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 | ... | 7 | 8 | 9 | 10 | 11 | 12 | 13

Mad Scientist
29760 posts

Uber Geek

Trusted

Lifetime subscriber

#1422601 6-Nov-2015 20:25

nunz: Any algorithm for an AI car should possibly use our human algorithm namely:

Can I avoid it? NOPE. Hit the brakes and see what happens.

No need to kill and until AI learn to pray random chance and blind luck should lead them as it does us.

Oh dear, ain't bode well if a 10 tonne 36 wheeler is heading your direction

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1422610 6-Nov-2015 21:04

LOL a fun thread!

Forget AI.

Keep it simple.

Imagine if the roads were like Scalextrix slot car tracks everyone obeying the rules, everyone in their own lane, everyone at the speed limit. Thats not AI, or ethical decisions, its coded instructions. Its not rocket science. The only thing omitted is our personal love of driving.

Batman

Mad Scientist
29760 posts

Uber Geek

Trusted

Lifetime subscriber

#1422615 6-Nov-2015 21:15

Yes try implementing that in India, Somalia, Sierra Leone ...

3rd world visitors to NZ -> sit in self drive cars.
1st world visitors (eg president of USA, Navy Seals, Pro Cycling group) to 3rd world ... can't drive (never learnt - a bit like grammar and spelling).

gzt

17104 posts

Uber Geek

Lifetime subscriber

#1422629 6-Nov-2015 21:54

nunz:
gzt: I think you are beginning to understand the difficulties with your ok to kill cyclists algorithm.

Personally the reason I find these 'jokes' in bad taste is because I know people who have been intentionally injured while commute cycling.

So to me some of this stuff is very similar to hate speech. Yeah you are free to say it, but I am also free to say that saying this stuff over and over really creates a social environment where some people feel supported to do very bad things.

and that would be me - Ive had more than my share of going over bonnets, bouncing off fenders and being cut off, tail gated, pushed out of lanes etc etc. - and yet I joke about it - Humour is how we handle the bad stuff in our lives. You should hear some of the suggestions I have made for dealing with cars as a cyclist and for dealing with cyclists as a car driver.

Humour is the politics of slaves and if some butt head uses it as an excuse to do bad stuff - then the likelyhood is they would do it anyway.

Without humour we become uptight, politically correct kill joys. and as we all know the politically correct are the ones you should NEVER make fun of - they have no sense of humour at all.

Well yeah I will often agree with that. The thing is the 'joking' becomes practical all too frequently and then it is funny only to one party. More than once I have been travelling with younger people and had to discourage this pointless and dangerous behaviour usually with appropriate swearing /endlecturemode

JWR

821 posts

Ultimate Geek

#1422652 6-Nov-2015 23:18

People don't make ethical decisions in an accident situation.

Conscious thought like that is far too slow. The 'real world' takes at least 1/2 second to process.

I say 'real world' because of course we don't see the real world. We just imagine/process it in our brains.

People act subliminally in accident situations.

If you have ever been in an accident situation, you should know what I mean. You will act without thinking and often be surprised afterwards.

I don't think we need ethical self-driving cars. We just need cars that don't get distracted, don't get angry, don't get drunk etc.. and react sooner and faster than we do.

bisr

227 posts

Master Geek

#1422684 7-Nov-2015 07:20

JWR: People don't make ethical decisions in an accident situation.

Conscious thought like that is far too slow. The 'real world' takes at least 1/2 second to process.

I say 'real world' because of course we don't see the real world. We just imagine/process it in our brains.

People act subliminally in accident situations.

If you have ever been in an accident situation, you should know what I mean. You will act without thinking and often be surprised afterwards.

I don't think we need ethical self-driving cars. We just need cars that don't get distracted, don't get angry, don't get drunk etc.. and react sooner and faster than we do.

I agree, and echo what somebody else said above. The answer is just 'hit the brakes', or bring the car to a stop safely as fast as possible. Don't overcomplicate it, that would actually be a stupid algorithm. If all the other cars are self-driving, then surely they are keeping a safe following distance from each other, unlike NZ drivers!

Rather than searching for the 'perfect mathematical solution' and force it, the answer is more testing in environments modelling these difficulties. See what the car does, see what learning occurs, rather than trying to enforce something top-down, try to train the neural net with different scenarios and see what happens.

It's not actually an ethical question we want the computer to consider. We want it to drive safely and predictably (and cleanly!), this will reduce accidents for everybody with all variables considered. If there is a panic condition, hit the brakes, come to a safe stop. That again becomes the 'predictable' behaviour which makes it easier for all other agents (human or computer) to accomodate for, including cyclists.

If an 18-wheeler is coming at you out of a blind corner and the computer 'knows' it could veer off and possibly safe the occupants by definitely kill a cyclist, then that is a bogus outcome. The answer for the car is to do nothing to endanger a third party, for whatever cause – just try and hit the brakes and come to a safe stop and hope for the best, not be a miracle worker. The driver of the 18-wheeler is guilty of vehicular manslaughter, not the self-driving car for not veering away in an attempt to possibly save the passengers but definitely kill/injure a third party. Sometimes it's your time, I would rather die than program a computer that considered that the occupants of the car were of more value than anyone else.

DizzyD

523 posts

Ultimate Geek
Inactive user

#1422692 7-Nov-2015 08:24

bisr:

I agree, and echo what somebody else said above. The answer is just 'hit the brakes', or bring the car to a stop safely as fast as possible. Don't overcomplicate it, that would actually be a stupid algorithm. If all the other cars are self-driving, then surely they are keeping a safe following distance from each other, unlike NZ drivers!

