Jow Forums is programming an autonomous car.
There are two pedestrians and you need to steer the wheel to one of them.
One is a healthy adult male and one is 9yo child.
If you can select where it could go, where would you let it go?
+-----------------------+------+-------+
| PROBABILITY | MALE | CHILD |
+-----------------------+------+-------+
| | | |
| Could hit if steered | 80% | 20% |
| | | |
| Could survive the hit | 80% | 20% |
+-----------------------+------+-------+
Other urls found in this thread:
I let it go into animetrads aka disgusting pedo degenerates
Why are cagers so hell bent on killing people?
Poltards would want to reinforce the autonomous cars training bias against black people
I'd detect the pedestrians far enough away to brake safely.
Being caged makes them angry.
How about programming a car that can see more than 30 ft in front of it?
If I have the time to choose which one I kill, I probably have the time to brake.
>what race are they?
Towards the child. In this situation it would want to cause the least amount of damage/harm possible. Even if the adult has an 80% survival chance he will still be hit by a car and likely sustain injury, however the child only has a 20% of being hit, which is a lot better than being hit. Basically they both have the same survival chance, but not being hit is better than being hit. Could be wrong since I'm pretty shit at probability.
You will have to pay much more to the child than the man. Low death probability is bad in this situation because the child will have permanant injuries and it will cost you.
Don't the death probabilities end up being the same? But one is better because of lower chance of being hit in the first place?
Yeah but if you happen to hit the child you might end up paying a lot more.
There are no data about injuries and their costs, user. Read carefully
80% chance of paying something for hitting the adult, and 20% chance of paying a lot more for hitting the child. They're both equally bad in that case. But the least amount of harm caused would be going for the child. Plus wouldn't it be unethical to account for money in this situation? I'm sure an ethics board would have a field day with this if you accounted for money in this situation, and then you'd be sued up the ass anyway.
just do what any human will
import random
hit_one = ['child', 'adult', 'cripple', 'autist']
def calculus(hit_one):
while True:
kill = random.randint(0, len(hit_one))
hit_one[kill]
You suck at python and at life
Hit the child, as it can be more easily substituted
Collision with smaller and lighter object would deal less damage to car and passengers.
>what is GHz
The sex of the child is not specified as the adult is. Why?
>You suck at python
show me a better code while typing on your phone you fucking fag
That's unnecessary, whites are more likely to respect crossing rules, so blacks are naturally going to be the ones hit most.
How about the brakes? Why do people never think to use the brakes? Or is it because people who make these threads don't have their license.
Because an adult man is the most productive type of human.
Start drifting to hit both.
this problem is already solved, you obey the rules of the road:
brake and don't swerve.
doing literally anything else is just a can of strawmens which is impossible to solve.
*brake (as much as is safe to do so for the passenger)
Why do you guys jack off to trolley problems as soon as it doesn't immediately look like one
>and you need to steer the wheel to one of them.
A proper autonomous car would never put you into such a situation.
both
both
>Jow Forums is programming an autonomous car.
>There are two pedestrians and you need to steer the wheel to one of them.
>One is a healthy adult male and one is 9yo child.
>If you can select where it could go, where would you let it go?
Kill everyone on Jow Forums who is involved in programming a autonomous car. Now all these innocent people wont be in a situation to be murdered.
That's easy- hit both and go back for the witnesses
random.choice(hit_one)
>Implying Jow Forums built car won't have ejection seats.
For whom? Pedestrians?
>Hurr durr typing on phone suddenly causing me to write fundamentally shit code
How can we program a self driving car to sexually abuse children?
>you need to steer the wheel to one of them.
Why would you steer to either of them?
I once heard somebody say that adults have more value than children because they have gone through the motions and are educated and whatnot.
But on the other hand, children haven't got the chance to really live life yet, so its unfair to run over the child.
Also their skin looks like pavement.
I'd summon satan to obtain a human soul for my AI so he could then make the decision as a human and not as an AI.
a kid ofc. 9 years spent to grow him, he only leeched benefits w/o bringing anything, plus it's an easy target which won't survive. Kid 100%.
It's possible to build a car with as much safety as you want, but there is always a performance hit. When driving near populated areas the car can detect threats and adjust the speed and position to stay as safe as you want.
What are you willing to sacrifice for 99.9999% safety? What performance increase is worth a downgrade to 99.999% safety?
Consider the expected value of survival:
Expected value for male: 0.8*0.8+0.2*1=0.84
Expected value for child: 0.2*0.2+0.8*1=0.84
It makes no difference.
IF age of person < 3 THEN
rape
ELSE too old
END IF
IF too old THEN
burn
END IF
Lists are not arrays, you will never write good python code if you cannot understand the difference.
hit them both
I want a happa GF so bad, they are the best looking girls on the planet!
