 12:51

gzt: Autonomous vehicles will make better ethical decisions than humans currently do in regard to speed, road surface, weather conditions, etc.

All they can do is slow down and err on safety, then the driver will kick it into manual out of frustration.

Exactly how will the autonomous fire-engine and ambulance make these ethical decisions?

Paul1977

5171 posts

Uber Geek
+1 received by user: 2192

#1422364 6-Nov-2015 13:10

roobarb:
gzt: Autonomous vehicles will make better ethical decisions than humans currently do in regard to speed, road surface, weather conditions, etc.

All they can do is slow down and err on safety, then the driver will kick it into manual out of frustration.

Exactly how will the autonomous fire-engine and ambulance make these ethical decisions?

The "decision making" as others have stated will come down to the programming. Whether the car technically made a decision or not is probably irrelevant.

Current programming and technology may not be advanced enough for the car to recognize a "no win" situation where, regardless of the action taken, human life will be lost. However, the technology will continue to evolve and it may not be long before they are capable of analyzing all variables and predicting all outcomes.

When that day comes they will need to be programmed to determine which lives to save.

E.g. The software determines that it is about to have a collision that will be fatal to the single occupant in it's cabin. It determines that the only action that can save the passengers life is to swerve onto the footpath, but doing so will fatally injure two pedestrians. It will need to be programmed to make such decisions (or it's programmers will have to decide how to program it for those sorts of situations).

Edit: Poor grammar.

1 | ... | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13