Can't say I agree

Will you be happy driving in a car that can take no action against a car careering into your lane from the other side (Think overseas, or drunk driver). Hitting the brakes will just cause a head-on collision with the other car. Swerving onto the footpath, or into the other lane while hitting the brakes is a far better option, and avoids a fatal accident.

Equinox IT

Cloud spending continues to surge globally, but most organisations haven’t made the changes necessary to maximise the value and cost-efficiency benefits of their cloud investments. Download the whitepaper From Overspend to Advantage now.

DizzyD

523 posts

Ultimate Geek
Inactive user

#1422719 7-Nov-2015 09:13

tdgeek: LOL a fun thread!

Forget AI.

Keep it simple.

Imagine if the roads were like Scalextrix slot car tracks everyone obeying the rules, everyone in their own lane, everyone at the speed limit. Thats not AI, or ethical decisions, its coded instructions. Its not rocket science. The only thing omitted is our personal love of driving.

Thats a problem right there. Cars need to be programmed to deal with the unexpected.

The odd person not obeying the rules, the odd car in my lane, the odd speed freak.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1422723 7-Nov-2015 09:24

DizzyD:
tdgeek: LOL a fun thread!

Forget AI.

Keep it simple.

Imagine if the roads were like Scalextrix slot car tracks everyone obeying the rules, everyone in their own lane, everyone at the speed limit. Thats not AI, or ethical decisions, its coded instructions. Its not rocket science. The only thing omitted is our personal love of driving.

Thats a problem right there. Cars need to be programmed to deal with the unexpected.

The odd person not obeying the rules, the odd car in my lane, the odd speed freak.

Yep it would. Sensors, closing speed, direction of impact, sensors decide what material it is, and algorithms decide the best cause of action. While there are many scenarios, the actuals are a limited range of speeds, direction and what the other vehicle is. Avoidance, breaking, traction control management etc all helps to avoid an impact, or reduce the damage.

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1422728 7-Nov-2015 09:37

DizzyD:
tdgeek: LOL a fun thread!

Forget AI.

Keep it simple.

Imagine if the roads were like Scalextrix slot car tracks everyone obeying the rules, everyone in their own lane, everyone at the speed limit. Thats not AI, or ethical decisions, its coded instructions. Its not rocket science. The only thing omitted is our personal love of driving.

Thats a problem right there. Cars need to be programmed to deal with the unexpected.

The odd person not obeying the rules, the odd car in my lane, the odd speed freak.

Expanding on that

I'd list al the possible scenarios that occur in the real world. 10, 20, 50 scenarios. The first step is to bring the physics in. Physics tell us that in any scenario, there are limited options available. The speed of the car, its inertia, tyres, etc, all these give a maximum range of braking and direction change that cannot be exceeded. Then, what is the best outcome? Program for that, thats all we can do, and that will well exceed what any human can do. As has been stated earlier in this thread about who to kill, its not about that, its about mitigation, what is the best choice to handle a situation, and doing that

Rikkitic

Awrrr
18657 posts

Uber Geek

Lifetime subscriber

#1422760 7-Nov-2015 10:49

Mitigation to bring about the least worst case result makes sense and is achievable, as long as there is acknowledgement of the Law of Unintended Consequences. Ethical decision-making, in cars or humans, is nonsense and should not even be considered.

Plesse igmore amd axxept applogies in adbance fir anu typos

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1422784 7-Nov-2015 11:57

Rikkitic: Mitigation to bring about the least worst case result makes sense and is achievable, as long as there is acknowledgement of the Law of Unintended Consequences. Ethical decision-making, in cars or humans, is nonsense and should not even be considered.

Exactly right. Use the available physics for the best outcome, not pick and choose who to injure/kill

SaltyNZ

8218 posts

Uber Geek

Trusted

2degrees

Lifetime subscriber

#1422788 7-Nov-2015 12:10

tdgeek: LOL a fun thread!

Forget AI.

Keep it simple.

Imagine if the roads were like Scalextrix slot car tracks everyone obeying the rules, everyone in their own lane, everyone at the speed limit. Thats not AI, or ethical decisions, its coded instructions. Its not rocket science. The only thing omitted is our personal love of driving.

That's a brilliant idea! Right up until some numbnuts town planner puts in an odd number of cross-over pieces, anyway.

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

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

SaltyNZ

8218 posts

Uber Geek

Trusted

2degrees

Lifetime subscriber

#1422792 7-Nov-2015 12:18

tdgeek:

Use the available physics for the best outcome

Yeah, see, you'd think that would be easy, right? We've had Newton's equations for about 400 years now. And we've been writing computer programs to simulate them for, oh, 50 years? But no, not necessarily. And those are for vastly simplified simulations where you can in principle know everything exactly, accurate to whatever precision you want!

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

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

tdgeek

29740 posts

Uber Geek

Trusted

Lifetime subscriber

#1422800 7-Nov-2015 12:41

SaltyNZ:
tdgeek:

Use the available physics for the best outcome

Yeah, see, you'd think that would be easy, right? We've had Newton's equations for about 400 years now. And we've been writing computer programs to simulate them for, oh, 50 years? But no, not necessarily. And those are for vastly simplified simulations where you can in principle know everything exactly, accurate to whatever precision you want!

Do you feel that humans can stop sooner or manage a change of direction to minimise damage better?

1 | ... | 7 | 8 | 9 | 10 | 11 | 12 | 13