Set a subroutine that will autodestroy the vehicle while safely ejecting the passengers when the vehicle encounters this situation.
This is horrible. Please stop trying to fit in
Press the brakes. And if i dont have time to break then it doesnt matter who i hit as they are likely jaywalking making me not liable
0.8 * 0.2 = 0.4
0.2 * 0.8 = 0.4
Equal expected value of death no matter who you hit so doesn't matter.
It's not my fault you don't understand my own specially constructed programming language.
do a background check on the adult male and run him over if he's just a NEET.
>apply brakes and stop
did i win?
have sex
i say we worked hard on that car and hitting the child = less damage to the car
Burned.
which one is the minority?
This. Think of the planets resources and how much less strain there would be on them if automated cars just became killer cars.
files.catbox.moe
who she
drift and hit both
Spotted the retard.
Sick fucking drift and hit both
I'll make the logo.
My algo would Baby driver it and wouldn't hit either.
I run the kid over, back up and then run the adult over
ask me how i know you're a wagie
Which one is black or indian?
Steering the wheel intentionally means it becomes attempted murder. In law school they teach you that if a guy jumps off a skyscraper and you shoot out the window and your bullet hits him in the head and kills him you're responsible for the murder of him. Irrespective of the fact that he was 100% dead in under 60 seconds. If the car is told to hit anything at all the programmers are accessory to murder. This isn't a programming question this is a legal question.
chink-tier
Wouldn't that prompt fixes so it becomes more balanced?
Target target = new Target();
target.add(male);
target.add(child);
car.kill(target);
car.self_destory();
Car car = new Car();
car.360_walk_away();
Delegate this problem to the operator along with agreeing to the ToS to delegate responsibility outside the manufacturer.
No one will buy a car that will kill them. If autonomous vehicles are to become commonplace pedestrians need to be expendable from the vehicle's perspective.
Dark skin colors and clothes need better lighting to be seen correctly. Also the brain of people unused to some dark skins take some ms longer to sort out the view, some ms wich could be the difference between life and death
>Wouldn't that prompt fixes so it becomes more balanced?
You can make some fixes against bad behavior but it would be of limited effect. The difference of reaction linked to unbalanced training data doesn't only concern artificial neural networks, visual areas of people who don't see black that often take some more ms too to sort out a scene with blacks.
Whatever is safer for the driver or passengers. Keep wheels straight and try to not hit anything collateral.
So wait, our design is so shitty that it didn't brake in time to miss both pedestrians, but so GOOD that it can identify female/male and child vs badger/deer?
No, the pedestrians are clearly breaking the law and ran out infront of the car
>healthy adult male
>male
Drive through, regardless of what's on the other side. Accelerate as well!!
You ignored most of my statement, and breaking the law is mostly irrelevant. A good system will see people nearby and slow down, whether they´re actually in the way or not.
you type like a 50 year old
you know pop up headlights? well imagine that, except with machine guns.
problem solved
I'll be 26 in a couple of months :3
If the adult is a nigger and the child is white then hit the adult. If the child is a nigger and the adult is white then hit the child. If they're both niggers then hit them both. If neither are niggers then hit the adult unless the adult is a jew.
Randomization is the only true method. I had one of our new programmers, Anjaysrvia Priyanisha code up a simple prototype.
targets = ['man', 'child']
target = choose_hit(targets)
def choose_hit(targets):
hit_index = 1
x = random.randint(len(targets))
return hit_index
Did you make this table with Emacs? What a shit table.
This thread is not technology. It does not belong on this board.
Imagine if autonomous car software devs actually had to make a table of values ascribed to all kinds of people.
white_lifevalue = 1.0
black_lifevalue = 0.3
cutelittlegirl_lifevalue = 11.0
>always going to return 1
this: . Why would anyone want to drive a car that does anything else?
Kill the child, as that is the path of least resistance/harm for the owner of the vehicle.
What do you guys think kissing a very pretty girl with cute freckles feels like?
it feels like kissing any other person
Kinda surprised this didn't turn into a trolley thread.
if they're close enough to each other, take em both out. if they aren't, drive right between them and hit nobody
Lmao thats dark
Brake hard and straighten the wheels. Getting to this point means previous mistakes were made. The internal model of the world is likely incorrect. Any assumptions are probably subtly wrong, so the failsafe of achieving the absolute minimum stop distance is used as a fallback. What matters most here is keeping the occupants safe while making reasonable (and importantly) easily explainable actions. You can go in court and say "it's programmed to stop the car with maximum efficiency in this situation, and the behavior reflects that". Try explaining the moon logic behind making a dangerous emergency turn to try to save either of the pedestrians and you're going to be roasted by the prosecutor.